Research On Underwater Target Recognition System Based On Optical Vision

Underwater target recognition technology is the focus in underwater exploration,and has broad application prospect.Although acoustic technology is dominant in underwater target recognition,optical vision technology is irreplaceable in close range underwater operations because of the more intuitive and richer visual perception of optical images.Consequently,Research on underwater optical visual target recognition technology has important theoretical significance and practical engineering value.The thesis is one of the sub-projects of research on multipurpose intelligent vehicle from institute of marine equipment of Jiangsu university of science and technology.The underwater target recognition techniques are deeply studied,and the main contents include:(1)Construction of hardware platform for underwater target recognition system.Underwater image acquisition and processing system has two kinds: outdoor and indoor.The outdoor collection and processing system uses underwater vision system of self-developed semi submersible unmanned vehicle to capture and handle underwater images under natural light and artificial light.Indoor collection and processing system is an underwater image acquisition and processing system composed of flume,computer,underwater camera and underwater light,it also carries out underwater image acquisition and processing in natural light and artificial light environment.(2)Research and analysis of underwater image enhancement algorithm.On the basis of in-depth analysis of common image enhancement algorithms,an underwater image enhancement method based on the combination of the physical model of underwater imaging and the dark original color prior principle is proposed.In order to improve the anti-jamming,the method of the background light estimation is improved,and the color partial blue and green phenomenon of underwater image is corrected to meet the visual habit of the observer.(3)Research on underwater target segmentation and multi-features extraction.The various edge detection operators are compared and analyzed,and an underwater target segmentation method based on Roberts edge detection operator and improved level set is proposed.In this method,the traditional level set method is improved,and the three color channel of the image is decomposed,mathematical operation and reorganized on the RGB color space,so that the segmented underwater target is more complete.This paper also do research on feature extraction from three aspects,such as shape,texture and color.(4)Design and experimental verification of underwater target recognition method.Aiming at the problems of low recognition rate and poor stability,an underwater target recognition method based on BP neural network is proposed.This method uses the target shape feature extracted by Hu invariant moments,the target texture features extracted by the gray level co-occurrence matrix and the color feature extracted by the color moments as the input of the BP neural network,then do sample training and testing.In order to verify the effectiveness of the work,the indoor image acquisition system is used to collect different azimuth images of underwater targets under natural light environment and artificial light environment,and the image sample library and test library are established,and the recognition experiment is carried out.