Abstract

The Finnish Geospatial Research Institute hyperspectral LiDAR (FGI HSL) was one of the first multichannel terrestrial LiDARs capable of producing simultaneous 3-dimensional topography with spectral data. Supercontinuum-based HSL instruments developed so far have suffered from portability and sensitivity issues, severely restricting potential applications. Recently, we have implemented a new robust field design of the FGI HSL together with an improved pulse digitizing scheme. Small size and significantly improved measuring accuracy of this new system enable a range of novel applications that so far have been impractical for multichannel terrestrial LiDARs. Particularly, this new design has enabled us to perform measurements in underground mines and detect minute spectral differences in various rock types. In this paper, we present the design of our new LiDAR and preliminary algorithms together with a brief performance assessment of the device. In addition, we provide example measurements of typical rock samples found in a ferrochrome mine.

© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

1. Introduction

Emergence of hyperspectral measuring technology within the past two decades has enabled a myriad of new applications. For example, chlorophyll content of vegetation can be successfully determined using visual spectrum [1]. Following the publication of a two wavelength LiDAR for vegetation analysis by [2], the interest in multispectral 3-dimensional measurements has also increased. Multispectral point clouds enable analyzing multichannel laser return intensity. In addition to laser backscatter intensity, the pulse waveform has also been analysed to retrieve structural parameters together with spectral data using different laser illumination, including tunable [3], multiple lasers [4, 5] and supercontinuum lasers [6–8]. Scanning LiDAR data products, i.e. full 3-dimensional data in the form of point clouds have so far been presented in two wavelengths in [4] and [5], while an eight-channel implementation presented in [6] is so far the only one based on supercontinuum laser technology. There are other supercontinuum based applications with even larger number of channels (such as [7,8]), but they have so far focused mainly on vegetation spectral indices rather than point clouds.

Despite steady advances and increasing use of hyperspectral techniques, there still remains a large number of uncharted application areas that are yet to be studied. Some of these applications require improvements and further development of the hyperspectral instruments themselves, particularly the retrieval and calibration of spectral data. One particularly challenging application can be found in mining industry, where hyperspectral techniques could be utilized for mineral and ore detection in rocks [9]. Quantitative retrieval of small differences in reflectance values has not yet been possible from intensity values retrieved with LiDAR instruments. This is partially because of the error values in currently available instruments and the complexity of the LiDAR radiometric response in general [10], but also because the sampling of laser pulses has not been optimized for intensity retrieval. The digitizing speed in full waveform laser scanners has so far been typically 1–2 GHz [11]. The deconvolution and sampling of LMS-Q560 waveforms was extensively studied in [12]. The challenges and benefits of full waveform digitizing TLS is also discussed from the manufacturer’s point of view in [13], but overall, limited information is available on the optimal sampling rate for systems aiming for both accurate timing and reflectance measurements.

We have recently started experimenting with rock ore content measurements using the Finnish Geospatial Research Institute hyperspectral LiDAR (FGI HSL). First experiments were carried out with our previously built HSL prototype developed in 2012 and utilized in vegetation measurements [6]. After a variety of tests, the sensitivity and measuring accuracy of this prototype was found insufficient for ore content estimation in rocks. The early prototype had a sampling speed of 1 GHz, which was used to record the roughly 1 ns FWHM (full width at half maximum) laser pulse waveforms. As a result, estimating the pulse characteristics was inaccurate, as only a few samples per laser pulse could be captured. This problem was exacerbated by the very low reflectivity of the tested rock samples. Inaccuracies due to the low sampling speed were addressed using pulse averaging, where a number of consecutive laser pulses were summed together to improve noise characteristics and increase the number of pulse sample points. If the device was moving during consecutive pulses, this technique would average not only noise and algorithmic inaccuracies, but also the sample space. For detecting minute differences in the sample space, either an accurate single pulse measuring or a quasi-static measuring setup was needed. The early prototype was also mainly designed for laboratory environment, which limited its field applications. Due to the aforementioned reasons, a total redesign of the instrument was implemented with emphasis put on improving the sensitivity and measuring accuracy. Further aims were to make the system compact for easy field operation and decrease the high power consumption in order to enable prolonged operation from battery. Major effort was invested in finding suitable digitization solution that would be compact, low power and fast enough. Also, a considerable effort was put to the miniaturization of optics and electronics. Together, these efforts enabled us to develop a relatively compact and significantly more accurate unit, which enables the retrieval of 3D hyperspectral features that have not been available so far. This paper presents the first examples of such data related to ore detection in rocks. Our new experiments on spectral and distance measurements also provide information on the limitations and challenges related to the operation of multispectral LiDAR in general, which has not been available thus far.

This paper is divided as follows: In Section 2, we will describe the optical, mechanical and electrical design of the instrument. In Section 3, we will provide background on the waveform processing and discuss the accuracy of the instrument. In Section 4, we will provide preliminary results of measuring typical rock samples from a mine and discuss the challenges of such measurements. We also discuss various aspects of the instrument and a number of future improvements that could be implemented. Finally, we draw conclusions in Section 5 by summarizing our research.

2. Design of the instrument

The new portable instrument in its operational state is depicted in Fig. 1. The instrument consists of a base structure and a pan-and-tilt-platform attached on top of this base. The base structure dimensions are 40×30×10 cm and it houses all the electronics and some other components, such as spectrograph, laser source and a photodiode sensor. The pan-and-tilt-platform performs scanning of the target surface and consists of two rotation stages and a custom aluminum housing containing optical components for transmitting and collecting the laser pulses. The total height of the unit is roughly 25 cm. In addition, two optical fibers are routed externally from the pan-and-tilt-platform to the base. The total weight of the system is 15 kg.

 figure: Fig. 1

Fig. 1 New portable and improved version of FGI hyperspectral LiDAR. An enclosure housing all the electronics forms the base of the unit. On top of this base is attached a 2-axis pan-and-tilt-type gimbal, which performs the 2-dimensional scanning. Light transmitting and collection optics are fitted in a custom housing attached to this gimbal.

