Visual Detection And Recognition Of Underwater Target

With the rapid economic development and population growth, resources on land get fewer, underwater resources development in the future has been placed in an important position. With the development, underwater activities, such as undersea exploration, monitoring of oil platforms, pipeline crack detection, etc. get more frequent. All these operations need target detection and recognition.Because of complex underwater environment, a number of confounding factors such as suspended solids, the images got under water have very low contrast and blurred edges. Experimental results show that some of the traditional method of image enhancement can not make the quality of underwater image better. In this situation, we attempt to enhance the image with MSR (multi-scale Retinex) and dark channel. The method clears the uneven illumination of underwater images using MSR, and then enhances underwater image by dark channel which has achieved remarkable results in the haze removal. Ultimately this method significantly improves image contrast, and makes the edge more clearly.Underwater because of noise and interference fringes, Canny edge detection will detect a large number of false edge and seriously affect target detection. So we change the fixed Canny threshold to variable threshold based on gradient histograms, and track the edge using the continuity of the edge gradient and reduce the false edge.?With moving of the target, the target area in the image will rotate and change size. So before target recognition we get the edge line parameters by Hough transform, according to these parameters rotate the image, and adjust the image size using bicubic interpolation. At last we use PCA (principal component analysis) method to recognize the target.