A two-stage single-tree crown segmentation method based on deep learning and watershed algorithm

Abstract

A two-stage tree crown segmentation method based on deep learning and watershed algorithm, relating to the field of image processing technology, includes: acquiring optical remote sensing images to be segmented; inputting the optical remote sensing images into a trained deep learning model to obtain preliminary tree crown segmentation maps and tree kernel density maps; generating a spatial constraint mask based on the semantic segmentation results of tree crowns; extracting marker points from the tree kernel density maps; generating a gradient image based on the optical remote sensing images; and, under the constraints of the spatial constraint mask, performing a marker-controlled watershed transformation on the gradient image using the marker points to obtain the final tree crown segmentation map. This method addresses the problem of low accuracy in traditional methods when segmenting tree crowns for different types of forests.

Images