3D Terrestrial lidar data classification of complex natural scenes using a multi-scale dimensionality criterion: applications in geomorphology

3D point clouds of natural environments relevant to problems in geomorphology\noften require classification of the data into elementary relevant classes. A\ntypical example is the separation of riparian vegetation from ground in fluvial\nenvironments, the distinction between fresh surfaces and rockfall in cliff\nenvironments, or more generally the classification of surfaces according to\ntheir morphology. Natural surfaces are heterogeneous and their distinctive\nproperties are seldom defined at a unique scale, prompting the use of\nmulti-scale criteria to achieve a high degree of classification success. We\nhave thus defined a multi-scale measure of the point cloud dimensionality\naround each point, which characterizes the local 3D organization. We can thus\nmonitor how the local cloud geometry behaves across scales. We present the\ntechnique and illustrate its efficiency in separating riparian vegetation from\nground and classifying a mountain stream as vegetation, rock, gravel or water\nsurface. In these two cases, separating the vegetation from ground or other\nclasses achieve accuracy larger than 98 %. Comparison with a single scale\napproach shows the superiority of the multi-scale analysis in enhancing class\nseparability and spatial resolution. The technique is robust to missing data,\nshadow zones and changes in point density within the scene. The classification\nis fast and accurate and can account for some degree of intra-class\nmorphological variability such as different vegetation types. A probabilistic\nconfidence in the classification result is given at each point, allowing the\nuser to remove the points for which the classification is uncertain. The\nprocess can be both fully automated, but also fully customized by the user\nincluding a graphical definition of the classifiers. Although developed for\nfully 3D data, the method can be readily applied to 2.5D airborne lidar data.\n