The Implementation Of Moving Target Detection And Tracking Algorithm Based On ARM Embedded Platform

Target detection and tracking has been a hot field in computer vision,by these algorithm,a vision system can capture the moving target area from an image sequence,and then calculate the location of this target.It has great application value and potential in civil and military industry.In the traditional way,we put the computation of a vision algorithm into PC,but with the development of industrial technology,this method cannot afford all computer vision applications.This paper studies how to detecting and tracking an object in dynamic background by practice,designs a target detection and tracking system based on Samsung Exynos 4412 microprocessor,and capture video data by a CCD sensor.The main contents are as follows:First,this paper introduces the technology related to object detection and tracking,and corner detection algorithm.Analyzing the main advantages and disadvantages of each object detection and tracking algorithm in dynamic background.Second,studies the moving object detection and tracking algorithms under the moving background.Based on the characteristics of target detection in dynamic background,choose a motion compensation strategy in this paper,use the matched feature point between two images,estimate the global motion parameters,finally use frame differencing method to compute the area of moving target.Then in the tracking module,use pyramidal Lucas-Kanade optical flow to real-time tracking.At last,bases on Exynos 4412 microprocessor,realizing a method of real-time moving target detection and tracking algorithm on ARM9 platform,includes porting Linux kernel,capturing video by a CCD camera,porting OpenCV vision library,and use QT library to display results into LCD module,then verifying the feasibility by porting these algorithm into embedded system.