Using Bispectral Full-Waveform Lidar to Map Seamless Coastal Habitats in 3d

Abstract

Abstract. Mapping coastal habitats is essential to their preservation, but the presence of water hinders seamless data collection over land-water interfaces. Thanks to its dual-wavelength and optical properties, topo-bathymetric lidar can address this task efficiently. Topo-bathymetric lidar waveforms contain relevant information to classify land and water covers automatically but are rarely analysed for both infrared and green wavelengths. The present study introduces a point-based approach for the classification of coastal habitats using bispectral waveforms of topo-bathymetric lidar surveys and machine learning. Spectral features and differential elevations are fed to a random forest algorithm to produce three-dimensional classified point clouds of 17 land and sea covers. The resulting map reaches an overall accuracy of 86%, and 65% of the prediction probabilities are above 0.60. Using this prediction confidence, it is possible to map coastal habitats and eliminate the classification errors due to noise in the data, that generate a clear tendency of the algorithm to over-estimate some classes at the expense of some others. By filtering out points with a low prediction confidence (under 0.7), the classification can be highly improved and reach an overall accuracy of 97%.

3DMASC: accessible, explainable 3D point clouds classification. Application to bi-spectral topo-bathymetric LiDAR data.

“…To the best of our knowledge, no research on the application of 3D neural networks to TB lidar data has yet been published. Existing approaches rely exclusively on handcrafted features extracted on full-waveforms (Launeau et al, 2018;Letard et al, 2022b), rasters (Wedding et al, 2008), or PCs (Hansen et al, 2021).…”

“…Their ease of use, efficiency, robustness to overfitting, generalization abilities and production of a feature importance metric (Breiman, 2001;Pal, 2007) explain their frequent use for 3D data classification. They have been used successfully in (Chehata et al, 2009;Hansen et al, 2021;Letard et al, 2022bLetard et al, , 2022a for point-based classifications of both topographic and TB lidar. In RF, since the decision trees are independent, one cannot compensate the potential weaknesses of another to improve the global performance of the forest.…”

“…To further improve the robustness of the classifiers, we also exploit the prediction probability output by RF and use it as a classification confidence indicator, as seen in (Brodu and Lague, 2012) and (Letard et al, 2022b). The prediction probability corresponds to the proportion of forest trees that voted for the class assigned to the point.…”

“…As a result, vertically highly dense data is condensed into equally spaced punctual values, losing the spatial point pattern information. Few studies provide 3D classifications of underwater environments using bathymetric lidar (Hansen et al, 2021;Letard et al, 2022bLetard et al, , 2022a. Additionally, most of them require full-waveform information (Letard et al, 2022b(Letard et al, , 2022a, which is complex to process and often unavailable or unpublished.…”

“…Few studies provide 3D classifications of underwater environments using bathymetric lidar (Hansen et al, 2021;Letard et al, 2022bLetard et al, , 2022a. Additionally, most of them require full-waveform information (Letard et al, 2022b(Letard et al, , 2022a, which is complex to process and often unavailable or unpublished. On the land covers, TB lidar datasets offer potentially new classification opportunities using the bispectral backscattered intensity.…”