Open Access
Add to BibSonomy
Abstract
Flash LiDAR is a photoelectric system that can acquire a 3D image by emitting a diffuse pulsed laser beam, and hence is suitable for both autopilot and spacecraft flight control. Achieving long-range and high-speed, especially in outdoor applications with strong solar background illumination, are challenging requirements. In this paper, a set of laser imaging prototype based on 2×6 VCSEL array and 32×32 MPPC array image sensor is developed, the range calibration is completed, and relevant experimental research is carried out. The frame rate of the system can reach 10kHz, the detection probability of 120m range can reach 86.23%, and the maximum walk error is about 0.6m under different reflectivity. The 3D imaging of the vehicle can be realized at about 70m, the horizontal spatial resolution is less than 5cm, and the ranging precision after ten shots average is within 10cm by calculating the centroid of a histogram. The detection probability can be improved by using the time-gating method. After multiple measurements, a 120m “laser imaging through window” can be realized in sunlight. This LiDAR system has the advantages of small volume, light weight and fast detection speed.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Light Detection and Ranging (LiDAR) is a 3D imaging technique, widely used in many applications such as augmented reality, automotive, machine vision, spacecraft navigation and landing [1–4]. Flood illumination schemes is a kind of Flash LiDAR using non-scanning laser imaging method, which can obtain a 3D image as the key output content [5]. The imaging principle is similar to that of the flash camera [6], that is, the target image in 3D space is reconstructed by measuring the time of flight from the laser to each pixel on the target surface, a laser array is emitting to illuminate the whole field of view in a short time, and then the echo signal is collected by the high-sensitivity array detector to construct the 3D image. This imaging technology has the advantages of fast imaging speed and small volume. Its image resolution depends on the total number of pixels of the detector, and the lateral resolution depends on the spacing of adjacent pixels [7,8].
In 2013, Niclass [9] digital multi-pixel photon counter(MPPC) array, with 32 pixels of 12 single photon avalanche diode(SPAD) each, was one of the first specifically designed for LiDAR application with special feature for background robustness. By 1D scanning they achieved an image resolution of 340×90, field of view of 170°×4.5°, angular resolution of 0.5°×0.05 ° and frame rate of 10Hz.
In 2017, Perenzoni [10] designed a flash imaging LiDAR using a 64 × 64 2D array MPPC detector (each pixel contains 8 SPADs). The maximum imaging distance is 367m, the measurement accuracy is 20cm and the frame rate is 7.7Hz.
In 2019, Hutchings [11] designed a 256×256 SPAD array, which can work in photon counting mode, each pixel contains 4×4 SPADs, and they achieved maximum range of 50m and FOV of 1.2°×1.2°, angular resolution of 0.02°×0.02° and frame rate of 30Hz.
In 2020, Seo [12] designed a 1D scanning LiDAR with maximum range of 48m, measurement accuracy of 0.85cm and field angle of 120° using a 63-channel linear array MPPC detector (each pixel contains 4 SPADs), its angular resolution is 0.05°×0.2°, image resolution is up to 2200×36, with frame rate of 1.18Hz.
In 2021, Padmanabhan [13] designed a 256×128 SPADs array that can work in photon counting mode to form into a 16×16 SPADs array, its test distance is 10m, field angle 2°×2°, angular resolution 0.16°×0.16°, maximum image resolution can be up to 128×12.
In 2021, Kumagai [14] designed a 183×600 SPADs array, which is able to work in 3×4 or 4×4 aggregation mode, its FSR can reach 300m and its precion is 30cm, the field of view is 25.2°×9.45°, angular resolution 0.15°×0.15°, image resolution can be up to 168×63. It works in line field scanning mode in which it scans one line at a time, and its frame rate is 20Hz.
In theory, compared with the single point laser ranging system [15,16], the laser emission energy needs to be increased by several times to compensate. Such a scheme suffers from unnecessary laser energy wastage in the invalid area between pixels, and can be called a “brute” scheme. In order to realize long-distance three-dimensional imaging and solve the problem of low echo power density, this paper proposed the methods to improve detection probability and detection accuracy from three aspects.
