|
This thesis is focused on the technology of vedio vehicles detection and tracking based on optical flow field.Currently,with the continonous developmet of economy and society,in order to meet all areas of the need for intelligent video surveillance system, intelligent surveillance is still a hot reaserch for our country,and video image on motion detection and tracking is one of the most central issues,it not only can be used in military and industry,also could be applied to public places,docks,banks and vehicles detetion for transportation systems,etc. This paper mainly including the following points:1.Duing to exterior and interior influences,the images usually are polluted by noise,which would greatly reduce the accuracy of moving targets detection based on optical flow algorithm,so the images should be smoothed before the process of moving targets detection,this paper introduce three classical denoising method: median-filter, mean filter and dynamic filter.2.The key point for moving target detection is how to extract target object, namely,getting the target objects from the complicated background.The traditional method of target detection should extract the static background image first,then compare the image sequence with background imgae to extract target objects,but this method with very large amount of data processing.This paper mainly introduce four optical flow algorithms:Horn-Schunk, Lucas-Kanade,block matching and phase-based algorithm.Judging by experiments, H-S and L-K algorithm can detect moving targets effectively,however,there are still certain advantages and dis- advantages exist both of them.3.Tracking the moving target with Horn-Schunk algorithm effectively through simulation experiment,it proofs this method significantly reduces the amount of data processing,could be better meet real-time traget objects tracking in reality.
|