Download Full Size | PPT Slide | PDF

2.1. Optical design

Optical principle of the instrument parallels the design of the previous prototype explained in [6]. In essence, a broadband laser pulse is sent to the target and a portion of the reflected signal is collected using 76.2 mm, 90° off-axis parabolic silver mirror. The parabolic mirror focuses the received pulses to a collecting fiber which is consequently routed to a spectrograph. Spectral output from the spectrograph is sensed using a linear avalanche photodiode array. Current component selection and the overall design of the system limit the usable wavelength measurement range to roughly 450–1000 nm. This wavelength range is divided by the spectrograph and photosensor combination so that we have measurement channels of roughly 50 nm wide. The main optical components and the setup are illustrated in Fig. 2.

 figure: Fig. 2

Fig. 2 Optical components and topology of the instrument.

Download Full Size | PPT Slide | PDF

We selected Leukos SCM-30 OEM supercontinuum laser as the source due to its compact size and low weight. The generated pulses are roughly 1 ns FWHM with selectable pulse frequency up to a roughly 25 kHz. Output from the supercontinuum laser is through a fiber and is first collimated and then passed through a beam sampler splitting the laser pulses with a 90/10 ratio. The parabolic mirror has two custom 3 mm holes drilled in such a way that light can pass through the mirror, coaxial to both, the collimating and focal axes. The 90% portion of the pulse passes through the hole coaxial to the collimating axis of the parabolic mirror and progresses then freely to the target. The 10% portion is used as a per pulse power level reference and as a trigger for the digitization. These trigger pulses are routed by two 90° silver coated 5 mm mirrors so that the pulses can pass through the hole coaxial to the focal axis of the parabolic mirror. The reflections from the target surface are focused by the parabolic mirror on to a circular-to-linear fiber bundle consisting of 7 discrete 100 μm fibers. The system is adjusted so that the 10% trigger pulses are also collected by this fiber bundle. Both the reflected echo pulses and the trigger pulses are thus fed to a Specim V10 spectrograph. The linear fiber bundle end is directly matched to the spectrograph slit.

The optical part is designed to have a minimal amount of adjustments. The only adjustable components in terms of alignment are the fiber bundle and the spectrograph. The light collecting input side of the fiber bundle is aligned using a 3 axis miniature translation stage. This allows the bundle to be precisely set in the focal axis of the parabolic mirror with easily varied focal distance for different measuring ranges. The output of the fiber bundle has a 2 axis translation stage with a custom made adapter allowing the precise positioning over the spectrograph slit. In addition, the spectrograph is mounted on a miniature goniometer allowing smooth selection of the wavelengths projected over the photodiode sensor. All the other components are fixed in a custom aluminum housing with special recesses and mounting points for the beam sampler and the 5mm silver mirrors, which are glued in place after alignment. This optical housing can be seen on top part of the instrument in Fig. 1 with a circular 76.2 mm hole in the front plate for the parabolic mirror.

The described optical design was selected in order to provide minimal amount of optical components and hence, minimal losses. Attaching the light collecting parabolic mirror directly to the moving part of the system avoids additional mirrors in the optical path, which would have increased light dispersion and losses and furthermore required additional alignment efforts. The current design was also perceived to have smallest overall size. However, due to the mass of the moving parts, this system is less agile and scanning speeds are lower.

2.2. Mechanics

The enclosure housing the electronics is made from a combination of off-the-shelf (OTS) and custom manufactured components. Aluminum profiles and breadboard plates are used for the enclosure frame, which is then covered with aluminum and carbonfiber sheets. The pan-and-tilt platform uses OTS rotation stages with an output resolution of 0.0005° per axis. According to manufacturer, the absolute accuracy per axis is ±0.0115° or better and maximum scanning speed is 80°/s.

Additionally, the laser output fiber and the light collecting fiber bundle are encased inside plastic tubing in order to limit bending radiuses and to provide strain relief. Due to the fibers, we have limited the pan-and-tilt platform operating angles to roughly ±60° vertically and ±100° horizontally.

2.3. Electronics

Design of the instrument electronics has focused largely on high speed data acquisition implementation. Traditional ADC-solutions have high power demand, which is problematic in compact portable systems due to heat generation and consequent cooling requirements and also due to preferred battery operation. After reviewing alternative solutions, we settled on switched capacitor array (SCA) based solution. These are a type of analog buffer, allowing the capture of short analog signal segments. After capture, these buffers are read with traditional ADC. However, the reading can be done with significantly lower speed and higher accuracy ADC. This essentially allows superior noise levels together with higher conversion accuracy for a fraction of the cost of traditional ADC solutions. The power consumption is also significantly lower.

We use two DRS4-evaluation boards manufactured by Paul Scherrer Institute [14] to capture 8-channel waveforms with sampling speeds in excess of 5 GHz. DRS4 chips can record 1024 samples per channel and the channels can further be cascaded for deeper buffers. We currently use 1024 deep buffer at 5.12 GHz sampling speed, which allows time-of-flight (ToF) distance range of roughly 30m. The two boards are connected via USB2 to an ×86-based miniature computer running Linux operating system. A high level schematic of the system is presented in Fig. 3.

 figure: Fig. 3

Fig. 3 High level schematic of the instrument electronics and power supply.

Download Full Size | PPT Slide | PDF

Avalanche photodiode array is connected via SMA-cables to these DRS-digitizer boards. We use First Sensor’s linear 16-element 16AA0.4-9 APD-array on MOD501568 evaluation board. This sensor is less than optimal due to its roughly 300 MHz cut-off frequency and a highly wavelength dependent sensitivity. The sensor’s limited impulse response to the 1 ns laser pulse causes distortion in captured pulse shape and post-pulse oscillation. Preliminary frequency domain analysis of waveforms captured with the given components indicate that up to a roughly 1 GHz signal components in the pulses can still be detected from the noise floor. Sensor sensitivity also decreases quickly when operated at either edge of the measuring range of roughly 450–1000 nm. If a wide wavelength range is measured, varying sensitivity causes challenges with dynamic range of the instrument. Currently, 8 channels of the APD are utilized.