Firstly, by reducing the scanning angle, the laser emission is focused on a smaller area, so that the laser pulse can transfer further under the same power. By changing the focal length, the LiDAR can adapt to different detection ranges.
Secondly, the laser transceiver adopts a vertical cavity surface emitting laser (VCSEL) array and MPPC array to form a transceiver array which both increases the transmission power and the reception area, hence it can detect photons for longer distance and weak echo signal condition.
Finally, the time-gating method [17] is used to further suppress the influence of background radiation, so as to maximize the utilization of laser power in each scan, and the laser detection distance is prolonged after the signal statistical processing.
2. Materials and methods
2.1 VCSEL array and its driver circuit
Figure 1(a) shows the 2×6 VCSEL array module composed of a total number of 12 VCSEL chips (BX110, Berxel, China). There are 110 emitting apertures in each chip, so the total number of emitting apertures is 1320. After homogenization by the lens, the output beam shape is close as a square. Fig. 1(b) shows a part of VCSEL array driver circuit, where switch Q is a GaN FET switch (EPC2034, EPC, USA). When the switch Q is closed, the capacitor C is the charged through the resistance R by DC power HV. When Q is open, C is discharged through Q. Table 1 shows this VCSEL module parameter. The laser peak power is over 1.2kW with a short pulse of 5.6ns while C is 2.2nF.
Table 1. VCSEL array module parametera
2.2 MPPC array and its evaluation circuit
The MPPC array detector (S15013-0125NP-01, Hamamatsu, Japan) is composed of a 32×32 channel SPADs cluster, each cluster is composed of 12 SPADs, and each channel is separately integrated with a time to digital converter (TDC) circuit, with resolution of 312.5ps. Table 2 gives detailed key technical parameters. The relationship between measurement distance d and TDC measurement time tdc_data can be expressed with Eq. (1).
Table 2. S15013-0125NP-01 parametera
Here, c is the speed of light, t is the time of flight of the laser, tdc_window is the time of TDC measurement window, tdc_data is the 16-bit output data of the TDC circuit, and TREF is 6.67ns.
As is shown in Fig. 2, the evaluation circuit of MPPC array detector is based on FPGA (EP4C55, Altera, USA) extended USB3.0 (cyusb3014, Cypress, USA) hardware architecture, which generate the trigger signal for the laser driving circuit, supply of high voltage bias for MPPC detector, serial peripheral interface (SPI) bus control of detector, parallel reading of the TDC row and column data, data cache, data processing and communication control of USB3.0 bus.
2.3 System design and analysis
2.3.1 LiDAR budget
Assuming that the optical system efficiency η is 0.8, the effective area of the receiving optical system is Ar, the receiving aperture of the detection system is 50mm, the target reflectivity ρ is 0.5, and the number of channels of the MPPC array Npix is 1024. It is assumed that 100% of the laser energy reaches the target after collimation, and the measured target conforms to the diffuse reflection model, based on the photon form of the plane array LiDAR equation, the relation between echo photon numbers of single channel Nr and the target distance R can be expressed with Eq. (2).
Here, Et is the laser pulse energy, h is the Planck’s constant, λ is the laser wavelength, and c is the speed of light.
It is assumed that the energy distribution of the echo beam is uniform, when the single pulse energy is 2µJ (corresponding to peak power of 1kW and pulse width of 2ns). As can be seen from Fig. 3(a), the number of photons received in each channel is more than 40, and the imaging distance is more than 250 meters.
Fig. 3. (a) The relation between the number of echo photons of each channel and distance (b) The relation between the number of fired pixels and the number of echo photons
The single-channel SPAD number of the MPPC array Ncell is 12, and the photon detection efficiency (PDE) is 6%. Based on Poisson distribution model with a small number of photons arrived, the relation between the number of echo photons Nr and the number of fired pixels Nfired can be expressed with Eq. (3).
