Research On Moving Target Detection And Tracking Method Under Complex Environment

As the basis of motion analysis and high-level semantic processing,moving object detection and tracking has been widely used in many fields such as military analysis,intelligent city and human-computer interaction,and has become one of the focuses in the field of computer vision.However,there still exist a number of complex environment disturbances in the real scene analysis,such as dynamic background,high noise,illumination,shadow and occlusion,which bring great difficulties and challenges for detection and tracking.In this paper,we focus on moving target detection and tracking methods under the complex environment.(1)This paper focuses on dynamic background disturbance,high noise and shadow interference scenes under complex environment.Aiming at the problem of target incompleteness caused by misdetection of large area disturbance into foreground area by traditional Vibe algorithm in the above scenarios,this paper improves the algorithm from three aspects:background model initialization,dynamic background adaptive calculation and shadow detection elimination.Experiments show that the method can eliminate the disturbance,which improves the detection algorithm accuracy and anti-interference ability.(2)Aiming at the problem that the KCF algorithm with single feature cannot accurately and stably describe the target and has poorer anti-interference ability,under complex environment of the background of similar interference,occlusion,motion blur and so on.A multi-layer fusion adaptive kernel correlation filter tracking algorithm(MALKCF)is proposed.The HOG,CN and convolution features are integrated from the feature level to improve the target description and representation ability.The adaptive fusion tracking response graph improves the tracking accuracy.The overall algorithm model update method is designed to enhance the robustness of the algorithm.After experimental simulation,the tracking accuracy and success rate of the algorithm are 85.4%and 78.2%,which are 34.3%and 32.3%higher than the KCF algorithm respectively.It has higher robustness in complex environment of background clutter interference,motion blur and fast motion.(3)Aiming at the problem that KCF algorithm lacks occlusion processing,which results in the tracking drift of classifier learning background and occluded object information.A cascade occlusion detection mechanism is proposed to perform occlusion judgment,an adaptive model updating strategy is used to enhance the anti-interference ability of the model.Kalman filtering is introduced to optimize and predictive tracking,which improve tracking accuracy.Compared with the KCF algorithm,the proposed COPKCF algorithm improves the tracking accuracy and success rate by 31.3%and 33.6%respectively.The overall tracking performance is greatly improved,and the fps of 46 has better real-time performance.