Remote sensing vessel target tracking method based on background self-selection

Abstract

The invention discloses a remote sensing vessel multi-target tracking method based on background self-selection, and the method comprises the steps: carrying out the high-precision target detection and background classification of a specific frame through a front module, and enabling the module to divide a sequence frame into a pure ocean background and a background containing sea and land; for a frame image with a pure ocean background, performing multi-target tracking by adopting a high-speed correlation filtering algorithm; and for a frame image containing sea and land backgrounds, adopting a high-precision neural network to carry out multi-target tracking. In addition, each module is provided with a tracking loss return mechanism, and tracking loss and scene switching phenomena of the tracking module are returned to the front module for re-detection and classification. According to the method, the feature information of the vessel remote sensing video is fully explored, the tracking speed is greatly improved while the detection precision is ensured, the network structure is complete, the process is clear, and the method has high field application value.

Description

technical field [0001] The invention relates to a remote sensing ship multi-target tracking method based on background self-selection, belonging to the technical field of multi-target tracking. Background technique [0002] Recently, with the frequent military activities at sea and the development of marine transportation and trade, ships at sea are the key targets for security monitoring and wartime strikes. Using technical means to quickly and accurately identify and track ship targets has great practical significance and application value. Remote sensing videos have the following characteristics: some frames in long videos are often of very poor quality, which manifests as motion blur and cloud occlusion in ship videos. The main consideration of video target detection and tracking is how to integrate more spatio-temporal features, so as to make up for the lack of features obtained in a single frame during training or detection. In addition, in terms of shooting scenes, ...

AI Technical Summary

Problems solved by technology

However, there are still two problems in this field: first, the current research is mostly based on target detection of ship remote sensing images, and there is a lack of in-depth research on ship remote sensing video target tracking models; second, ship remote sensing video has strong domain characteristics, traditional methods are difficult to generalize

Images