As can be seen from the Fig. 3(b), when the average received photon is 40, the average number of fired pixels is more than 2. Meanwhile, compared with single channel wide dynamic multi-pixel photon counter [15], as the number of pixels of each channel is decreased, the dark count rate (DCR) will also decrease accordingly. In theory, the high-speed detection mode can be realized by setting the threshold to 1∼2 p.e.
2.3.2 System design
As is shown in Fig. 4, the coaxial optical transceiver with a 940nm narrow band filter (FWHM = 10nm) is adopted in this paper. The focal length of transceiver lens is adjustable, up to 120mm, and the optical aperture is 46mm. As the 32×32 MPPC array photosensitive area is about 4mm×4mm, according to the basic principle of geometric optics, the detection area of the imaging system at 120m is about 4m.
As is shown in Fig. 5(a) and Fig. 5(b), the spot of VCSEL array in near field and far field are tested independently. As the emitting area of the 2×6 VCSEL array is about 3.2mm, the focal length of the lens is 120mm, so the divergence angle is 26mrad. Theoretically, the light spot at 7.3m is about 20cm, which is consistent with the actual test result. The light spot near 120m is about 3m, while the height of the window in the figure is 2.3m, which is also consistent with the test result.
3. Results
3.1 Effects of dynamic threshold on detection probability
Due to the array MPPC image sensor, a single pixel contains 12 SPADs, so each pixel can correspond to up to 12 photon events. As shwon in Fig. 6, if the threshold voltage is set externally to suppress background noise, the detection probability can be improved. At both 12:00 a.m. and 12:00 p.m., the detection probability of the target is tested respectively under these two scenarios, as is shown in the table below.
It can be seen from Table 3 that under the weak background condition, the dark count is lower than the one-photon event, and the maximum detection probability (86.23%) can be obtained by setting the threshold voltage to less than 1p.e. However, under the strong background condition, there is a maximum value (7 p.e) in the setting of the threshold level. The selection of threshold level depends on the intensity of the background radiation intensity. The maximum detection probability (30.7%) can be obtained by setting the appropriate threshold. The experiment can also prove the theory that MPPC detector can obtain the maximum detection probability by setting the threshold in a few-photon scene [18,19].
Table 3. Detection probability and threshold voltage statistics (DCR = 6.88kcps)
3.2 Effects of laser intensity on laser imaging
This paper has conducted a target imaging experiment by changing the voltage of the capacitor C in the laser driving circuit, which hence changed the laser output power accordingly. The target distance is about 5m, the focal length of the receiving lens is set to 70mm, the HV voltage regulation range on the capacitor is from 70V to 15V, and the corresponding discharge current through the single switch is from 100A to 25A. Since one switch drives three VCSEL lasers, the current passing through each laser chip is about one third, about 33.3A to 8.3A, Since the electro-optic conversion slope efficiency of the VCSEL laser is 2.7W/A, the variation range of the optical power of one single VCSEL laser can be calculated as from 89.9W to 22.4W.
It can be seen from Fig. 7 that when the laser intensity is very strong, the target reflection has great interference with the background area, but with the weakening of the laser intensity, the background noise gradually decreases, and the target contour reaches clearest at 25V. Then, due to the weakening of the light intensity, the target contour gradually blurs.
It can be concluded that if dynamic threshold adjustment is added to the detection system and the background interference is deducted, the imaging system can adapt to the change of laser intensity.
3.3 Effects of echo laser intensity on detection probability
In this paper, the detection probability experiment is carried out by using the extinction coefficient method [20] in a weak background scene. The actual measurement distance is about 120 meters, and the attenuation ratios are 1×,10×,100×,1000×respectively. The detection probabilities with thresholds set at 1 p.e, 2 p.e and 3 p.e are measured respectively. The statistical results are shown in table 4.
Table 4. statistical data of detection probability and attenuation multiple
It can be seen from table 4 that by increasing the attenuation ratio, the detection probability of the system decreases continuously, the detection probability is 1.40% with attenuation ratio of 1000×.
