Target tracking algorithm based on optical flow and dynamic cascade RPN

Abstract

The invention discloses a target tracking algorithm based on optical flow and dynamic cascade RPN. The target tracking algorithm sequentially comprises the steps of adaptive video sampling, construction of an optical flow feature module, fusion of multiple types of features, construction of a dynamic RPN structure and construction of a tracking framework. According to the invention, the optical flow features are used to obtain the time sequence information, and the local shielding problem and the similar object interference problem are solved; the robustness of the model is improved by fusing various features, and target tracking under a complex background is completed; the operation speed of the algorithm is optimized by increasing constraint conditions, and the real-time performance of the algorithm is further improved by using a dynamic planning method; a self-adaptive video sampling mode can solve the problem that batch training cannot be carried out in a fixed sampling interval mode, and can also solve the problem that optical flow constraints are damaged in a fixed sampling total frame number mode; the optical flow information and the appearance information are combined for modeling, and semantic features, optical flow features and low-level fusion features are fused, so that the discrimination capability of the model is improved, and the target can be successfully tracked in a complex background.

Description

technical field [0001] The invention belongs to the technical field of computer vision and relates to a target tracking algorithm based on optical flow and dynamic cascaded RPN. Background technique [0002] Target tracking is widely used in intelligent monitoring and human-computer interaction. For example, banks, supermarkets and other occasions often use target tracking analysis technology. By locating objects and analyzing their behavior, once abnormal behavior is found, the monitoring system will send out Alarm, reminding people to pay attention and deal with it in time, improving problems such as distraction of manual supervision, slow response time, waste of human resources and so on. In addition, target tracking technology also has important practical value in many fields such as virtual reality, industrial control, military equipment, medical research, video surveillance, and traffic flow observation and monitoring. Especially in the military, automatic tracking te...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to address the above problems and provide a target tracking algorithm based on optical flow and dynamic cascaded RPN. Through adaptive video sampling, it can not only solve the problem that the fixed sampling interval cannot be trained in batches, but also solve the problem of fixed sampling. The total frame number method destroys the optical flow constraint; combined with optical flow information and appearance information modeling, fused semantic features, optical flow features and low-level fusion features, aiming to improve the discriminative ability of the model so that it can successfully track in complex backgrounds Target; add a judgment branch to the cascaded RPN to form a dynamic cascaded RPN, and predict the position of the target to be tracked in simple samples in advance

Images