The Research On Detection And Tracking Algorithms Of Motion Object

The detection and tracking of moving target is an important research in the field of visual machinery. With the rapid development of the technology of detection and tracking of moving target, people want to operate efficient detection and tracking of algorithm. How to detect and track moving objects accurately and efficiently is a key and difficult research issue in the field of visual machinery. The research of this paper includes the following aspects:(1) Pointing at the problem that the traditional W4 background model algorithm fails to eliminate the shadow of moving objects, a moving target detection algorithm which can remove the shadow is proposed. Primarily, in the way of the W4 algorithm of background model we can get each pixel in video image sequences of maximum and minimum gray value and maximum between adjacent frames difference value; secondly, conduct the image of each pixel of the maximum gray value and minimum gray value weighted linear processing; thirdly, remove the affect shadows simplified LBP texture feature. Finally, apply similar extract multiple moving target detection background model in the principle of Gaussian mixture background model. Obtaining streamline type LBP texture feature extraction and reducing complexity meets the real-time requirement. The analog result show that the algorithm compared to homogeneous methods can effectively dislodge the distraction of shadow detection of moving target, but also meets the real-time requirements.(2) For moving target tracking and target background often appearing in similar, object occlusion, background interference and other issues, this article conducts multi feature fusion and two quadratic curve fitting algorithm based on Camshift. In order to solve the problem of background interference, this article applies the W4 model improved background modeling. The model of moving target is described by motion information, color feature, LBP texture features. In the tracking process, in order to adapt to the change of background and target, this article combines three frame difference methods to eliminate the interference of background noise to obtain the exact location of target. The target is occluded when using the two curve fitting method for trajectory prediction to find lost back moving target. Compared with the existing algorithms of tracking, the method in complex background and occlusion of target location is more accurate and has better robustness.(3) Propose a combination of sparse representation, spatial Pyramid model and the mean shift tracking. First, extend the classical mean shift algorithm to make it estimate the change of the spatial position and estimate the change of direction and scale space; secondly, introduce pixel density block sampling technique and trivial template design model, the histogram matching more accurate, which effectively overcomes the light changes. Finally, replace the original algorithm with either the global model or local models to represent the feature of the target, and spatial pyramid model to combine the two representation methods to effectively solve the occlusion problem. Analog results demonstrates that the proposed results run tracking algorithm is superior to homogeneous algorithms, and can precisely solve the problem of the scene illumination changes, part missing occasion and the scale change of target.