Canopy Density of Broad-leaved Forest by Watershed Segmentation Algorithm

Four groups were photographed at different height and distance on 5 sample plots of broad-leaved forest with 50 m×70 m in Linan, Zhejiang province by unmanned aerial vehicle. They were made into orthophotoquad, which were pretreated by gray level and median filter, tree canopy was extracted by improved tag-control watershed segmentation algorithm, and compared with that by manual one. The experiment showed that algorithm had higher accuracy, and the error was about 5%. Canopy density was calculated by line transect method on canopy estrated, and the results demonstrated that the canopy density between 0.5 and 0.9 was accurate.