We implemented a custom motor controller platform on a dedicated microcontroller, which allows precise real-time control of the pan-and-tilt-platform. Previous prototype suffered from timing and synchronization issues in the angular data. This was caused by nondeterministic delays between high level operating system and discrete motor controllers interfaced through RS232. In the new implementation, motor encoders are read with hardware counters in the microcontroller and used for PID-based control of motors via PWM controlled power stages. This microcontroller platform is also connected to the laser via isolated RS232 converter and an additional trigger signal. The laser has an internal photodiode producing short pulses when the laser pulse is sent out. We read this photodiode signal with the microcontroller and capture the exact angles from the hardware counters. This provides submicrosecond synchronization between the gimbal angles and the outgoing pulse. The microcontroller communicates with the ×86-based host via ethernet, allowing control from the ×86-side and transfer of the angular information from the motor controller. Use of Ethernet provides galvanic isolation between the ×86 host and the microcontroller and a more deterministic data transfer compared to buses like USB. The laser pulse rate and operation is controlled from ×86 host with the microcontroller acting as a low level abstraction layer for RS232 based interface.

The electronics are designed for battery operation, allowing a roughly 14–18 VDC supply voltage, which can be provided by typical batteries with optional regulation using DC/DC-converter. Power consumption depends on operating parameters, such as laser pulse rate and speed of scanning. A typical average power consumption while scanning is roughly 50 W or less.

A custom software written for ×86 performs recording of the captured pulse waveforms from the digitizers and angular data from the microcontroller. Precise timestamps are used to synchronize waveforms with angular data. In addition, this software has a user interface functionality, which allows the instrument to be controlled via Wi-Fi connection from remote device, such as, a tablet-PC.

Main issue with the current electronics design is the USB2 connections in the digitizers. This connection type limits the system to pulse speeds of only a few hundred pulses per second. This issue could be solved with higher speed interfaces. However, when the pulse rate increases, the storage speed becomes a problem, as the amount of data is very high. This could be further solved with real-time processing of the raw data, but most likely requires fine tuning of the algorithms to specific application. For our research settings, low pulse capture speed has not been an issue so far and hence, we have settled for the speed of 200 pulses per second.

3. Data processing methods and instrument accuracy

During the development of the instrument, we have implemented simple processing algorithms for testing and prelimary measurement purposes. These algorithms are based on curve fitting methods, performed individually on both, the trigger and the echo pulses of each channel. From the fitted curves we can estimate peak places and values of the trigger and echo pulses and possibly additional parameters, such as the width of the pulses. Normalizing the backscattered laser pulse intensity (amplitude) with the trigger intensity, we get normalized pulse intensities for each channel. To convert these intensities into backscattered reflectance, a distance and spectral calibration can be applied as presented in [6]. Applying these calibrations will also remove any differences in normalized intensities between channels, which may occur due to the optical performance being wavelength dependent.

We have tested mainly polynomial- and Gaussian-based curve fitting and compared the speed and accuracy of these algorithms. Due to the high sampling speed of the digitizers, the waveforms are captured with relatively high resolution, as depicted in Fig. 4. Oversampling allows us to use noise averaging in the form of curve fitting, which essentially improves the signal-to-noise ratio. This makes it possible to estimate very low intensity signals with complex curve fitting models requiring more than three sample points. According to our preliminary tests, for higher intensity pulses, 7 to 15 sample points around the peak value provide a decent estimation of the pulse characteristics. With low level pulses with the peak at, for example, twice the noise level, we can still get between 3 to 5 sample points that lie above this noise level. Hence, even these low level signals can be estimated, albeit with a degraded accuracy due to the signal-to-noise ratio.

 figure: Fig. 4

Fig. 4 An example waveform captured on a single channel. Peak on the left is the 10% trigger portion separated by the beamsampler while the peak on the right is the reflected echo from the target. The oscillation following these pulses is attributed to the suboptimal bandwidth of the sensor. Signal range extends from −50 to 950 mV.

Download Full Size | PPT Slide | PDF

We have evaluated the polynomial- and Gaussian-based curve fitting methods in terms of ToF and intensity estimation precision and computational speed. These calculations are done offline with mathematical software using typical desktop computer and without using any hardware acceleration. Preliminary evaluations indicate, that the polynomial curve fitting provides a reasonable compromise between speed and accuracy. Gaussian-based curve fitting is significantly slower and seems to provide only marginal improvements in precision. Based on the preliminary tests, it seems that various noise sources in the system are of such nature, that both polynomial model and Gaussian model capture this noise floor. Due to these reasons, we have mostly relied on polynomial-based curve fitting. However, use of hardware accelerated curve fitting could make the computational cost of different models so small, that we will switch to using only Gaussian models. For example, Levenberg-Marquardt based Gaussian curve fitting can be accelerated on a graphical processing unit (GPU), reaching speeds of over 100000 fits per second on a modern consumer level hardware [15].

Another aspect of the curve fitting is that the pulses themselves are not strictly Gaussian. This is largely due to the finite impulse response of the APD sensor, causing skewing of the pulse and the after pulse oscillation. It is possible, that a further improvement in the precision could be achieved by using a custom curve model accommodating these nonidealities, but we have not yet tested this hypothesis. One of the challenges for more complicated models is the after pulse oscillation, which appears to change on a wavelength and temporal basis. Hence, different wavelength channels might require different models. The oscillation of the trigger pulse is furthermore superimposed on the echo pulse. As the trigger pulse oscillation quickly decreases, this only becomes problematic when the target is very close. According to our tests, targets closer to 2–3 meters tend to suffer from artifacts caused by these oscillations. While theoretically possible, it is difficult to reliably redact this oscillation on a pulse-by-pulse basis due to temporal variations in power and pulse shape.

ToF distance estimation on a multichannel instrument has few different options for implementation. ToF can be calculated for each individual channel and then combined using mean or median, or a combination of both. It is also possible to sum the individual waveforms together, and perform curve fitting on this average waveform. The former methods seem more sensible, as the waveforms have channel-wise differences and biases, making an average waveform somewhat problematic solution.

