A non-feature matching motion trajectory optimization method and system

Abstract

The invention discloses a non-feature matching motion trajectory optimization method and system, The method comprises the steps of: step S1, Depending on the angle of view and the working distance ofthe camera, calculating a lateral field of view width, calculating a trajectory window N according to the ratio of the human moving distance to the screen, the matching frame rate of the classic walking speed and the moving trajectory, calculating the moving speed and the moving direction of each point of the moving trajectory sequence with N frames as the calculating width, and further calculating the changing rate of the moving direction of the sequence; Step S2, determining abnormal inflection points in the trajectory according to the moving speed and the changing rate of the moving direction of the moving trajectory sequence, cutting the moving trajectory sequence into a plurality of pieces of moving trajectory sequences according to the abnormal inflection points, and re-combining themoving trajectory sequences into a new set of moving trajectory sequences; Step S3, judging the disconnection, coincidence and separation of the motion trajectory according to the new motion trajectory sequence set to obtain the repaired motion trajectory sequence set.

Description

technical field [0001] The invention relates to the technical field of people counting in video image processing, in particular to a motion trajectory optimization method and system for non-feature matching of people counting. Background technique [0002] At present, in video image processing, target tracking methods mainly include: matching methods based on target static features and matching methods based on target distance and geometric features. Since the method based on target static feature matching needs to calculate the feature information of all targets in the entire video sequence, and needs to perform cyclic matching with all feature data, it has strong matching and does not cause tracking failure due to target loss at local time. In the case of multi-target tracking, the characteristics of each target can be effectively distinguished. At the same time, such a global strong matching method also leads to a huge amount of calculation; while the matching method base...

AI Technical Summary

Problems solved by technology

Since the method based on target static feature matching needs to calculate the feature information of all targets in the entire video sequence, and needs to perform cyclic matching with all feature data, it has strong matching and does not cause tracking failure due to target loss at local time. In the case of multi-target tracking, the characteristics of each target can be effectively distinguished. At the same time, such a global strong matching method also leads to a huge amount of calculation; while the matching method based on distance and geometric features has a fast matching speed, but in multi-target tracking In the case of , due to the lack of feature information of different targets, the matching accuracy is low

[0003] At present, in the field of people flow statistics, the front-end computing terminal is generally used, resulting in weak computing power of the computing terminal, so the matching method of distance and geometric features can only be used for tracking and counting, and the accuracy rate is not high.

Images