Research On Tracking Algorithm For Moving Object

Moving object tracking which fused the technology of image processing, pattern recognition, artificial intelligence and automatic control, is one of the core research in the field of computer vision and has have a wide variety of applications such as the military guidance, vision navigation, safety monitoring, intelligent transportation, video coding, medical diagnosis and weather analysis ,etc.In this paper , several classical tracking algorithm for moving object are summarized and profoundly analyzed, which are mainly feature-based or correlation-based or active contour-based. The accuracy of the moving targets extraction is the key of the performance for target tracking . Because classical algorithm can not obtain complete moving object in complex scenes, there are several improved algorithm method based on mixture Gaussian, the improved detection algorithm for moving target based on mixture Gaussian , was used, which is the result of our laboratory. This method can succeed in getting the complete segmentation of the moving object in complex scenes.Particle filtering based on the Bayesian estimation of Particle Filter as a nonlinear filtering algorithm brings original superiority of dealing with the nonlinear and non-Gaussian problems. In this paper, a modified mixture Gaussian model is used to establish model for monitoring scene and then the particle filter is applied to track the target in the modeling scene. Through the experiments of the outdoor and indoor scenes, it shows that the algorithm can be successful to track the moving target even under the conditions of interference of light, trees swing and nonlinear kinematic. Experiments show that the algorithm is effective.