3.1. Precision of intensity estimation

We have conducted normalized intensity estimation tests using Spectralon as a target surface. We have performed tests both on static targets and while scanning a target surface. We have not noticed any difference in the results, which indicates that the instrument does not suffer from movement related mechanical or optical stability issues. Tests conducted using 12.5% and 25% Spectralon plates at 4.5 m distance give roughly 2% standard deviation for intensity in channels residing between 520 and 1000 nm. Sample size for these statistics was 100. Channels at the APD edges can achieve standard deviation of 4% for intensity. Intensity variations for each channel is depicted in Fig. 5, which shows raw normalized intensity values, before any calibration. These results are acquired using second order polynomial with 11 samples around the peak for both trigger and echo pulses.

 figure: Fig. 5

Fig. 5 Pulse intensities of each of the channels while scanning a Spectralon surface consisting of 12% and 25% Spectralon plates. Samples between 1258 and 1278 are from the 12%, and samples between 1280 and 1305 are from the 25% Spectralon, respectively. Presented values are raw values of the echo-to-trigger ratios, before calibration and without removing static biases. Channel at 473 nm shows slightly higher variations due to it being at edge of the sensor detection range.

Download Full Size | PPT Slide | PDF

There are variations between channels based on the wavelength the channels are set to. If a channel wavelength is at the extreme of the APD sensor range, then the variance of the particular intensity increases slightly. This variance does not seem to change significantly whether we use polynomial or Gaussian models. We assume it is related to the lower signal-to-noise ratio of these channels.

3.2. Precision of distance estimation

We have tested the distance estimation precision using the same tests done with intensity estimation. While distance precision may seem less important, it has a crucial role for accurate calibration of intensity values. Received pulse intensity decreases as a square of the distance to the target and echo values require readjusting based on this distance. Noise related variations in the distance estimate therefore directly impact the calibrated intensity estimates. Providing accurate distance estimation is therefore an integral part of a high quality measurement. Furthermore, it affects the 3D point cloud precision, which in many cases is of great importance.

Figure 6 presents ToF from a scanning of rocks against a Spectralon background. Left side of the Fig. 6 presents overall distance over a 5 cycle scan. On the right side of Fig. 6 is a detail of the middlemost peak, where the target material has been 50% Spectralon. The ToF varies within ±1.5 mm on this short scan segment. These results were obtained using polynomial model with 11 samples for the trigger and 7 samples for echo pulse and combining all 8 channels using a combination of both median and mean calculations. In essence, four of the eight channels were dropped using median values, and the remaining four were used for the average ToF presented in the Fig. 6. In general, standard deviation of ToF remains below 2 mm for planar targets when the reflectivity of the target and the distance to the target allows us to get 7 samples roughly 2–3 times above the noise floor on all channels. Incidence angle, target shape, if not planar and channel wavelength configuration among other things affect this estimation. Thorough characterization of ToF accuracy is therefore left for future work and the previous example only showcases the best achieved results so far. Eventually, ToF accuracy is a statistical characteristic when measuring actual targets, as it then depends on the combination of intensities of different wavelength channels. However, the accuracy can in most cases be improved by adding more channels. Use of mean and median and their combination can also be altered, giving varying results. Averaging all eight channels gives generally smaller variance, but allows occasional outliers. With a combination of mean and median, we essentially get smaller absolute error with slightly higher variance. Hence, the optimal solution also depends on the application needs. Further improvements in ToF estimation could be achieved by taking into account, for example, differences in biases and ToF precisions of individual channels.

 figure: Fig. 6

Fig. 6 Example of ToF distance estimation precision. On the left, a ToF for a scanning of rocks against a Spectralon surface is given. Highest peaks on this graph are the Spectralon surfaces. On the right side, a detailed view of the middlemost peak is given, depicting the millimeter level accuracy of the ToF estimation.

Download Full Size | PPT Slide | PDF

4. Results and discussion

We have tested the performance of the instrument and algorithms on a real application by measuring a variety of rock samples in laboratory settings. The measured rocks represent typical samples collected from mines for distinguishing ore from gangue in mineral separation. For laboratory tests, the rocks are dried and gently brushed to clean off dust. Figure 7 depicts a scan of one such rock, showing the level of detail achievable with the instrument. In laboratory setup, most of the cleaned rocks provide reasonable reflectance, allowing both, precise point cloud construction and spectral characteristics estimation.

 figure: Fig. 7

Fig. 7 A point cloud of a sample rock is plotted with intensity values of one measurement channel. Picture of the rock is provided for reference. Z-coordinate indicates the distance from the instrument. The measurement distance in this case was roughly 4.65 m.

Download Full Size | PPT Slide | PDF

We have also made rock measurements in an underground mine, where conditions were significantly more demanding due to humidity, dust and moisture covering the rocks. Consequently, the reflectances of the measured rocks were significantly lower compared to the measurements performed at laboratory settings. Figure 8 presents one example of a rock surface scanned in actual mining conditions. A 99% Spectralon disk was placed on the rocks to provide reference and calibration.

 figure: Fig. 8

Fig. 8 Intensity point cloud of a moist Serpentine rock surface scanned in an underground mine. A bright red target on the left figure is a 10 cm diameter 99% Spectralon disc, which was placed on the rocks for reference and calibration purposes. Compared to the Spectralon, the reflectance of the rock surface is very low. This represents typical intensity level for all the samples we have tested. The same point cloud is presented on the right side with the values scaled so that the details of the dark regions are more observable. The scanned rocks were moist and partially covered with dust. Measuring distance was roughly 3.4 m.

