Detection method based on head identification and tracking characteristics

Abstract

The invention provides a detection method based on head recognition and tracking characteristics in the technical field of pattern recognition. The detection method comprises the following steps: acquiring images from the video frequency, detecting acquired images by adopting a head categorizer to the head area, clustering the detected head area in the frame with consecutive numbers according to the western-style distance so as to gain corresponding relationship between head areas in the adjacent frames; carrying out one-one corresponding relationship between head areas in the adjacent frames intersected and recognized by adopting a bar chart, so as to obtain the track of head movements; and finally, judging in and out behaviors according to import and export marks and the track of head movements so as to detect number of the people. The invention utilizes abundant samples to train the head categorizer, the detection method is more accurate than the method of simply adopting ellipse matching to recognize the head, and the detection precision of the invention is about 90%.

Description

technical field [0001] The invention relates to a detection method in the technical field of image processing, in particular to a detection method based on head recognition and tracking features. Background technique [0002] Computer vision inspection technology uses engineering technology methods to start from the analysis of the functions of the human visual process, and does not deliberately simulate the internal structure of the human visual system, but only considers the input and output of the system, and uses existing feasible means to realize system functions . Video-based people detection is a branch of computer vision applications, and it replaces manual detection of people entering and leaving through computer understanding of video images. [0003] After searching the existing technical literature, it was found that the patent application number is 200510025186.7, and the name is: Entrance and Exit People Counter, which involves a method for automatically count...

AI Technical Summary

Problems solved by technology

However, this method is based on the background subtraction method to detect the moving area, which is seriously affected by the light, especially the door opening and closing phenomenon often occurs at the entrance and exit, causing sudden changes in light, which reduces the detection accuracy of the background subtraction method. The statistical accuracy will be reduced because the human body model is more complex and easy to occlude each other

[0004] After searching, it is also found that the special application number is 200810116312.3, and the name is: intelligent management method and system based on video people counting. Although this technology has added ellipse-like detection to judge the head area, its core method is still based on background construction. In terms of the modeling method, the influence on the illumination is not very robust. At the same time, the ellipse model is too simple to simulate the head, and it has not passed a large number of head feature detection.

Images