3.4 Effects of target reflectivity on walk error
As is shown in Fig. 8(a), the reflectivity of the diffuse reflection target is 90%, 10% and 2% respectively, and the imaging effect of the distance between three regions A, B and C is shown in Fig. 8(b). The TDC output data is obtained by filtering out the invalid noise data in software, and then calculating the difference between the measured distances of the three areas.
Fig. 8. (a) Walk error experimental platform (b) Pseudo color display of TDC data (c)Pixel(H10, V26) depth data histogram (d) Pixel(H10, V15) depth data histogram
As is shown in Fig. 8(c) and Fig. 8(d), under the conditions of high reflectivity (90%) and low reflectivity (2%), the walk error of the two histograms is about 0.6m by calculating the centroid.
3.5 Laser imaging for vehicle without time-gating
As is shown in Fig. 9, without setting the threshold and time-gating, this paper conducted a three-dimensional vehicle imaging experiment at the distance of about 70 meters. The length of the vehicle license plate is 44 cm while the width is 14 cm, and the corresponding pixels of the license plate in the 3D image are 3×7 pixels, it can be calculated that the horizontal spatial resolution of the system is less than 5 cm. In table 5, the adjacent pixel distance error is within 0.5m for single measurement, since reflected light intensity depends on the cosine of the angle of incidence of the light, which is related to Lambert’s cosine law [21, 22]. As is shown in Fig. 10(c) and Fig. 10(d), the precision of the license plate center position after ten shots average is within 10cm by calculating the centroid.
Fig. 10. Histogram statistics of distance data of license plate center position (H6, V17) (a)Single shot (b) Ten shots average
Table 5. Distance data of license plate (unit: meter)
3.6 Laser imaging with time-gating under strong background
By setting the threshold, the detection probability in strong background scene can be improved, but the maximum range will be reduced. As is shown in Fig. 11(a), this paper verifies the application of time-gating method for outdoor sunlit laser imaging. Here, the target is the bulge of the building with a depth of about 2 meters. As is shown in Fig. 11(b), by predicting the target position, the TDC enable time is set as 793ns, and the TDC measurement window is set within 100ns to decrease the false trigger probability of the photons coming from background noise. As is shown in Fig. 11(c) and Fig. 11(d), the method based on time-gating and multiple measurements can realize the target ranging imaging at 120m under strong background through glass while the window is closed, that is, “laser imaging through window”.
Fig. 11. (a) Laser imaging with time-gating (b) Time setting of time-gating method (c) Pseudo color display of TDC data (d) Single line depth data display
4. Discussion and conclusion
Table 6 lists the performance of some SPAD and MPPC arrays when used in the LiDAR application. Medium ranges (about 50m FSR) and centimeter precisions can be reached, at least theoretically most of those architectures. However, the actual achievable range and precision depend not only on the detection chip but also on laser power, target reflectivity, and background. With respect to flash LiDAR, scanning setups typically provide higher image resolution (few kilo pixels in flash LiDAR vs tens of kilo pixels in scanning setups).
Table 6. Main specifications of selected SPAD and MPPC arrays for pulsed-LiDAR.
In this paper, based on 32 × 32 array MPPC image sensor and 2×6 VCSEL array, a kind of laser imaging prototype is developed, the range calibration is accomplished, and several relevant experiments are carried out. The frame rate of the LiDAR system can reach 10kHz, by setting the threshold at 1 p.e, the detection probability of 120m can reach 86.23%, and drop to 1.40% with attenuation ratio of 1000×. Since the detection ability can be improved by multiple shots, it can be calculated that the detection range of the system can reach 1km, the detection probability is not less than 1%, and the frame rate can reach 100Hz for a laser repetition frequency of 10kHz. The walk error of the system is within 0.6m under different reflectivity. The walk error is mainly affected by the laser pulse width and light intensity. Further reducing the laser pulse width and increasing the laser output power will help to reduce the walk error. The 3D imaging of the vehicle can be realized at the distance of about 70m, the horizontal spatial resolution is less than 5cm, and the ranging precision is within 10cm after ten shots average by calculating centroid of a histogram. The detection probability can be improved by using the time-gating method. After multiple measurements, the 120m “laser imaging through-window “ can be realized in sunlight.
