Monitoring scene multi-target tracking method and system

Abstract

The invention discloses a monitoring scene multi-target tracking method and system, and the method is based on the multi-hypothesis tracking after pruning optimization, and comprises the following steps: receiving a target detection frame of a current frame; based on the target detection box of the current frame, generating assumptions of possible positions of each target; solving the maximum independent subset and problem based on all hypotheses of a plurality of recent frames, obtaining hypothesis paths which can be selected by each target and do not conflict with each other, and outputtingthe position of each target before the plurality of frames; and performing pruning optimization by combining the maximum independent subset, the solution of the problem and a Hungary algorithm to obtain a multi-target tracking result. The invention provides a novel pruning optimization method, so that the complexity of a multi-hypothesis tracking algorithm is greatly reduced. For multi-target tracking of a monitoring scene, CPU resources can be fully utilized, GPU resources are not occupied, and multiple targets are tracked in real time; and the adjustment of algorithm complexity and precisionbalance can be realized by changing the length of information integration time.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a real-time tracking method and system for multiple targets in a monitoring scene. Background technique [0002] In the process of intelligent analysis of surveillance video, the usual process is to extract the detection frame of the target, and give each target its own unique identification code through the multi-target tracking method to distinguish different targets. Multi-target tracking method is a very classic problem in the field of computer research. Usually the algorithm converts the tracking problem into a graphical model for solution. However, it is usually difficult to flexibly adjust the balance between the accuracy and speed of the algorithm. Most of the improvements in multi-object tracking methods in recent years are based on deep learning-based feature extraction methods. By introducing a deep feature extraction network into the multi-target tracking method,...

AI Technical Summary

Problems solved by technology

However, in the original multi-hypothesis tracking algorithm, although the maximum independent subset sum algorithm (MWIS) will prune the hypothesis tree, the number of hypotheses for each target will still multiply geometrically over time before reaching the pruning time point. growth, it is difficult to make judgments based on long-term hypothetical information

Images