Individual Tree Crown Extraction from High Resolution Image Based on a Threshold-Improved Watershed Algorithm

In order to achieve 3D modeling of campus scenes and monitor the growth of campus trees, completing individual tree crown segmentation is an important research content. The purpose of this study is to perform individual tree segmentation on high-resolution campus scene images taken by unmanned aerial vehicle (UAV) to improve the accuracy. Therefore, this study proposes an improved watershed algorithm based on distance threshold and area threshold. In this way, over-segmentation and incorrect segmentation problems are improved. And the final accuracy of segmentation in both dense and sparse areas of tree distribution reaches above 86%, which shows that the method is robust and effective. It has practical applications in such areas as forest resource management.