Filtering vegetation from terrestrial point clouds with low-cost near infrared cameras

In applications relating to the reconstruction of a rock face's surface by Terrestrial Laser Scanning (TLS), the overgrown vegetation does not allow one to correctly accomplish this task. In standard Airborne Laser Scanning surveys, the vegetation is filtered out by using spatial filters that exploit the availability of multiple echoes. The same approach does not work efficiently in the case of a rock face. This is due to the morphological complexity, that is typical of such surfaces. For this reason a new system for the automatic recognition of vegetation using a NI R camera was designed and implemented. It is based on a set of images acquired with a low-cost SLR digital camera modified to capture also the NIR component. Such camera is integrated and calibrated with respect to the TLS sensor. A vegetation filter based on the analysis of the NIR component allows one to locate vegetated areas, that can be automatically removed from the data set. In the paper we would like to give an introduction to the procedure used for camera setup, calibration, and the filtering algorithms implemented.