Method and device for visual tracking based on correlation filtering

Abstract

The embodiment of the present invention discloses a method and device for visual tracking based on correlation filtering, which can improve the robustness and accuracy of tracking. The method includes: S1, extracting a sample whose size is the target scale at the target position in the first frame of the video, extracting multi-channel features from the sample, using the multi-channel feature to update position filter parameters, extracting multi-scale features at the target position, using multi-scale Feature update scale filter parameters; S2. Extract the corresponding target position and target scale in other frames of the video except the first frame. Multi-channel features that are the same as the first frame, and determine a new one according to the multi-channel features and position filter parameters. Extract the multi-scale feature at the new target position in other frames, determine the new target scale according to the multi-scale feature and scale filter parameters, and use the new target position and new target scale to update the target position and target scale ; S3, judging whether the tracking is over, if not, execute step S1.

Description

technical field [0001] Embodiments of the present invention relate to the field of visual tracking of video moving objects, and in particular to a method and device for visual tracking based on correlation filtering. Background technique [0002] Moving target visual tracking technology is of great significance in the field of UAV vision. Visual tracking technology refers to estimating the state of the target in subsequent frames of the video based on the position and size of the target in the first frame of the given video. Visual tracking is a challenging problem in the field of computer vision. The challenge is that the prior knowledge is only the state of the object in the first frame of the video, and there is no explicit modeling of the object. Motion blur, occlusion, shape and scale changes during target movement will cause loss of tracking targets. In addition, the real-time requirement of UAV visual tracking also poses a challenge to the complexity of the tracking...

AI Technical Summary

Problems solved by technology

However, this type of algorithm has the following problems: (1) It cannot adapt to the scale change of the target, and performing KCF calculations on multiple scales will inevitably affect the tracking speed, and it cannot accurately estimate the scale change of the target

(2) The color information and grayscale information of the target are not fully utilized, and the cumulative error of the model is likely to cause the loss of the tracking target after being occluded.

[0004] In short, the current correlation filter tracking algorithm has made great progress, but there are also great limitations in scale, aspect ratio adaptation, and feature selection. Further research work is needed to improve the accuracy of algorithm position and scale estimation. and robustness against occlusions, deformations, etc.

Images