Research On Underwater Image Pre-Processing Based On Entropy And Recognition Of Underwater Object

The subject stems from the project of "Information processing and understanding technology of underwater optical vision" in military intelligent underwater robot technology. It is important to smoothly perform the tasks of detecting the underwater objects and getting their position in the complex condition of the underwater.The purpose of thesis is mainly concentrated on the pre-processing of the underwater image. Including image enhancement, edge detection, image segmentation. Then binary image that used to be extracted features can be obtained. At last finishing recognition by Back Propagation Neural Net.The main work of the thesis is that the entropy information is widely used in the pre-processing of underwater image. Firstly, the method of Fuzzy Enhancement of Underwater Image based on Relative Entropy is proposed. It well deals with the problem of defining the cross point using the combination of the theory of image segmentation and relative entropy. Secondly, according to the characteristic of underwater image a new method of edge detection based on the entropy operator is proposed. It not only removes the noise of the image, but also well detects the edges of the image by the information of the windows of the underwater image. At the step of the pre-processing the Particle Swarm Optimization algorithm is introduced to deal with the problem of the time-consuming computation using the entropy to segment the underwater image. And the efficiency is much higher than the original methods.After the feature extraction and selection, the eigenvectors with better clustering effect are obtained. The influence of scale factor on moment invariant features in discrete are considered, six moment invariants with scale, translating and rotating invariance are proposed.Improved Back Propagation Neural Net method is applied to recognize underwater objects. To make the BP efficient, adaptive learning speed and momentum are introduced. The experimental results show that the precision of...