Research And Implementation Of Underwater Image Enhancement And Target Detection Algorithm Based On Deep Learning

Currently,computer vision-based underwater image enhancement and target detection technology plays an important role in marine economic areas such as marine product fishing and marine garbage cleaning.However,due to the complexity of the underwater environment and the absorption and scattering effects of water on light,underwater images often suffer from serious degradation,leading to issues such as blurry details,serious color distortion,and low contrast.Additionally,underwater organisms are usually small,diverse in scale,and densely populated,posing significant challenges to underwater target detection technology.In recent years,with the development of deep learning technology,deep learning-based underwater image enhancement and underwater target detection methods have been widely studied.Deep learning algorithms have the ability to learn features from data,and can achieve better results in underwater image processing.Therefore,this paper analyzes and researches the problems of existing methods for underwater image enhancement and underwater target detection based on deep learning technology.The main work and innovations of this paper are as follows:(1)Building the underwater image enhancement dataset RUIED and the underwater target detection dataset URTD.Existing underwater image enhancement datasets have a single scene,resulting in insufficient model stability and generalization ability.Based on two real underwater datasets,this paper constructs the RUIED dataset.Existing underwater target detection datasets have few types and insufficient samples.In this paper,the URPC challenge datasets from recent years are integrated,and images with marked errors in the dataset are re-marked,constructing the URTD dataset.(2)Proposing a CMFU-Net underwater image enhancement algorithm.Based on the CMF-Net network,this paper improves the U-Net module to fully utilize the detailed information of deep convolution feature layers.Through attention module fusion with shallow feature layers,more highresolution information contained in the feature maps can be retained.The improved attention module uses multiple residual connections to effectively preserve the overall features and local information of underwater images.The channel attention,pixel attention,and spatial attention effectively solve the problems of low contrast,edge details blur,and severe color distortion in the underwater images.The improved loss function module solves the problem of the original network being insensitive to the generated image’s color by adding a color balance loss.The proposed CMFU-Net network not only enhances the color,contrast,and clarity of underwater images but also adapts to different natural scenes,preserving their original content and structural information.It improves the underwater object detection accuracy to a certain extent.(3)Proposing an improved YOLOv8 underwater object detection model.Based on the YOLOv8 model,the improved Neck module improves the problem of missed detection of small marine organisms by adding the CBAM attention mechanism and using CBAM attention to the local information of the image.The improved C2F module adds Coordinate Attention mechanism,which can not only capture inter-channel information of underwater images,but also capture direction perception and positionsensitive information of the image,making the improved YOLOv8 model more accurate in locating and recognizing target areas.By replacing the CIoU loss of the original model with the SIoU loss,the improved loss function module effectively improves the training speed and inference accuracy of the original model.Compared with the original model,the proposed improved model only increases the parameter volume by 0.4MB and the FPS remains almost unchanged,with the average precision and recall rate increased by 2.8%and 1.2%,respectively,fully verifying the effectiveness of the proposed method in this paper.(4)Design and implementation of an underwater image enhancement and object detection system.Based on the PyQt5 framework,this system integrates the proposed underwater image enhancement algorithm and underwater object detection model,providing users with functions like image upload,image enhancement,image detection,video detection,and download.