Advantages of using object-specific knowledge at an early processing stage in the detection of trees in LIDAR data

Abstract

In some imaging setups the following assumptions hold: the objects are opaque and viewed only from one point, their surface is continuous at least piecewise, and the occluding objects are small with respect to the viewed objects. In addition, in the application of our interest the images can be treated similarly to the case of the plane of light. This made it possible to design algorithms with some desired features: the segmentation based on sorting the data according to angle and the version of the object verification method using fuzzy voting with the positive and negative evidence. The algorithms have some opposite and complementary features which could be used in application to LIDAR data in the measurements of trees and forest.