Pedestrian Counting Method Based on Group Context

Abstract

The invention provides a pedestrian counting method based on a group context. The method comprises the following steps of: extracting a moving foreground image according to an image sequence acquired by a fixed monitoring camera; establishing a group distance matrix for groups in adjacent image frames; establishing a group relativity matrix by using the group distance matrix; detecting a group status by using the group relativity matrix, tracking groups in the image sequence and identifying a group relation in a group event to establish a group context within a time interval from the occurrence of the groups to the disappear of the groups; establishing a context mask by using an internal group context and an external group context; and establishing a maximum combination posterior probability estimation problem according to the context mask and popularizing a pedestrian counting problem based on a single frame image to a multi-frame image basis to count the group-based pedestrian number in a monitoring scene. According to the method, errors of pedestrian counting caused by blockage, image depth and attitude vibration are lowered effectively, and the counting result is consistent ona time domain.

Description

technical field [0001] The invention belongs to the technical field of intelligent video monitoring, and in particular relates to a pedestrian counting method based on group context. Background technique [0002] Pedestrian counting is a central problem in the intelligent video surveillance neighborhood. Counting the number of pedestrians in the scene can guide the flow of people in public places and warn of abnormal events in public places, so as to maintain the order and safety of public places. The participation of computers and surveillance cameras liberates human eyes from multi-scene and long-term surveillance tasks, and realizes automatic counting of pedestrians. [0003] Most of the traditional pedestrian counting methods achieve the purpose of pedestrian counting by detecting the human body in a single frame image. The introduction of human body part detectors can deal with the problem of partial occlusion; other methods of pedestrian counting are based on the regr...

AI Technical Summary

Problems solved by technology

However, these methods often only use a single frame of still images, ignoring the introduction of information (historical information) in the time domain, resulting in large errors in the results of pedestrian counting

like Figure 7 Shown is the monitoring scene example 1, Figure 8 The second example of the monitoring scene is shown. Even for the same group of pedestrians, when there is serious occlusion or the monitoring scene has a large image depth, the pedestrian counting results obtained by the traditional method are often inconsistent.

Images