Abstract:Forest harvesting is a forest carbon source. Accurate estimation of forest harvesting biomass is helpful for accurate measurement of forest carbon sinks. Aiming at the challenging problem of using single timephase visible light UAV image to estimate the biomass of highdensity forest harvesting, a high-precision estimation method of forest harvesting biomass was studied based on multi-temporal visible light UAV image before and after logging. Taking a coniferous forest in Fuzhou City of Fujian Province Baisha forest cutting small class as the experimental zone, collecting resolution better than 10cm long before and after cutting, unmanned aerial vehicle (UAV) visible light image, the local maximum dynamic window method was adopted to get high precision of cutting plants and single tree height information, and then based on the UAV image after cutting, detection and extraction by the method of YOLO v5 cut pile diameter of information, the DBH information of the cut wood was estimated according to the DBH-pile diameter model, and the biomass of the cut wood was estimated by using the binary biomass formula of tree height and DBH, which was verified by the measured data. The precision of tree number and average tree obtained by dynamic window local maximum method was 96.35% and 99.01%, respectively. The overall accuracy of pile cutting target detection by YOLO v5 method was 77.05%, and the accuracy of average DBH estimated by pile cutting diameter was 90.14%. Finally, the accuracy of forest harvesting biomass was 83.08%. The results showed that this method had great application potential. Using multi-temporal UAV visible light remote sensing before and after harvesting can realize effective estimation of forest harvesting biomass, which can help to reduce the cost of manual investigation, and provide effective technical support for the government and relevant departments to accurately measure carbon sinks.