The system has the advantages of small volume, light weight and fast detection speed. It can be used in the field of autopilot and spacecraft flight control. Its high-speed and high sensitivity characteristics can further exploit other potential applications, which will help to facilitate the continuous development of this technology.
Acknowledgements
The authors acknowledge insightful discussions with Dr. Linhai Huang, Mr. Zhenbao Liu, Dr. Zihan Yi, Mr. Chong Li, Mr. Zhuoli Naer, Dr. Yilong Zeng, Dr. Caihui Zhu, Dr. Haotian Liu, Ms. Hui Chen and Dr. Simone Wang. Furthermore, the authors would like to express sincere thanks for supports by Berxel and Gudatek 2019 Corporation.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. T. Mizuno, H. Ikeda, T. Nagano, T. Baba, M. Mita, Y. Mimasu, and T. Hoshino, “Three-dimensional Image Sensor with MPPC for Flash LIDAR,” Trans. Jpn. Soc. Aeronaut. Space Sci. 63(2), 42–49 (2020). [CrossRef]
2. T. Mizuno, H. Ikeda, K. Makino, Y. Tamura, Y. Suzuki, T. Baba, S. Adachi, T. Hashi, M. Mita, Y. Mimasu, and T. Hoshino, “Geiger-mode three-dimensional image sensor for eye-safe flash LIDAR,” IEICE Electron. Express 17(11), 20200152 (2020). [CrossRef]
3. A. J. Brown, T. I. Michaels, S. Byrne, W. Sun, T. N. Titus, A. Colaprete, M. J. Wolff, G. Videen, and C. J. Grund, “The case for a modern multiwavelength, polarization-sensitive LIDAR in orbit around Mars,” J. Quant. Spectrosc. Radiat. Transfer 153, 131–143 (2015). [CrossRef]
4. A. J. Brown, “Equivalence relations and symmetries for laboratory, LIDAR, and planetary Mueller matrix scattering geometries,” J. Opt. Soc. Am. A 31(12), 2789–2794 (2014). [CrossRef]
5. T. Baba, Y. Suzuki, K. Makino, T. Fujita, T. Hashi, S. Adachi, S. Nakamura, and K. Yamamoto, “Development of an InGaAs SPAD 2D Array for Flash LIDAR,” in QUANTUM SENSING AND NANO ELECTRONICS AND PHOTONICS XV, (2018).
6. Z. M. Li, H. Zhou, Z. Y. Li, Z. Q. Yan, C. Q. Hu, J. Gao, S. B. Wu, S. Y. Yin, and X. M. Jin, “Thresholded single-photon underwater imaging and detection,” Opt. Express 29(18), 28124–28133 (2021). [CrossRef]
7. F. Villa, F. Severini, F. Madonini, and F. Zappa, “SPADs and SiPMs Arrays for Long-Range High-Speed Light Detection and Ranging (LiDAR),” Sensors 21(11), 3839 (2021). [CrossRef]
8. G. Chen, C. Wiede, and R. Kokozinski, “Data Processing Approaches on SPAD-Based d-TOF LiDAR Systems: A Review,” IEEE Sensors J. 21(5), 5656–5667 (2021). [CrossRef]
9. C. Niclass, M. Soga, H. Matsubara, and S. Kato, “A 100m-range 10-frame/s 340×96-pixel time-of-flight depth sensor in 0.18µm CMOS,” in 2011 Proceedings of the ESSCIRC (ESSCIRC), 2011), 107–110.