Download Full Size | PPT Slide | PDF

Many of the measured rocks exhibit for some or all of the channels an average intensity values below 1/10th of the 99% reference Spectralon measured at equal distance. The rock in Fig. 8 presents above average values, selected to better visualize the details. Nevertheless, the sensitivity of the instrument allows reliable spectral comparisons at these low levels. For a simple comparison, we can plot recorded intensity values of three measurement channels to form 3D scatter plots. In Fig. 9, we have provided such a comparison between two different rock types. These scatter plots show observable clustering patterns between certain rock types. Observability of these clusters varies among rock types and some of the rock types encountered seem to have practically identical spectral responses. Extensive testing on the use of various wavelengths and channel widths needs to be carried out. It is highly likely that the measurement wavelengths need to be optimized individually for various rock types.

 figure: Fig. 9

Fig. 9 Intensities of two different rock samples as a function of three different measurement channels. Two different channel sets are given in order to show the distinctive clustering patterns of the samples.

Download Full Size | PPT Slide | PDF

We have also tested the instrument in ambient conditions, including direct sunlight. We have not observed any degradation in performance caused by ambient lighting. Due to the AC-coupled signal chain and the fast, submicrosecond measurement time, comparatively slower natural changes in ambient lighting will not interfere with the measurements. Fast phenomena, such as sparking, or light sources modulated above few megahertz could potentially affect measurements, but these sources would have to reside within the scanned area, as the optics are designed for a very narrow field of view.

Challenges with the low reflectance targets could be alleviated, at least in theory, by increasing amplification and instrument sensitivity. However, the main obstacle for this are the optical sensors, which have a highly wavelength dependent sensitivities. Channels at the edge of the wavelength range remain close to the noise level, which significantly degrades the dynamic range of the system. Logarithmic amplification between the sensor and digitizers could be utilized to provide better dynamic measurement range. With individually settable amplification for each channel, sensitivity and level differences between channels could also be decreased significantly, further increasing the dynamic range. Due to channel-wise differences in trigger-echo ratios, full equalization is not possible. Furthermore, implementing high speed logarithmic amplifiers come with certain drawbacks, such as increased cost and measuring noise.

Number of measuring channels could also be increased by adding more digitizers. With SCA technology, the previously significant barrier to high channel count, namely the digitizer cost, is no longer a major factor. This is also an important issue considering the future use of multispectral laser scanning in consumer applications, such as vehicles and robotics. Currently, the SCA-technology provides significantly cheaper solutions compared to traditional analog-to-digital converter technology with similar performance. A more fundamental challenge in increasing the channel count is to find suitable, high bandwidth, high sensitivity sensor arrays. Currently, there does not seem to exist suitable integrated sensors having channel counts in excess of 32. Moreover, even the existing low channel count sensors tend to have relatively poor wavelength and pulse responses. This essentially means that for higher channel counts, a custom sensor would have to be constructed from individual sensor elements, which is challenging from the optics perspective.

Sensors with higher bandwidth could potentially improve both ToF and intensity estimation as these would introduce less distortion and oscillation in the pulse, which are significant sources of uncertainty in the measurements. However, higher bandwidth sensor would incorporate more noise and is generally achieved by smaller physical sensor surface area, possibly affecting optical reception. It is hence difficult to assess the possible improvement quantitatively and thorough benchmarking of different sensors would be required with custom DRS analog frontends designed for each sensor.

Use of high speed digitizers increases requirements for storage capacity and recording bandwidth. These could consequently become the limiting factor for the achievable pulse rate. For most research tasks, low pulse rates do not constitute significant issues. If higher pulse rates are required, developing custom electronics with adequate interconnections is straightforward, albeit a considerable undertaking. Increasing the channel count would also eventually mean decreasing measuring channel wavelength widths, which would improve spectral resolution. It is highly likely that increased channel count would require further improvements in the optical efficiency and sensor sensitivity. For these reasons it is likely that the future implementations of multiwavelength LiDARs will operate on a low number of channels. On the other hand, many applications do not necessarily require more channels.

The design of the instrument presented here is still a compromise between modularity and compactness. In its current form, the instrument allows easy modifications, for example, testing of different sensors. Considering the nature of the instrument, this is a highly desirable feature as the instrument will be employed on a variety of tasks. If required, significant reductions in weight and size could still be achieved by implementing fully customized electronics, mechanics and optics. With the given noise characteristics and signal bandwidth of the system, the sampling rate could likely be decreased somewhat without significant degradation in measuring performance. This drop would likely show first on the ToF estimation and hence, the optimal rate depends on the application needs.

5. Conclusions

In this paper, we have presented the design of a compact, portable, high accuracy hyperspectral LiDAR instrument. Furthermore, we have described algorithms used for preliminary bench-markings and field tests together with some accuracy estimates achievable with the instrument and algorithms. To the authors knowledge, this is the first truly portable and field operable supercontinuum based hyperspectral LiDAR capable of producing precision 3D point clouds comprising spectral data. Compared to earlier prototypes, this device is also capable of accurate single pulse measurements from very low reflectance targets. In practice, 12% Spectralon target can now be measured with equivalent or better precision than 99% Spectralon target on our old prototype with 1 GHz sampling. The wavelength range is also increased as we can now operate 30 – 50 nm closer to the edges of the APD range. Angular accuracy is also improved roughly tenfold compared to the old prototype. The presented unit is also capable of producing very precise time-of-flight estimates, improving accuracy of calibrations and corrections. Due to these characteristics, this new instrument allows novel application areas, which require a brightness resolution high enough to distinguish minute differences in backscattered reflectance. For such an application, we have provided an example in ore content and rock type estimation in mines for which we have shown some preliminary results. These results show that certain rock types can be identified using measurements provided by the instrument. Furthermore, this identification can be accomplished even with extremely low intensity values, as the clustering is observable from the lowest captured light levels.

We have already started implementing full benchmark and calibration of the instrument and we intend to provide a thorough performance and accuracy assessment of the device in the future. Furthermore, our future work will include testing of different novel photosensors and improving the dynamic range of the instrument. On the application area, we will be using the instrument for various ore type measurements and ore content estimation in mining conditions.

Funding

Business Finland (project 1515/31/2016: ’Efficient and safe identification of minerals: smart real-time methods’).

