Identifying treetops from aerial laser scanning data with particle swarming optimization

Abstract

In this study, the particle swarming optimization procedure was applied to parametrize two Local Maxima (LM) algorithms in order to extract treetops from LiDAR-data in a test area (10 km2) of heterogeneous forest structures of conifers in the Alps. The obtained results were compared with those of a widely used variable-size window LM algorithm calibrated using literature values. Quantitative statistical parameters like matching, extraction, omission, and commission rates were calculated. The experimental results showed the effectiveness of the proposed method, which was capable to identify the 91% of the trees and estimate the 92% of the real above ground biomass with a total extraction rate close to 1. Almost all the dominant and codominant trees were extracted, while the extraction rate of the dominated trees averaged over 50%

Keywords

• Lidar
• automatic calibration
• forest inventory
• single tree extraction
• matching
• airborne laser scanning