Marine Target Detection and Recognition Method Based on YOLO Neural Network in Embedded System

This paper introduces a marine target detection and recognition system based on an embedded platform. The frame difference method is used to perform a series of image processing on two consecutive frames of images to detect marine moving objects and the YOLO network is used to complete the identification of marine targets, so as to achieve accurate detection and identify of marine objects. This paper optimizes the The recognition function of the YOLO network uses the recognition frame as a frame to crop and store the recognized marine target image and name it with the time, category and confidence of the recognized target. By deploying the automatic sampling module of moving objects to embedded devices near the coast, automatic sampling and storage are realized, which enriches the data set of the recognition model and realizes the local transmission and storage of the data set through intranet penetration.