Research Of Motion Detect Based On Pulse Coupled Neural Network

With the development of using information technology and the uplifting of people’sliving standard in recent years, people are pushing for intelligent monitoring system moreand more urgently. The aim of intelligent monitoring system is to automatically do theobject detection, identification, tracking and behavior understanding. The intelligentsystem has a large potential value to reinforce the social public order. Moving objectdetection as the first step of the video processing which is a part of intelligent videosurveillance plays a really important role and is a technical focus as well as difficulty at thesame time.For monitoring system based on a camera, the coterminous frames differencing, opticalflow, and background differencing are all the staple algorithm for moving object detection.This article has presented a method of the PCNN moving object detection on the basis ofresearching the classical algorithm. It also has improved the basic algorithm. It hasresearched the following parts:1Considering the major noisy points on the original control images and the inaptitudeto extract object, this article has presented the improved image de-noising method withmedian filter technology which can eliminate impulse noises effectively and keep thetexture information of images.2Summed up algorithm’s principles of coterminous frames differencing, optical flow,Gaussian mixing background modeling, and then analyzed the problems and applicableconditions of classical algorithms by experiments and comparison.3Descanted on mathematical models and working principles of the PCNN. As theGaussian mixing background modeling has the problems including low-speed modelingand without guarantee of pixel spatial correlation, etc. This paper has presented thealgorithm of moving object detection based on PCNN. It has designed the algorithm flowof moving object which aims to simplify the PCNN model. Comparing with traditionalalgorithm, this one has the advantages like fast background modeling and highanti-interference, etc.4This article has presented the improved algorithm of the PCNN parameter by usingthe OFGA to solve the problem that the PCNN parameters are complicated and hard to bechosen. It has improved genetic operator with PCNN parameters’ feature and the frequencywith the optimized parameters. Comparing with the optimized algorithm experiments ofGA parameters, it has proved that OFGA has a fast convergence and is hard to ripe early.5Under the framework of OFGA, this article has designed the algorithm flow ofPCNN moving object detection based on optimized OFGA parameters. In order to verify the algorithm validity, this article has realized the algorithm in this text and the concerningcontrastive algorithms by using the VS2008, CUDA, OpenCV2.2on PC. The experimentalresults proved that the improved PCNN algorithm could process the backgroundmultimode areas better and increase the segmenting precision of the moving object. Itcould eliminate the interference and hold the shape as well as the marginal information ofthe moving object at the same time.