Research And Design On Target Recognition Technology Of Submersible Optical Vision

There is a lot of mineral resource and bio-energy in the vast ocean which is a valuable asset to the development of human society. However, the development and utilization of marine resources is very limited due to the hard and complex environment of the ocean. According to the feature of underwater image, more attention is paid on the further study of the target recognition technology for submersible vehicle optical vision. It has becoming the critical link of intelligent process to design the intelligent submersible vehicle which has the skill of automatic recognition.This paper analyses the present and development situation of submersible optical vision technology and the main problems in the area first and then research on the technologies of image processing and target recognition. On the basis of deeply researching on the structure of the submersible vehicle's optical vision system, the main solution of the target recognition technology and the main problems in designing the software modules have been put forward.Analysis of the features of underwater images deeply, design the contrast enhancement algorithm based on histogram and image filter algorithm according to the problem of low contrast and a great deal of water molecules noise. The technology improves the quality of underwater images effectively.Focused on analyzing the segmentation problems of underwater images, research and design improved Sobel edge detection solution and threshold segmentation algorithm based on Otsu. According to the features of underwater images, select the invariant moments as the characteristic quantity of the underwater images. Analyzing the clustering ability of invariant moments in the target recognition process, invariant moments are proved to have the ability of distinguishing different targets.On the basis of researching on the factors which impact the accuracy of underwater target recognition, the target recognition algorithm based on neural network has been put forward. Setting up underwater target recognized experiment environment based on VC++6.0, this paper has tested the research results. The experiment shows that the solution which has been put forward in the paper can recognize the underwater target accurately and the operation of the system is stable.