References

1. M. Kalacska, M. Lalonde, and T. Moore, “Estimation of foliar chlorophyll and nitrogen content in an ombrotrophic bog from hyperspectral data: Scaling from leaf to image,” Remote Sens. Environ. 169, 270–279 (2015). [CrossRef]  

2. J. A. R. Rall and R. G. Knox, “Spectral ratio biospheric lidar,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2004), 1951–1954.

3. I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011). [CrossRef]  

4. E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

5. F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014). [CrossRef]  

6. T. Hakala, J. Suomalainen, S. Kaasalainen, and Y. Chen, “Full waveform hyperspectral lidar for terrestrial laser scanning,” Opt. Express 20, 7119–7127 (2012). [CrossRef]   [PubMed]  

7. W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014). [CrossRef]  

8. L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016). [CrossRef]  

9. R. J. Murphy, S. T. Monteiro, and S. Schneider, “Evaluating classification techniques for mapping vertical geology using field-based hyperspectral sensors,” IEEE Trans. Geosci. Remote Sens. 50, 3066–3080 (2012). [CrossRef]  

10. L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016). [CrossRef]  

11. M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017). [CrossRef]  

12. A. Roncat, G. Bergauer, and N. Pfeifer, “B-spline deconvolution for differential target cross-section determination in full-waveform laser scanning data,” ISPRS J. Photogramm. Remote Sens. 66, 418–428 (2011). [CrossRef]  

13. A. Ullrich and M. Pfennigbauer, “Echo digitization and waveform analysis in airborne and terrestrial laser scanning,” Photogramm. Week 11, 217–228 (2011).

14. S. Ritt, R. Dinapoli, and U. Hartmann, “Application of the drs chip for fast waveform digitizing,” Nucl. Instrum. Meth. A. 623, 486–488 (2010). [CrossRef]  

15. A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017). [CrossRef]   [PubMed]  

References

  • View by:
  • |
  • |
  • |

  1. M. Kalacska, M. Lalonde, and T. Moore, “Estimation of foliar chlorophyll and nitrogen content in an ombrotrophic bog from hyperspectral data: Scaling from leaf to image,” Remote Sens. Environ. 169, 270–279 (2015).
    [Crossref]
  2. J. A. R. Rall and R. G. Knox, “Spectral ratio biospheric lidar,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2004), 1951–1954.
  3. I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
    [Crossref]
  4. E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.
  5. F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
    [Crossref]
  6. T. Hakala, J. Suomalainen, S. Kaasalainen, and Y. Chen, “Full waveform hyperspectral lidar for terrestrial laser scanning,” Opt. Express 20, 7119–7127 (2012).
    [Crossref] [PubMed]
  7. W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014).
    [Crossref]
  8. L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
    [Crossref]
  9. R. J. Murphy, S. T. Monteiro, and S. Schneider, “Evaluating classification techniques for mapping vertical geology using field-based hyperspectral sensors,” IEEE Trans. Geosci. Remote Sens. 50, 3066–3080 (2012).
    [Crossref]
  10. L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016).
    [Crossref]
  11. M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
    [Crossref]
  12. A. Roncat, G. Bergauer, and N. Pfeifer, “B-spline deconvolution for differential target cross-section determination in full-waveform laser scanning data,” ISPRS J. Photogramm. Remote Sens. 66, 418–428 (2011).
    [Crossref]
  13. A. Ullrich and M. Pfennigbauer, “Echo digitization and waveform analysis in airborne and terrestrial laser scanning,” Photogramm. Week 11, 217–228 (2011).
  14. S. Ritt, R. Dinapoli, and U. Hartmann, “Application of the drs chip for fast waveform digitizing,” Nucl. Instrum. Meth. A. 623, 486–488 (2010).
    [Crossref]
  15. A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017).
    [Crossref] [PubMed]

2017 (2)

M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
[Crossref]

A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017).
[Crossref] [PubMed]

2016 (2)

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016).
[Crossref]

2015 (1)

M. Kalacska, M. Lalonde, and T. Moore, “Estimation of foliar chlorophyll and nitrogen content in an ombrotrophic bog from hyperspectral data: Scaling from leaf to image,” Remote Sens. Environ. 169, 270–279 (2015).
[Crossref]

2014 (2)

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014).
[Crossref]

2012 (2)

R. J. Murphy, S. T. Monteiro, and S. Schneider, “Evaluating classification techniques for mapping vertical geology using field-based hyperspectral sensors,” IEEE Trans. Geosci. Remote Sens. 50, 3066–3080 (2012).
[Crossref]

T. Hakala, J. Suomalainen, S. Kaasalainen, and Y. Chen, “Full waveform hyperspectral lidar for terrestrial laser scanning,” Opt. Express 20, 7119–7127 (2012).
[Crossref] [PubMed]

2011 (3)

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

A. Roncat, G. Bergauer, and N. Pfeifer, “B-spline deconvolution for differential target cross-section determination in full-waveform laser scanning data,” ISPRS J. Photogramm. Remote Sens. 66, 418–428 (2011).
[Crossref]

A. Ullrich and M. Pfennigbauer, “Echo digitization and waveform analysis in airborne and terrestrial laser scanning,” Photogramm. Week 11, 217–228 (2011).

2010 (1)

S. Ritt, R. Dinapoli, and U. Hartmann, “Application of the drs chip for fast waveform digitizing,” Nucl. Instrum. Meth. A. 623, 486–488 (2010).
[Crossref]

Armitage, R. P.

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

Bates, M.

A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017).
[Crossref] [PubMed]

Bergauer, G.

A. Roncat, G. Bergauer, and N. Pfeifer, “B-spline deconvolution for differential target cross-section determination in full-waveform laser scanning data,” ISPRS J. Photogramm. Remote Sens. 66, 418–428 (2011).
[Crossref]

Chakrabarti, S.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Chen, Y.

Cook, T.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Culvenor, D.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Danson, F.

L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016).
[Crossref]

Danson, F. M.

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

Dinapoli, R.

S. Ritt, R. Dinapoli, and U. Hartmann, “Application of the drs chip for fast waveform digitizing,” Nucl. Instrum. Meth. A. 623, 486–488 (2010).
[Crossref]

Disney, M.

