Abstract
The capability of the parameters derived from waveform data in discriminating objects is assessed and the effect of the relative calibration of full-waveform data in discriminating land-cover classes is evaluated. Firstly, a non-linear least-squares method with the Levenberg-Marquardt algorithm is used to fit the return waveforms by a Gaussian function. Gaussian amplitude, standard deviation, and energy are extracted. Secondly, a relative calibration method using the range between the sensor and the target based on a radar equation is applied to calibrate amplitude and energy. The change in transmit pulse energy is also considered in this process. A support vector machine classifier is used to distinguish the study area into non-vegetated area (including roads, buildings, and vacant lots), grassland, needle-leaf forests, and broad-leaf forests. The overall classification accuracy ranges from 79.33% to 87.6%, with the combination of the two groups of the three studied parameters. Calibrated data classification accuracy is improved from 1.20% to 6.44%, thus resulting in better forest type discrimination. The result demonstrates that the parameters extracted from the waveforms can be applied effectively in identifying objects and that relative calibrated data can improve overall classification accuracy.
© 2013 Chinese Optics Letters
PDF Article
More Like This
Cited By
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
Contact your librarian or system administrator
or
Login to access Optica Member Subscription