Impact of optimization of ALS point cloud on classification

Airborne laser scanning (ALS) is one of the LIDAR technologies (Light Detection and Ranging). It provides information about the terrain in form of a point cloud. During measurement is acquired: spatial data (object’s coordinates X, Y, Z) and collateral data such as intensity of reflected signal. The obtained point cloud is typically applied for generating a digital terrain model (DTM) and a digital surface model (DSM). For DTM and DSM generation it is necessary to apply filtration or classification algorithms. They allow to divide a point cloud into object groups (e.g.: terrain points, vegetation, etc.). In this study classification is conducted with one extra parameter–intensity. The obtained point groups were used for digital spatial model generation. Classification is a time and work consuming process, therefore there is a need to reduce the time of ALS point cloud processing. Optimization algorithm enables to decrease the number of points in a dataset. In this study the main goal was to test the impact of optimization on the results of a classification. Studies were conducted in two variants. Variant 1 includes classification of the original point cloud where points are divided in the groups: roofs, asphalt road, tree/bushes, grass. On variant 2 before classification, an optimization algorithm was performed in the original point cloud. Obtained from these two variants object groups were used to generate a spatial model, which was then statistically analyzed.