Method and device for video multi-target tracking

Abstract

The invention discloses a method and device for video multi-target tracking. The method includes: when the tracking result of the last frame contains the location frame of the tracking target, for the previous frame, execute: distribute dense grids on the overall image, Each grid corresponds to a feature point; for all the position boxes of the tracking target, determine the tracking feature point corresponding to the position box of each tracking target, record the corresponding relationship and the position of each tracking feature point; according to each tracking feature in the previous frame Point position, deduce the position information of each tracking feature point in the current frame in the previous frame; according to the calculated position information of each tracking feature point in the current frame and the corresponding relationship recorded in the previous frame, generate the current frame The position boxes of each tracking target in the system are saved separately; after the tracking result is output, the next frame of the video data is determined as the current frame, and the above processing is performed cyclically to realize multi-target tracking. By applying the invention, the processing efficiency and accuracy can be greatly improved.

Description

technical field [0001] The invention relates to video monitoring technology, in particular to a method and device for video multi-target tracking. Background technique [0002] With the continuous advancement of smart city construction, a large-scale video surveillance network in the national sense has been initially formed. At the same time, one of the most direct consequences of the popularization of large-scale video surveillance networks is the generation of massive video data. Massive video data poses a huge challenge to traditional intelligent video analysis techniques. [0003] A very important task in intelligent video analysis is dynamic target capture and tracking, and many very important applications such as video retrieval, target recognition, behavior analysis, and video enrichment require the use of multi-target tracking technology. [0004] Existing multi-target tracking technologies can usually be divided into two categories: The first type of method is a d...

AI Technical Summary

Problems solved by technology

At present, the algorithm for processing multiple targets as a whole is relatively complex and the processing speed is slow

[0006] Because there are a lot of repeated calculations in the first type of method, and the second type of method has high computational complexity and slow processing speed, so these two types of methods are difficult to be practically applied in engineering

In addition, due to the complex movement of real-world targets, existing tracking methods usually lack the ability to select effective tracking features to balance tracking stability and accuracy.

Images