Complex Context Of Target Detection And Tracking Technology Research

Dynamic object tracking has received considerable attention in image processing areadue to its wide application prospects and implicaitons in visual effects. Meanwhile thediversity and complexity of visual enviornment make it difficult to effectively and stableycomplete the tracking task. this paper proposes that MeanShift algorithm, combined withthe Kalman filter, is able to achieve real-time tracking steadily.There exits numerous tracking algorithm in the visual processing area from simple,low-speed backgroud to complex high-speed background. The algorithm of MeanShift isbased upon the estimation of kernel density and the distribution of Histogram. Besides it isunsensitive to the change of a target's shape and size and has such advantages as trackingdynamic object fast, stably and effectively. This paper is based on MeanShift to achieve thetarget tracking mission under complex background. Using Kalman Filter to forecast thestate of the target and update the template in time, then the solidity of the improvedMeanShift is strengthened.As MeanShift algorithm needs to Conduct exhaustive image matching, and is hard totrack rapid moving targets.Kalman Filter can effectively forecast the state of moving target,it can reduce the number of matching operations. The forecasting results of Kalman Filtercan be the starting point of MeanShift, then we can get the precise target location afterseveral matching operation, and the task of fast moving target tracking can be improved.Even if the target is completely occluded, the forecasting results can be used as a targetlocation awaiting goal again. Under certain matching criteria, the establishment of adaptiveKalman Filter group can update template of a target timely, avoiding trackingenvironmental changes which leads to the failure of tracking task.LabVIEW pogram is used to actualize the algorithm.The tracking algorithm in this paper is able to complete the target tracking tasks under complex background duringnumbers of experimental tests. We got excellent tracking effect, the images processed andalgorithm analysis is presented.