Research on Innovation and Model of Underwater Visual Detection System for Nuclear Contaminated Seafood Based on Deep Learning

In this paper, an innovative underwater visual detection system for nuclear-contaminated seafood is proposed, aiming to achieve accurate identification and detection of nuclear-contaminated seafood in the Marine environment by using deep learning technology. The system combines advanced color texture feature fusion technology and linear discriminant analysis to effectively improve the accuracy and robustness of image recognition. By training the deep neural network, the model can extract the features of seafood from the complex underwater environment, including shape, texture, and color, and then determine whether it is contaminated by nuclear. In order to improve the detection efficiency, the K-nearest neighbor (KNN) algorithm is introduced to quickly locate suspected contaminated seafood by comparing with the features of known nuclear contaminated seafood samples. The experiment was conducted in a real Marine environment, and many seafood samples were collected for model training and testing. The simulation results show that the accuracy of the system in detecting nuclear contaminated seafood reaches 95% and the recall rate reaches 93%, which is significantly better than the traditional detection method. Data analysis shows that the deep learning model can effectively distinguish between normal and contaminated seafood, providing a strong technical support for Marine ecological security and food safety. This study provides a new idea for the application of underwater visual inspection technology in Marine environment monitoring, which has important theoretical and practical value.