Research And FPGA Realization Of Real-time Motion Video Sequences To Achieve Target Detection Method

Moving object detection is an important and fundamental problem in thefield of computer vision. With the intellectualized and digitized development ofvideo frequency Surveillance technology,The technology of moving object hasbeen widely applied in many fields in video detection.Therefore, it become oneof valuable research topic,And Many scholars researched and many effectivemethods were put forward about moving object detection.This paper analysis and compare the commonly algorithm of moving objecton the basis of overseas research present situation. Analyzes the problemsexisting in the traditional moving target detection algorithm,and put forward anew improved algorithm based on researching Gaussian Mixture Model(GMM)moving object detection algorithm. In this paper, the main work is as follows:Firstly, A new moving object algorithm was proposed about The traditionalGMM. This paper focuses on adaptive of background update rate and accuracyof matching about gaussian distribution and the background model in GMM.Firstof all, combined Kalman Adaptive Background Model with Gaussian MixtureModel,and using Kalman Filter to predict moving object locus, and adaptiveselection update rate according to the situation of object.Secondly, using theprinciple of double match mode to match,and improving the accuracy of thematching.Second,A shadow removal method is proposed based on local color spaceconversion.In this algorithm,to determine the active rectangular box offoreground area using the detected moving object, and removing the shadowwithin rectangular box use of in HSV color space.Third, Design of image capturing and display system based on FPGA. Thesystem use Cyclone II series EP2C35F672C6FPGA device on DE2platform. Itused CMOS image sensor to collect real-time image, and the CRT monitor to display the results. This design can realized640×480@30fps video moving objectdetection.