Research On Underwater Moving Target Detection And Recognition Technology Based On Optical Vision

The rapid development of underwater optical imaging equipment has made optical vision widely used in underwater monitoring,ocean exploration,underwater security,ecological research and other fields.And underwater target detection and recognition is the focus of the optical vision system.Therefore,it is of great significance and value to study it for academic field and engineering applications.The thesis deeply studied the underwater image enhancement,moving target detection and recognition technology.The main research contents are as follows:Firstly,on the basis of analyzing the principles of several commonly used underwater image enhancement algorithms,an algorithm based on improved CLAHE is proposed to solve the problems of low contrast and uneven illumination caused by the absorption and scattering of light by water medium.The corresponding clipping amplitude is determined according to the pixel distribution of each sub-block to clip and distribute the histogram.The portion beyond the clipping amplitude is distributed in the second time to enhance image details and homomorphic filter is introduced to eliminate uneven illumination.The experimental results show the effectiveness of the proposed algorithm by comparing with other algorithms.Secondly,the commonly used moving target detection algorithms are analyzed and an algorithm based on improved Vi Be is proposed.In order to improve the real-time performance of the algorithm,the Gaussian pyramid is used to scale the video sequence image.Five-frame difference method is combined with the Vi Be algorithm to eliminate ghost regions.The first four frames of video sequence are used to fill the sample set to overcome the shortcomings of traditional Vi Be algorithm in filling the sample set with the first frame image.And the second update mechanism of the sample set is introduced to enhance the adaptability of the background model.The comparison experiment results show the effectiveness of the proposed algorithm.Thirdly,in view of the difficulties of underwater target recognition and the shortcomings of traditional classifiers,tiny-dnn deep learning framework is used to realize underwater target recognition.The basic principles of convolutional neural network and structure of the network in tiny-dnn are studied and analyzed.The corresponding data set is made for the target objects and data augmentation is used to expand the data set to improve the generalization ability of the model.Leaky Re LU function is selected as the activation function through analysis and comparison.And the data set is used to complete the training and testing of the convolutional neural network.Finally,an underwater moving target detection and recognition system is developed by completing its interface design and function realization.The software and hardware test environment is built and the water tank experiment is carried out to test the functions such as video sequence display,underwater image enhancement,moving target detection,target recognition,image storage and other functions.