An Improved Strategy for Object-Oriented Multi-Scale Remote Sensing Image Segmentation

In order to enhance the accuracy of classification of high-resolution remote sensing data, it is necessary to utilize the rich scale-dependent information and geographical information contained in high-resolution image. In this paper, an improved object oriented multi-scale image segmentation method based on the mean shift and the fractal net evolution approach (FNEA) is introduced. In this method, the useful information contained in high-resolution image is utilized, and the efficiency of segmentation is also considered. Besides, the experiment in this paper shows that it is effective for extracting feature ground object whose feature is obvious in the original image. While, after this improved segmentation, the number of image object is less compare with the traditional FNEA, which would be helpful for the following classification.