Research On Underwater Image Recovery And Sea Urchin Recognition Based On Machine Learning

At present,with the increasing demand for aquatic products,it is a critical way to research and use underwater robots to detect marine organisms and complete the salvage of aquaculture products in the future.Sea urchin is one of the most nutritious and essential aquatic products.Its living environment is very complex,so it can be used to represent most of the marine products.Therefore,in this paper,sea urchin is selected as the experimental object of an underwater robot for aquaculture research,and the related problems of sea urchin detection and recognition are studied.The underwater image will be affected by water quality absorption and light scattering,which leads to the disadvantages of color deviation,low contrast,and low definition.Because these shortcomings will seriously affect the detection and recognition performance of sea urchins,it is necessary to enhance the underwater image and improve the quality of the underwater image.In this paper,the existing underwater image enhancement algorithm is optimized to enhance the underwater image in the view of the problems of color deviation,low contrast,and low definition,so that the algorithm can adaptively enhance the underwater image according to the underwater environment to obtain a more stable image enhancement effect,and can be used in engineering practice to achieve a stable enhancement of the real-time vision of underwater robots.Then,it classifies the objects in the image and locates the two-dimensional coordinate regression frame through the depth learning detection algorithm.Finally,the regression box as the region of interest for visual positioning can significantly improve the accuracy of sea urchin detection and classification.The main contents of this paper are as follows:1.Because of the problems of color deviation and low contrast of the image obtained in the underwater environment,these shortcomings will seriously affect the detection and recognition ability of the sea urchin.In this paper,the restoration of the underwater image is studied,and a dark channel first underwater image enhancement algorithm based on optimization is proposed.Firstly,the degradation modeling of the color deviation and atomization of the underwater image is carried out.In this study,a strategy of minimizing the difference between light and dark channels is proposed to obtain the depth of field image.This strategy is convenient for estimating the background color of the water body and obtaining the transmitted image.Then,we use the adaptive factor and depth of field factor selection strategy to post-processing the transmission image to get a better image restoration effect.In addition,the method of color correction is used to remove the color deviation and improve the overall brightness of the picture.Finally,the GSA optimization algorithm is used to automatically optimize the image quality by way of no-reference image quality score,so as to find the optimal value of the depth of field factor.In this paper,although the proposed ODCP algorithm can remove the color deviation and improve the contrast of the underwater image,it appears the phenomenon of over red compensation,resulting in image distortion.2.In view of the over red compensation problem of the ODCP algorithm in underwater image processing,which will cause image distortion,this paper proposes an improved multiscale Retinex underwater image enhancement algorithm.Based on the correlation between the underwater background and the scenery,this algorithm proposes to use the gray color scale of each channel as the scenery correlation image.By using the multi-scale Retinex enhancement algorithm to enhance the RGB grayscale of each channel of the underwater image,and to stretch the color of the three channels to make the enhanced image conform to the normal distribution,and then uses the scenery correlation image to enhance the image color restoration.In addition,the color correction algorithm is also needed to remove the color deviation so as to enhance the underwater image further.Finally,the GSA algorithm is used to optimize the image quality automatically to find the optimal parameters of the Gaussian weighting factor.In this paper,the IMSR underwater image enhancement algorithm can not only remove the color deviation and improve the contrast of the underwater image but also solve the problem of over red compensation in the ODCP algorithm.3.The underwater environment has problems with low contrast and color deviation,which makes the detection and recognition of sea urchins more difficult.In order to solve this problem,this paper combines the OMSR underwater image enhancement algorithm and SSD target detection algorithm to realize the detection and recognition of underwater sea urchins.In order to solve the problem of the poor detection effect of the small target in the SSD algorithm,this paper proposes the ISSD algorithm to enhance the detection ability of the small target.ISSD uses resnet50 as the basic network architecture,which can improve the ability of feature expression.At the same time,this paper also proposes a feature cross-level fusion method,which can strengthen the semantic information of context.Finally,according to the shape and texture characteristics of sea urchins,the edge feature enhancement strategy is proposed,which can strengthen the edge texture information of sea urchins,and improve the detection performance of sea urchins by combining the ISSD algorithm.In this paper,the proposed ISSD algorithm can improve the detection performance of small targets.OMSR image enhancement algorithm and edge multi-channel fusion strategy can improve the accuracy and robustness of underwater sea urchin detection.