Research On Stereoscopic Matching Algorithm Of Underwater Binocular Visual Images

With the gradual improvement of international trade,the shipping industry has also developed rapidly.However,it cannot be ignored that the development of the shipping industry is accompanied by a sharp increase in the number of maritime traffic accidents.The primary task of handling maritime accidents is to understand the sea environment around the accident.Especially for the shipwreck accident,because the underwater environment of the accident site is unknown and complex,the traditional manual search and rescue mode can neither guarantee the absolute safety of search and rescue personnel nor detect the dead angle.However,unmanned detection equipment such as underwater robot is small in size and highly flexible.Its binocular optical detection system can obtain the distance from the target to the underwater robot through stereo matching technology,so as to determine the position of the trapped.In addition,the underwater robot has strong detection ability and can detect dead corners that cannot be detected manually,taking into account safety and reliability,and has gradually become the protagonist of underwater search and rescue.The underwater robot equipped with binocular detection system is also widely used in underwater projects such as resource detection,submarine optical cable detection,and so on,with broad prospects.However,there are difficulties in the development of underwater robots,mainly in two aspects.First,due to the complex underwater environment,the underwater optical detection often has problems of quality degradation due to the underwater environment’s absorption,scattering and attenuation of light,such as blurring and color distortion,which brings difficulties to the subsequent stereo matching work.The second is that the current stereo matching technology of underwater images cannot meet the needs of both matching effect and matching speed.Starting from meeting the requirements of underwater image stereo matching technology for matching effect and matching speed,this thesis focuses on underwater image enhancement and underwater image stereo matching.In underwater image enhancement,by studying the transmission characteristics of light in water and underwater imaging model,the image enhancement method based on Retinex model is used for image processing and underwater image experiment,Experimental results show that this algorithm is superior to other algorithms in underwater image enhancement.In the aspect of underwater image stereo matching,a stereo matching algorithm based on image segmentation and multi-cost fusion is proposed according to the characteristics of underwater environment,and experiments are carried out using underwater images.The results show that this algorithm greatly reduces the matching time while maintaining the matching effect,and provides a theoretical possibility for real-time stereo matching.The innovative achievements of this thesis include:(1)By analyzing the underwater imaging model and the principle of image enhancement algorithm,it is concluded that the image enhancement method based on Retinex model is more suitable for underwater images,and it is proved by experiments.(2)According to the characteristics of underwater images,a stereo matching algorithm based on image segmentation and multi-cost fusion is proposed.The experiment shows that this algorithm is better than other algorithms.