Sensor Fusion of LiDAR and Stereo Camera for Forest Mapping

Abstract

Forests, covering 30.8% global land area, play a critical role in providing resources, storing carbon, and supporting biodiversity. Effective management and conservation require precise mapping and monitoring. Traditional methods rely on manual measurements and statistical techniques, and mobile robots are a promising alternative, with existing works that show good results using LiDAR or Stereo Cameras equipped on terrestrial or aerial robots. This study explores the fusion of LiDAR and stereo camera data to improve the accuracy of forest mapping, measuring the Diameter at Breast Height (DBH) of trees. The proposed method combines initial depth estimation from stereo images with LiDAR data and refines it with a neural network. Experimental results demonstrate that this method reduces DBH estimation error from 0.9cm(3.4 % of the measured actual DBH) using LiDAR alone to 0.6cm(2. 5%), validating the hypothesis that sensor fusion enhances mapping accuracy.