Segmentation Algorithm For Road In High Resolution Remote Sensing Images

The application of high resolution remotely sensed data has come into a new era since SPOT-5, IKONOS and QuickBird satellites were launched. As a result, new methods are required to obtain information from the high-resolution imagery. Extracting road features automatically from remote sensing images has proved to be a difficult task. It is also one of the key problems of photogrammetry and remote sensing, which plays a crucial role on photogrammetry and automating the analysis of the remote sensing image. A binary road image can be obtained by means of segmentation, and extracting roads in such an image is of course easier than in a gray one. In fact, extracting roads in a binary image is a representative kind of methods.In this paper, we mainly pay our attention to the segmentation algorithms of the road images. The specific characteristics of the typical urban roads in the high resolution remote sensing images for typical urban roads are considered, and the corresponding algorithms have been used to distinguish the road from other objects. This process provides the preparation for the final extraction. Major research contents are the following:(1) Some of the existing segmentation approaches for high resolution remote sensing images have been analyzed in detail, and their advantages and disadvantages have been summed up.(2) The segmentation algorithm for multi-category target images which is based on OTSU has been researched carefully, and its limitations while being used to segment roads directly has been summarized as well.(3) Two new methods suitable for the urban high resolution remote sensing images have been proposed. The first is base on the gray consistency of the regional road and the second is based on the integrated features.The method based on the gray consistency of the regional road uses a combination of gray and space position information to segment images. Experimental results show that this method is more effective, as it can be used to deal with the complex high resolution remote sensing images rapidly.The method based on the integrated features uses Fuzzy clustering for images segmentation, which overcomes the limitations of the original use of a single character.