Study Of Positioning Moving Human Based On Shape Complexity

With the rapid development of the technology of machine vision and digital image processing,the intelligent video surveillance technology has got more and more attention and played animportant role in traffic, community, military target, public and other places. Therefore, howto detect and position moving target in real-time monitoring system, then to track andrecognize, and to analyze the behavior of the target, has become the focus in the study. Thispaper mainly studied the technology of detecting and positioning moving human, it is the keytechnology of intelligent video surveillance.In this paper, we proposed the method of positioning moving human based on shapecomplexity. In order to improve the speed and accuracy of processing and analysis, first of allvideo image sequence collected by monitoring equipment should be preprocessed, thepreprocessing methods include gray, denoising, histogram equalization and binarization, etc.Gray can compress the image storage space, and improve the operation speed; Denoising canbe used to reduce the influence of monitoring equipment and natural conditions, make thedetection of moving regions more accurate; Histogram equalization increases the contrastbetween and background; Binarization stick out the foreground, ignore the background, whichis benefit for processing and analysis in the follow-up work, greatly reduces thecomputational complexity at the same time. In order to detect the moving regions frombackground image, this paper adopts the background modeling method based on framedifference method. Comparing three background images in different light intensity, early inthe morning, at noon and in the evening, we set the scheme to update background by variablespeed coefficient and then put the current video frame substract the background model to getthe moving regions. In view of possible shadow along with the moving regions, this paperadopts the model of c1c2c3to detect and remove the shadow; Contrapose the small area ofnoise in the detection results, we set a threshold by the experiment, when the area is less thanthe threshold, we think that it is sudden noise and remove it; Aiming at the existence of themoving regions are not connected, incomplete edge and burr, this paper uses the morphology method to eliminate the tiny burrs, filled the holes and repair damaged edges by the method oflocal minimum, obtaining the complete contour in the end. To recognize the human body fromthe detected regions, this paper proposes a weighted shape complexity analysis method,greatly improving the discrimination between the human body and other moving objects. Theexperiment results show that the method of positioning moving human based on shapecomplexity can reduce the influence of natural conditions and position the moving humanrapidly and accurately.