Research on Segmentation Scale of Multi-Resources Remote Sensing Data Based on Object-Oriented

To verify a method of choosing the optimal segmentation scale, the Guangzhou Higher Education Mega Center is selected as test area for surface features extraction. Buildings, roads, waters and vegetations are extracted from 4 different resolution images using different scales. It is found that the optimal segmentation scale selected based on standard deviation of the means and mean of standard deviations of image objects’ brightness is almost consistent with the optimal segmentation scale selected based on actual classification. The classification accuracy is related with spatial resolution. But for specific applications, higher resolution doesn’t definitely get better classification accuracy.