In order to accurately monitor the growth status of trees in forest surveys, single tree crown segmentation is an important research topic. Taking the spruce forest image taken by UAV as the research object, aiming at the problem of crown adhesion in the image, combining watershed algorithm and Chan Vese Active contour model, a set of automatic crown contour positioning, segmentation and optimization system was established. This improved algorithm inputs the preprocessed image into the labeled watershed segmentation, and the resulting watershed serves as the basis for the CV model. The curve is evolved to a specific edge of the image, maximizing the extraction of the complete crown contour. The research method in this paper solves the problem of serious over segmentation and under segmentation of watershed algorithm, and the problem of difficult to determine the initial location of regional Active contour model, thus obtaining a convenient and complete algorithm for tree crown segmentation. The experimental results show that compared with traditional watershed and Chan Vase models, the accuracy of the crown extraction method studied in this paper reaches 82.62%, achieving a high improvement. And there are good improvements in addressing issues such as crown misjudgment, missed judgment, under segmentation, and over segmentation. This method can automatically and effectively extract independent tree crown contours with better extraction accuracy.