Over the last decade point cloud derived from laser scanner has become the mainstream of the individual tree canopy extraction research. However, the high cost of airborne laser scanning acquisition makes it unsuitable for repeated surveys and small-scale forest mappings. Individual tree canopy was extracted from unmanned aerial vehicle images matching point cloud, aiming to provide a cost-effective method which can complement or even partly replace LiDAR forest mapping in small area. Choosing young Osmanthus and Podocarpus trees growing in a nursery as the study objects, the method was tested in two samples of images matching point cloud. An inexpensive commercial off-the-shelter drone with built-in camera was used to acquire overlapping nadir-viewing images. These images were then used to generate dense point cloud in photogrammetry software. After preprocessing, canopy height model was firstly built from the point cloud;a local maximum filter was applied to detect the canopy positions and marked as the seed points;then the initial area of regional growth can be formed from these seeds;in an iterative calculation process of all image matching points were classified. The canopy contours extracted by the algorithm were inputted into ArcGIS to obtain canopy contour vectors, and were validated by comparing with the manually drawn individual tree crown polygon (reference crown). The F score of segmentation results was higher than 89%, and the errors of individual tree crown diameter extraction results were less than 0.14m (root mean square error) in both sample plots.