Multi-human body tracking method based on attribute relational graph appearance model

Abstract

Aiming at the problem of multi-human bodies tracking in fixed monitoring scene, the invention provides a multi-human body tracking method based on an attribute relational graph appearance model. The method comprises the following steps: firstly, setting up the attribute relational graph appearance model for a current frame human body detection region; secondly, computing the similarity between the attribute relational graph appearance model of tracking human body at the current frame and the attribute relational graph appearance model of tracking human body at the former frame; thirdly, ensuring the matching of interframe human body according to the similarity so as to ensure human body tracking condition and acquire motion track. The method can greatly improve the accuracy, the reliability and the real-time performance of the multi-human body tracking in the fixed monitoring scene.

Description

technical field [0001] The invention relates to a multi-human body tracking method in a fixed monitoring scene, in particular to a multi-human body tracking method based on an attribute relationship diagram appearance model. Background technique [0002] In public places with high security requirements such as banks, hotels, and subways, it is necessary to detect suspicious persons with abnormal behavior in time. At present, most of these places use fixed cameras to monitor the scene. Therefore, computer vision monitoring needs to accurately track the complex movements of multiple human bodies in a fixed scene. This is also the basis for high-level visual processing such as follow-up behavior understanding and abnormal track analysis. Because the tracking process is to identify different human bodies in continuous frames of surveillance video, and human appearance features are an important basis for distinguishing different human bodies in most cases, so tracking methods bas...

AI Technical Summary

Problems solved by technology

However, because the color histogram only obtains color statistics and loses spatial information, it is easy to cause errors, such as a person wearing a red jacket and green pants and a person wearing a green jacket and red pants, the color histogram established for their appearance The graphs are almost identical, and the two cannot be distinguished based on their color histogram alone

2) The method of combining color histogram and spatial information: adding spatial information when calculating the color histogram, such as proposing to divide the shape of the human body into three parts, describing its color characteristics with color histogram for each part, and using Condensation (condensation) ) algorithm to realize human body tracking, but this method lacks the description of the structural relationship between various parts of the human body, and the tracking accuracy and reliability are low; 3) Color correlation map method: the color correlation map uses a co-occurrence matrix to calculate the distance between two pixels probability of distance

Although the color correlation map can reflect the spatial correlation between different color pairs, but because of the need to use a three-dimensional array to represent different color pairs and their distances, the calculation complexity of this method is very high, and it cannot be used for real-time monitoring and tracking. Analysis of target movement, which affects tracking accuracy

[0003] Therefore, the main problems in the method of tracking the human body based on the appearance model are: how to establish a more accurate appearance model to describe the human body, and how to improve the accuracy of multi-body tracking

Images