Selection of the Optimal Segmentation Scale in High-resolution Remote Sensing Image

Abstract

The selection of the optimal segmentation scale is of great significance in high-resolution remote sensing image multiple scale segmentation, which directly influences the information extraction precision from image. In order to find the optimal segmentation scale, a calculation model of the optimal segmentation scale in terms of several factors have been suggested. This method firstly calculates the best segmentation scale on the consideration of multi-band information, therefore, avoiding the potential subjectivity of human visual. Secondly, the quality evaluation function is put forward using the method of samples controlling which considering area and perimeter as factor s to evaluate consistency between segmented objects and samples. This method of the optimal segmentation scale based on multiple scale and the evaluation of the consistency between the sample and the segmented object in terms of the spectral band is proposed and verified. Results show that the method can obtain the optimal segmentation scale of remote sensing image quickly and efficiently. The algorithm of the model is simple, and is easy to realize, and it is a practical algorithm model.