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

Douglas, E. S.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Du, L.

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

Entwistle, N.

L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016).
[Crossref]

Gao, S.

W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014).
[Crossref]

Gaulton, R.

L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016).
[Crossref]

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

Gong, W.

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

Gunawan, O.

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

Hakala, T.

Hancock, S.

L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016).
[Crossref]

Hartmann, U.

S. Ritt, R. Dinapoli, and U. Hartmann, “Application of the drs chip for fast waveform digitizing,” Nucl. Instrum. Meth. A. 623, 486–488 (2010).
[Crossref]

Hollaus, M.

M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
[Crossref]

Jack, J.

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

Jupp, D.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Kaasalainen, S.

Kalacska, M.

M. Kalacska, M. Lalonde, and T. Moore, “Estimation of foliar chlorophyll and nitrogen content in an ombrotrophic bog from hyperspectral data: Scaling from leaf to image,” Remote Sens. Environ. 169, 270–279 (2015).
[Crossref]

Keller-Findeisen, J.

A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017).
[Crossref] [PubMed]

Knox, R. G.

J. A. R. Rall and R. G. Knox, “Spectral ratio biospheric lidar,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2004), 1951–1954.

Lalonde, M.

M. Kalacska, M. Lalonde, and T. Moore, “Estimation of foliar chlorophyll and nitrogen content in an ombrotrophic bog from hyperspectral data: Scaling from leaf to image,” Remote Sens. Environ. 169, 270–279 (2015).
[Crossref]

Lewis, P.

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

Li, W.

W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014).
[Crossref]

Li, Z.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Lovell, J.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Malthus, T. J.

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

Marshall, R.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Martel, J.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Mendillo, C.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Milenkovic, M.

M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
[Crossref]

Monteiro, S. T.

R. J. Murphy, S. T. Monteiro, and S. Schneider, “Evaluating classification techniques for mapping vertical geology using field-based hyperspectral sensors,” IEEE Trans. Geosci. Remote Sens. 50, 3066–3080 (2012).
[Crossref]

Moore, T.

M. Kalacska, M. Lalonde, and T. Moore, “Estimation of foliar chlorophyll and nitrogen content in an ombrotrophic bog from hyperspectral data: Scaling from leaf to image,” Remote Sens. Environ. 169, 270–279 (2015).
[Crossref]

Morsdorf, F.

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

Murphy, R. J.

R. J. Murphy, S. T. Monteiro, and S. Schneider, “Evaluating classification techniques for mapping vertical geology using field-based hyperspectral sensors,” IEEE Trans. Geosci. Remote Sens. 50, 3066–3080 (2012).
[Crossref]

Newnham, G.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Nichol, C.

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

Niu, Z.

W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014).
[Crossref]

Patenaude, G.

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

Pearson, G.

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

Pfeifer, N.

M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
[Crossref]

A. Roncat, G. Bergauer, and N. Pfeifer, “B-spline deconvolution for differential target cross-section determination in full-waveform laser scanning data,” ISPRS J. Photogramm. Remote Sens. 66, 418–428 (2011).
[Crossref]

Pfennigbauer, M.

A. Ullrich and M. Pfennigbauer, “Echo digitization and waveform analysis in airborne and terrestrial laser scanning,” Photogramm. Week 11, 217–228 (2011).

Przybylski, A.

A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017).
[Crossref] [PubMed]

Qiao, H.

W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014).
[Crossref]

Quast, R.

M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
[Crossref]

Rall, J. A. R.

J. A. R. Rall and R. G. Knox, “Spectral ratio biospheric lidar,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2004), 1951–1954.

Ramirez, A. F.

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

Ressl, C.

M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
[Crossref]

Ritt, S.

S. Ritt, R. Dinapoli, and U. Hartmann, “Application of the drs chip for fast waveform digitizing,” Nucl. Instrum. Meth. A. 623, 486–488 (2010).
[Crossref]

Roncat, A.

A. Roncat, G. Bergauer, and N. Pfeifer, “B-spline deconvolution for differential target cross-section determination in full-waveform laser scanning data,” ISPRS J. Photogramm. Remote Sens. 66, 418–428 (2011).
[Crossref]

Schaaf, C.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Schneider, S.

R. J. Murphy, S. T. Monteiro, and S. Schneider, “Evaluating classification techniques for mapping vertical geology using field-based hyperspectral sensors,” IEEE Trans. Geosci. Remote Sens. 50, 3066–3080 (2012).
[Crossref]

Schofield, L.

L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016).
[Crossref]

Shi, S.

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

Sinclair, P.

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

Song, S.

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

Stock, B.

A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017).
[Crossref] [PubMed]

Strahler, A.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Sun, G.

W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014).
[Crossref]

Sun, J.

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

Suomalainen, J.

Thiel, B.

A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017).
[Crossref] [PubMed]

Ullrich, A.

A. Ullrich and M. Pfennigbauer, “Echo digitization and waveform analysis in airborne and terrestrial laser scanning,” Photogramm. Week 11, 217–228 (2011).

Wagner, W.

M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
[Crossref]

Woodcock, C.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Woodhouse, I. H.

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

Yang, J.

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

Yang, X.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Zhu, B.

