Neural-network-refined fusion of LiDAR and stereo depth for precise forest DBH estimation

Forests occupy 30.8% of Earth’s land surface and are essential for resource provision, carbon storage, and maintaining biodiversity. Accurate mapping and continuous monitoring are crucial for their effective conservation and management. While traditional techniques depend on manual data collection and statistical analysis, recent advances have shown that mobile robots equipped with Light Detection and Ranging (LiDAR) or stereo cameras can significantly enhance forest mapping. Additionally, combining LiDAR and stereo camera data has yielded positive outcomes in urban settings. This research investigates the fusion of LiDAR and stereo camera information to enhance the precision of forest mapping, with a particular focus on estimating the diameter at breast height (DBH) of trees. The proposed approach integrates initial depth measurements derived from stereo images with LiDAR readings and applies a neural network for further refinement. Experimental findings show that this fusion method reduced the DBH estimation error by approximately 33%, confirming the hypothesis that sensor fusion leads to greater mapping accuracy.