Research On Moving Target Tracking Method Based On Video Complex Conditions

With the development of the Internet of Things and video surveillance technology and the wide application of machine vision,high-intelligence video surveillance and corresponding smart cities and safe cities have made great progress at the same time.With the extensive application of these fields,massive amounts of video surveillance data have been generated at the same time.In order to obtain interesting motion information in these video data,and at the same time replace the traditional information screening methods,how to obtain the moving targets in these videos is the current research.Hot spots and difficulties.Under normal circumstances,the background of the target is often not single,the complexity of the environment,the morphological changes of the tracked target,occlusion issues,etc.,resulting in the research of mobile target tracking algorithm in many aspects there are problems that must be resolved,so this article The dissertation will focus on the corresponding research on various methods of mobile target tracking in complex video-based environments.The specific work is as follows:1.Proposed a moving target detection algorithm adapted to the rainy environment.Aiming at the complicated environment that exists in the video due to the weather,rainy days and raindrops,the rain line detects the moving object to form noise,and based on this,a mobile target detection algorithm that adapts to the rainy environment is combined with the raindrop detection algorithm and the Gaussian mixture model.2.Proposed an improved Mean Shift moving target tracking algorithm under occlusion conditions.An optimization algorithm is proposed for the shortcomings of the classical Mean Shift algorithm.The color histogram of the target window after dividing the original target area is used as the feature describing the target,the adaptive window size,and the goal of the Kalman filter fusion.The state prediction mechanism,to some extent,improves the weakness of the original algorithm when the target is occluded.3.Proposed an improved particle filter moving target tracking algorithm under complex conditions.The algorithm of the classical particle filtering algorithm is improved in response to the defects such as target occlusion,light variation,and targetmorphological change.The particle number selection and redistribution strategy and template update method in the particle filter algorithm are proposed,and the occlusion strategy is proposed.The effectiveness of the algorithm under the conditions of occlusion and noise interference is analyzed through simulation.4.Proposed a particle filter algorithm based on mean clustering.An improved particle filter tracking algorithm using the mean clustering algorithm to enable particles to move to local extrema points.Absorption of the mean shift algorithm to improve the particle filter algorithm to solve the large amount of defects,simulation,compared with the previous single mean shift algorithm and particle filter algorithm,reflecting the performance of the improved algorithm.