The Research Of Moving Targets Detection And Distance Measurement Technologies In Intelligent Surveillance System

Intelligent video monitoring system is using computer vision technology toprocess, analysis and understand the observed video information, so as to realizethe monitoring of moving targets. Along with the computer and image processingtechnology increasing development, the research and design of intelligent videomonitoring system in the industrial production, transportation, security monitoringand other fields has received widespread attention and has become an importantresearch topic in the field of computer vision.The design of moving targets detection, tracking and distance measurementalgorithms makes a major impact on the effect of intelligent video surveillancesystem. Reasonable and effective algorithm can greatly improve the identificationprecision of moving targets and the performance of the intelligent videosurveillance system. Therefore, this thesis mainly studies the issue mentionedabove from the perspective of algorithm. The main work and contributions are asfollows:In terms of target detection, firstly several kinds of commonly used detectionmethods (such as: frame difference method, background difference method, opticalflow method, etc.) and the algorithms based on background modeling are comparedin view of the dynamic background motion target detection accuracy and rapiditydemands. However above methods can’t obtain excellent results in complexdynamic environment, so an improved codebook modeling of moving targetdetection approach which is on the basis of codebook model. By the algorithm,each pixel is converted into HSV color space distortion degree calculating the colorfrom RGB color space, so that the background of code word is isolated andregarded as the background model. Finally experimental results show that, underthe complex dynamic background, the method can quickly and accurately detectmoving targets, and lay a foundation for the follow-up tracking and ranging movingtargets.In the aspect of target tracking, tracking moving targets based on the detectedmoving targets. First this paper simply introduces the basic principle of movingtargets tracking algorithm and some mainstream tracking algorithms. Then thispaper focuses on the moving target tracking algorithm based on kalman filter, and successfully tracked the moving target according to the moving objects’ motionproperty. Finally through the simulation experiment, results show that movingtarget tracking algorithm based on kalman filter can well predict moving targetinformation, realize the stability of the moving target tracking.In terms of target distance measurement, first the advantages anddisadvantages of the monocular, binocular and multi vision distance measurementare simply analyzed. Then the movement target distance measurement methodbased on monocular vision is raised combined with the need of practical problems.Based on the principle of small hole imaging, the method establishes the linearmodel under which mutual transformation relationship between World CoordinateSystem, Camera Coordinate System, Retinal Coordinate System and PixelCoordinate System is calibrated. In order to accurately calculate the coordinates ofmoving targets in two-dimensional pixel coordinate system, the results ofcalibration are distortion processing. Experimental results illustrate that thedistance measurement of a moving target based on monocular vision method canmore accurately and rapidly measure the horizontal distance between the cameraand moving targets, so this method has practical significance.