Underwater target tracking method and system, storage medium, equipment, terminal and application

Abstract

The invention belongs to the technical field of underwater target tracking, discloses an underwater target tracking method and system, a storage medium, equipment, a terminal and an application. The method includes: designing a structure based on a twin network in combination with an attention mechanism and a scale pyramid, wherein the algorithm is composed of a modified deep convolutional neuralnetwork VGG and an attention module; performing related operation on the image high-dimensional semantic information features extracted by the template branch and the search branch; and obtaining thetarget position of the next frame, and adaptively selecting the optimal scale in combination with the scale. The invention provides an attention mechanism and scale self-adaption-based underwater target tracking method, which achieves accurate and robust tracking of an underwater target, provides a basic guarantee for marine ecological protection and marine fishery, and promotes marine sustainabledevelopment. The invention provides an underwater target tracking method based on an attention mechanism and scale self-adaptation, and the problems in the underwater tracking process are further relieved.

Description

technical field [0001] The invention belongs to the technical field of underwater target tracking, and in particular relates to an underwater target tracking method, system, storage medium, equipment, terminal and application. Background technique [0002] At present: Target tracking is one of the hot topics in the field of computer vision. It refers to monitoring the target in the video, recording the target's activity process, obtaining the target's trajectory, and providing support for video analysis and understanding. After decades of research, target tracking technology has made some progress, such as the correlation filtering method, which can achieve better tracking results in many scenarios. However, target tracking based on underwater vision is still a very challenging problem. The main reason is that underwater scenes have their particularities, such as poor underwater lighting, low contrast, underwater fish targets, shrimp targets, etc. The speed is fast and non-...

AI Technical Summary

Problems solved by technology

Due to the change in scale of the target during the movement, it usually brings about constant changes in image features, especially the size. This characteristic makes it difficult for traditional fixed-scale detectors to adapt to changes in target size and cannot obtain optimal extraction results.

Images