基于增强后的图像做第一次多尺度分割，将影像内各对象单元进行合并且保证各

Do enhanced image based on the first multiscale segmentation, within the image of each object unit for engagement and guaranteed minimum degree of heterogeneity between objects; based on the result of the first multi-scale segmentation, for each object divided using nearest neighbor rule set distinguishing canopy and ground, the ground select visual interpretation, background sample and crown band configured to mean nearest neighbor feature, classifies [36-38]; merge crown categories based on the extracted object, once again, its multi-scale segmentation, get better segmentation results of individual tree.

Based on the enhanced image to do the first multi-scale segmentation, the object units in the image are combined and the heterogeneity between the objects is minimal, based on the results of the first multi-scale segmentation, the nearest neighbor classification of the divided objects is divided between the crown and the ground, visual interpretation selects the ground background, The sample of the crown and the configuration band mean are the nearest neighbor characteristics, classified, combined based on the extracted crown category objects, and then multi-scale segmentation of them, to obtain the ideal single-wood segmentation results.

Based on the first multi-scale segmentation of the enhanced image, each object unit in the image is merged and the heterogeneity between objects is minimized; based on the results of the first multi-scale segmentation, the tree crown and ground surface are distinguished by the nearest neighbor classification of rule set for each object after segmentation, and the samples of ground background and tree crown are selected by visual interpretation and the band mean value is configured as the nearest neighbor feature Based on the combination of the extracted crown category objects, the multi-scale segmentation is carried out again, and the ideal single tree segmentation results are obtained.