Study And Application Of Algorithms For Remote Sensing Image Segmentation

Remote-sensing image segmentation is the process to extract the objective through processing and analyzing the remote-sensing image. The characters of remote-sensing image are higher gray level, tremendous image information, faint border and complex structure. And there is no reliable model to guide the remote-sensing image segmentation. Therefore, the image segmentation technology is not used in the remote-sensing field well. This thesis has studied the remote-image segmentation algorithm based on watershed transform and the remote-sensing image threshold segmentation algorithm based on genetic algorithm. And the prototype system of remote-sensing image segmentation has been designed and implemented. The work and contributions in this thesis are as follows.First, a survey of remote-sensing image segmentation methods is presented. The author summarizes the remote-sensing image segmentation methods from three aspects systematically, it is referenced that embedded study in the field.Second, a new method of multi-threshold selection method for remote-sensing image segmentation based on genetic algorithm is proposed. With the application of multi-threshold selection method, the effect and speed of image segmentation could be improved. The method is extended in such three aspects: code method, fitness function and the operation of crossover and mutation. The experiment shows that the threshold selection efficiency of our method is rather better than the efficiency of other conventional algorithms.Third, a new smooth method to solve the excessive segmentation is mended. The author implements the V-S algorithm, analyzes the problem of excessive segmentation in image segmentation, and brings forward two improvement methods. One is using the floating-point image to replace the morphologic gradient image to get the more distinct and prominent image border. The other is combination the region using gray and area as the control rules. The experiment shows that the excessive segmentation problem can be solved using our improved algorithm efficiently.Fourth, a prototype system of remote-sensing image segmentation has been implemented. As to the practice requirement, the author modifies the computing mean value method of image gray to increase the computing rate. And the basic operation in image process has been encapsulated using class.