Research On Moving Human Target Detection And Tracking In Particular Scene

The moving target detection and tracking which have been widely used in thepublic security, intelligent transportation and other areas are two core problems inthe field of computer vision. But in the realistic environment, the existing of manyinterference factors, such as illumination, occluded and similar targets, will lead tothe inaccuracy of the moving target detection and tracking. Aimed at this problem,this paper researched on the moving target detection and tracking respectively, andthe main work is as follows:Firstly, a novel algorithm based on Gaussian mixture model and three-framedifference is proposed to improve the deficiency of the lower processing capacity inan illumination mutation environment in the traditional object detection. Thismethod based on Gaussian mixture background model, determining whether there isillumination change through statistical methods, using three-frame difference todetect targets in the illumination mutation environment, and dividing the detectedscene into two areas, such as a static area and a dynamic area. It detects everydiscontinuous frame for static area, reduces computational burden.Secondly, a multi-feature fusion and particle based tracking algorithm is putforward for solving the problem of losing target which is caused by using singlecolor characteristic to tracking targets in object tracking. Extracting colorinformation and statistical characteristic information of objects, describing movingtargets through adaptive fusion of characteristics, and adjusting feature weightsadaptively according to the changes of environment, realizing robust tracking ofmoving targets in illumination mutation environment.Finally, in order to validate the effectiveness of the proposed algorithm, theobject tracking experiments in different scenarios are designed and the data isanalyzed. The results show that the proposed method can achieve a good trackingeffect.