Study On Segmentation And Recognition Algorithms For Underwater Images Based On Fuzzy Theory

With the rapid progress of exploring the sea, manned submersible plays an important role in the exploration of sea. The system of computer vision is one of the important tools to get the environment information. So advanced techniques of information manipulation and pattern recognition are needed, so as to utilize the valuable information. This paper studies on the subject of national 863 project –7000 meters manned submersible special task (the target recognition technique of undersea hydrothermal vent image). Because of the speciality of imaging environment in the deep sea, the quality of image is poor, traditional image technique cannot meet the need of application. Since the appearance of fuzzy theory, fuzzy techniques have developed rapidly, and have used in various domains. Fuzzy techniques transform fuzzy issues into certain ones or ones those can be coped with easily by means of fuzzy transformation and fuzzy inference. They show better performance in coping with fuzzy object compared with traditional techniques. In this paper, study of applying fuzzy technique to image processing are carried out, the main contents are described as follows: 1. Study on the way of deciding the parameters in the membership function and propose a fuzzy enhancement algorithm base on maximum entropy. This algorithm uses maximum entropy as criterion for parameters selection in the membership function, so the parameters can be decided adaptively according to different targets. This algorithm can mainly remain the original information after fuzzy transformation, and at the same time, it can make the difference between background and target apparent, the segmentation can be done easily. So this algorithm has good adaptability. 2. Propose an information combination method of gray and chroma based on OTSU in order to overcome the impact of the uneven illumination and other disadvantage factors. This method exploits the usable information fully and facilitates the segmentation. 3. Discuss the algorithms for underwater target recognition by way of texture analysis. on the basis of analyzing all kinds of texture analysis method, extract some texture features those are sensitive to underwater targets as classification characters. At last, an effective statistic-fuzzy classification method is employed, and this method distinguishes the chimney from others well. 4. Integrating cell algorithms above, the whole segmentation and recognition procedure of chimney for undersea hydrothermal vent image is proposed.