Real-time video target tracking method and system

Abstract

The invention discloses a real-time video target tracking method and system. The method comprises that obtained video training set and video test set are processed; a video target tracking network model composed of a twinning convolutional network, a single-layer convolutional network, a bidirectional long/short-time memory network and a full connection layer network is established, a network parameter of the model is initialized, and a loss function of the model is determined; picture pairs in the processed video training set are output to the model successively, and a predicted position coordinate of a video target is determined; a trained network parameter is determined according to the predicted position coordinate, an obtained practical position coordinate of the video target and theloss function; and the video target in the processed video test set is tracked in real time according to the trained network parameter and model. The video target tracking network model is used to track the video target effectively in real time, the video target is tracked in higher precision, and the application range and scene are broadened.

Description

technical field [0001] The invention relates to the technical field of video target tracking, in particular to a method and system for real-time tracking of video targets. Background technique [0002] With the development of deep learning technology, video target tracking technology based on deep learning is more and more widely used. Video object tracking is an object tracking technology based on frame tracking of objects in video. The video object tracking method integrates various professional technologies such as deep learning, pattern recognition and digital image processing. There are two key points in video target tracking: (1) extraction of target features; (2) tracking algorithm. Deep learning can realize the implementation of image feature extraction and tracking algorithms, which greatly improves the accuracy of video target tracking. The design of the deep learning network structure used in the video object tracking method often directly affects the effect of...

AI Technical Summary

Problems solved by technology

Long-term practice has found that although the above methods have achieved good tracking results, they cannot effectively track the movement of video targets in real time, and the accuracy of video target tracking is not high.

Images