Deep learning integrated tracking algorithm based on target movement trajectory prediction

Abstract

The invention discloses a deep learning integrated tracking algorithm based on target movement trajectory prediction. The algorithm comprises the steps: firstly reading a video frame sequence, and obtaining an initial frame image; then initializing a target tracking position, and calculating semantic information pre_label of a tracking target by using a detection algorithm; calculating a color feature average template S; judging whether a detection algorithm can be started or not; predicting the moving direction of the target object according to the t-5 frame and the t-15 frame, and screening k potential target objects of which the semantics is pre_label; and finally, respectively extracting three color features of the K target objects and performing similarity calculation on the three color features and the template S, selecting a maximum similarity value SDr from K similarity results, and if SDr is greater than tp, performing deviation correction on a tracking result by using a target object tracking frame until tracking is finished. According to the method, the problems of tracking drift and the like caused by overlarge scale change, shielding and the like in tracking are effectively solved, and the robustness and the accuracy of target tracking are improved.

Description

technical field [0001] The invention belongs to the technical field of target tracking, and in particular relates to a tracking algorithm based on deep learning integration of target moving trajectory prediction. Background technique [0002] Object tracking is an important research direction in the field of computer vision, which has attracted the attention of scholars at home and abroad. It has great research and application value in intelligent video surveillance, intelligent transportation, public safety, and military and civilian fields. Object tracking is the continuous positioning of an object in a series of changing video sequences. Researching an algorithm with high accuracy and good robustness has always been a key issue in the field of computer vision. There are many challenges in target tracking, such as occlusion, scale transformation, illumination change, out of view range, fast movement, rotation and background disturbance, among which scale transformation is...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problems of tracking drift and loss caused by excessive scale changes and occlusion in the existing algorithm in the tracking process, the present invention provides a tracking algorithm based on deep learning integration of target movement trajectory prediction, using detection algorithms and target movement Trajectory prediction, combined with historical frame tracking results and color features to calculate the target used to correct the tracking results

Images