Underwater sound image detection tracking method based on deep learning

The invention discloses an underwater sound image detection tracking method based on deep learning, and belongs to the technical field of image detection tracking. The method comprises the following steps of: 1, improving a Yov7 algorithm into a Yov7 + algorithm; 2, selecting a target tracking network; 3, combining the target tracking network with a Yolov7 + algorithm to realize target tracking; 4, setting model parameters, selecting data of a training set and a test set, and training the model; and 5, the model is detected, and identification performance is checked and recorded. According to the method, a feature extraction branch uses a Yolov7 + algorithm, feature information extracted from a single-frame image is input into a motion matching branch, and target association of a front frame and a rear frame is completed through a Strongsort algorithm; and a Yov7 + detector can be combined with a StrongSort tracker, so that target detection and tracking integration is realized, and real-time detection of the