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

Agric. For. Meteorol. (1)

F. M. Danson, R. Gaulton, R. P. Armitage, M. Disney, O. Gunawan, P. Lewis, G. Pearson, and A. F. Ramirez, “Developing a dual-wavelength full-waveform terrestrial laser scanner to characterize forest canopy structure,” Agric. For. Meteorol. 198–199, 7–14 (2014).
[Crossref]

IEEE Geosci. Remote Sens. Lett. (1)

I. H. Woodhouse, C. Nichol, P. Sinclair, J. Jack, F. Morsdorf, T. J. Malthus, and G. Patenaude, “A multispectral canopy lidar demonstrator project,” IEEE Geosci. Remote Sens. Lett. 8, 839–843 (2011).
[Crossref]

IEEE Trans. Geosci. Remote Sens. (1)

R. J. Murphy, S. T. Monteiro, and S. Schneider, “Evaluating classification techniques for mapping vertical geology using field-based hyperspectral sensors,” IEEE Trans. Geosci. Remote Sens. 50, 3066–3080 (2012).
[Crossref]

Int. J. Appl. Earth Obs. Geoinf. (1)

L. Du, W. Gong, S. Shi, J. Yang, J. Sun, B. Zhu, and S. Song, “Estimation of rice leaf nitrogen contents based on hyperspectral lidar,” Int. J. Appl. Earth Obs. Geoinf. 44, 136–143 (2016).
[Crossref]

ISPRS J. Photogramm. Remote Sens. (2)

M. Milenković, W. Wagner, R. Quast, M. Hollaus, C. Ressl, and N. Pfeifer, “Total canopy transmittance estimated from small-footprint, full-waveform airborne lidar,” ISPRS J. Photogramm. Remote Sens. 128, 61–72 (2017).
[Crossref]

A. Roncat, G. Bergauer, and N. Pfeifer, “B-spline deconvolution for differential target cross-section determination in full-waveform laser scanning data,” ISPRS J. Photogramm. Remote Sens. 66, 418–428 (2011).
[Crossref]

Nucl. Instrum. Meth. A. (1)

S. Ritt, R. Dinapoli, and U. Hartmann, “Application of the drs chip for fast waveform digitizing,” Nucl. Instrum. Meth. A. 623, 486–488 (2010).
[Crossref]

Opt. Express (1)

Photogramm. Week (1)

A. Ullrich and M. Pfennigbauer, “Echo digitization and waveform analysis in airborne and terrestrial laser scanning,” Photogramm. Week 11, 217–228 (2011).

Remote Sens. Environ. (1)

M. Kalacska, M. Lalonde, and T. Moore, “Estimation of foliar chlorophyll and nitrogen content in an ombrotrophic bog from hyperspectral data: Scaling from leaf to image,” Remote Sens. Environ. 169, 270–279 (2015).
[Crossref]

Remote Sens. Lett. (2)

W. Li, G. Sun, Z. Niu, S. Gao, and H. Qiao, “Estimation of leaf biochemical content using a novel hyperspectral full-waveform lidar system,” Remote Sens. Lett. 5, 693–702 (2014).
[Crossref]

L. Schofield, F. Danson, N. Entwistle, R. Gaulton, and S. Hancock, “Radiometric calibration of a dual-wavelength terrestrial laser scanner using neural networks,” Remote Sens. Lett. 7, 299–308 (2016).
[Crossref]

Sci. Rep. (1)

A. Przybylski, B. Thiel, J. Keller-Findeisen, B. Stock, and M. Bates, “Gpufit: An open-source toolkit for gpu-accelerated curve fitting,” Sci. Rep. 7, 15722 (2017).
[Crossref] [PubMed]

Other (2)

J. A. R. Rall and R. G. Knox, “Spectral ratio biospheric lidar,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2004), 1951–1954.

E. S. Douglas, A. Strahler, J. Martel, T. Cook, C. Mendillo, R. Marshall, S. Chakrabarti, C. Schaaf, C. Woodcock, Z. Li, X. Yang, D. Culvenor, D. Jupp, G. Newnham, and J. Lovell, “Dwel: A dual-wavelength echidna lidar for ground-based forest scanning,” in Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IEEE, 2012), 4998–5001.

Cited By

OSA participates in Crossref's Cited-By Linking service. Citing articles from OSA journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (9)

Fig. 1
Fig. 1 New portable and improved version of FGI hyperspectral LiDAR. An enclosure housing all the electronics forms the base of the unit. On top of this base is attached a 2-axis pan-and-tilt-type gimbal, which performs the 2-dimensional scanning. Light transmitting and collection optics are fitted in a custom housing attached to this gimbal.
Fig. 2
Fig. 2 Optical components and topology of the instrument.
Fig. 3
Fig. 3 High level schematic of the instrument electronics and power supply.
Fig. 4
Fig. 4 An example waveform captured on a single channel. Peak on the left is the 10% trigger portion separated by the beamsampler while the peak on the right is the reflected echo from the target. The oscillation following these pulses is attributed to the suboptimal bandwidth of the sensor. Signal range extends from −50 to 950 mV.
Fig. 5
Fig. 5 Pulse intensities of each of the channels while scanning a Spectralon surface consisting of 12% and 25% Spectralon plates. Samples between 1258 and 1278 are from the 12%, and samples between 1280 and 1305 are from the 25% Spectralon, respectively. Presented values are raw values of the echo-to-trigger ratios, before calibration and without removing static biases. Channel at 473 nm shows slightly higher variations due to it being at edge of the sensor detection range.
Fig. 6
Fig. 6 Example of ToF distance estimation precision. On the left, a ToF for a scanning of rocks against a Spectralon surface is given. Highest peaks on this graph are the Spectralon surfaces. On the right side, a detailed view of the middlemost peak is given, depicting the millimeter level accuracy of the ToF estimation.
Fig. 7
Fig. 7 A point cloud of a sample rock is plotted with intensity values of one measurement channel. Picture of the rock is provided for reference. Z-coordinate indicates the distance from the instrument. The measurement distance in this case was roughly 4.65 m.
Fig. 8
Fig. 8 Intensity point cloud of a moist Serpentine rock surface scanned in an underground mine. A bright red target on the left figure is a 10 cm diameter 99% Spectralon disc, which was placed on the rocks for reference and calibration purposes. Compared to the Spectralon, the reflectance of the rock surface is very low. This represents typical intensity level for all the samples we have tested. The same point cloud is presented on the right side with the values scaled so that the details of the dark regions are more observable. The scanned rocks were moist and partially covered with dust. Measuring distance was roughly 3.4 m.
Fig. 9
Fig. 9 Intensities of two different rock samples as a function of three different measurement channels. Two different channel sets are given in order to show the distinctive clustering patterns of the samples.

Metrics