10. M. Perenzoni, D. Perenzoni, and D. Stoppa, “A 64 $\times$ 64-Pixels Digital Silicon Photomultiplier Direct TOF Sensor With 100-MPhotons/s/pixel Background Rejection and Imaging/Altimeter Mode With 0.14% Precision Up To 6 km for Spacecraft Navigation and Landing,” IEEE J. Solid-State Circuits 52(1), 151–160 (2017). [CrossRef]
11. S. W. Hutchings, N. Johnston, I. Gyongy, T. A. Abbas, N. A. W. Dutton, M. Tyler, S. Chan, J. Leach, and R. K. Henderson, “A Reconfigurable 3-D-Stacked SPAD Imager With In-Pixel Histogramming for Flash LIDAR or High-Speed Time-of-Flight Imaging,” IEEE J. Solid-State Circuits 54(11), 2947–2956 (2019). [CrossRef]
12. H. Seo, H. Yoon, D. Kim, J. Kim, S. J. Kim, J. H. Chun, and J. Choi, “A 36-Channel SPAD-Integrated Scanning LiDAR Sensor with Multi-Event Histogramming TDC and Embedded Interference Filter,” in 2020 IEEE Symposium on VLSI Circuits, 2020), 1–2.
13. P. Padmanabhan, C. Zhang, M. Cazzaniga, B. Efe, A. R. Ximenes, M. J. Lee, and E. Charbon, “7.4 A 256×128 3D-Stacked (45 nm) SPAD FLASH LiDAR with 7-Level Coincidence Detection and Progressive Gating for 100m Range and 10klux Background Light,” in 2021 IEEE International Solid- State Circuits Conference (ISSCC), 2021), 111–113.
14. O. Kumagai, J. Ohmachi, M. Matsumura, S. Yagi, K. Tayu, K. Amagawa, T. Matsukawa, O. Ozawa, D. Hirono, Y. Shinozuka, R. Homma, K. Mahara, T. Ohyama, Y. Morita, S. Shimada, T. Ueno, A. Matsumoto, Y. Otake, T. Wakano, and T. Izawa, “7.3 A 189×600 Back-Illuminated Stacked SPAD Direct Time-of-Flight Depth Sensor for Automotive LiDAR Systems,” in 2021 IEEE International Solid- State Circuits Conference (ISSCC), 2021), 110–112.
15. Y. Tang, R. Yang, J. Qiu, and K. Liu, “Wide dynamic laser ranging based on diodelaser and photon counting techniques,” Appl. Opt. 60(9), 2716–2721 (2021). [CrossRef]
16. R. Yang, Y. Tang, Z. Fu, J. Qiu, and K. Liu, “A Method of Range Walk Error Correction in SiPM LiDAR with Photon Threshold Detection,” Photonics 9(1), 24 (2022). [CrossRef]
17. E. Vilella and A. Dieguez, “Dynamic range extension of SiPM detectors with the time-gated operation,” Opt. Express 22(10), 12007–12012 (2014). [CrossRef]
18. L. Cohen, E. S. Matekole, Y. Sher, D. Istrati, H. S. Eisenberg, and J. P. Dowling, “Thresholded Quantum LIDAR: Exploiting Photon-Number-Resolving Detection,” Phys. Rev. Lett. 123(20), 203601 (2019). [CrossRef]
19. E. Lantz, “Comment on “Thresholded Quantum LIDAR: Exploiting Photon-Number-Resolving Detection”,” Phys. Rev. Lett. 124(19), 199302 (2020). [CrossRef]
20. J. Hu, “The index and testing of extinction ratio for pulsed laser rangefinder,” in International Conference on Optoelectronic Science and Engineering’90, (SPIE, 1990), 186–188.
21. K. Tan, X. Cheng, X. Ding, and Q. Zhang, “Intensity data correction for the distance effect in terrestrial laser scanners,” IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing 9(1), 304–312 (2016). [CrossRef]
22. J. Folkesson, H. Chang, and N. Bore, “Lambert’s Cosine Law and Sidescan Sonar Modeling,” in 2020 IEEE/OES Autonomous Underwater Vehicles Symposium (AUV), (IEEE, 2020), 1–6.
