Dynamic Target Capture And Real-time Tracking Algorithm In Amphibious Environment

As the key basic technology of motion analysis and high-level semantic analysis,dynamic target capture and real-time tracking are widely used in many fields such as defense military,public safety,smart city,human-computer interaction,etc.It is one of the core topics in the field of computer vision.In amphibious application scenarios,there are usually water mist,high noise,occlusion,illumination and other disturbances,which bring great challenges to the capture and tracking of dynamic targets.To this end,this paper studies the dynamic target capture and real-time tracking algorithm in amphibious environment.This paper first summarizes the current research status of dynamic target capture and dynamic target tracking at home and abroad,and then makes an in-depth study on the problem of incomplete target area captured by dynamic targets in amphibious environment and the inadequate performance of tracking algorithms in the tracking process.Proposed a corresponding solution.The research methods and results of the paper are summarized as follows:1)In this paper,pre-processing algorithms for video images of people,vehicles,and ships are studied for the effects of high noise and water fog in the amphibious environment.Considering the attenuation effect of red light on the water surface during pretreatment,a dark channel prior defogging algorithm suitable for amphibious environment is adopted,and the feasibility of the algorithm is proved through experiments.Then,through experimental comparison and analysis of several commonly used image noise reduction algorithms,the wavelet transform noise reduction algorithm in the frequency domain noise reduction algorithm is finally selected.This algorithm has obvious noise reduction effect and can better retain the edge of the image,which is beneficial to the subsequent The unfolding of work.Next,three image segmentation algorithms are studied and compared experimentally.In order to accurately and efficiently locate the edges of dynamic targets,the Canny differential operator method in the segmentation method based on edge detection is selected,and morphological operations are used to make up for the shortcomings that the Canny differential operator method cannot guarantee the edge continuity and closedness.2)In order to meet the requirements of strong environment adaptability and complete target area in dynamic target capture algorithm in amphibious environment,an algorithm combining ViBe algorithm and improved three-frame difference method is designed to capture dynamic targets.First,analyze and summarize the advantages and disadvantages of the current classic target capture algorithm,and then design a dynamic target capture algorithm suitable for amphibious environment according to the experimental effect.Using the ViBe algorithm in the background difference method to extract a more complete target area;using the Canny operator for edge detection to improve the traditional three-frame difference method and enhance the traditional three-frame difference method to capture dynamic target edge information.Finally,the CDNET data set proves that the algorithm combining ViBe algorithm and the improved three-frame difference method has obvious advantages for dynamic target capture in an amphibious environment.3)Aiming at disturbances such as background clutter and illumination changes in amphibious environments,an improved background perception related filtering algorithm based on multi-feature fusion and automatic model updating is designed.Analyze the advantages and disadvantages of the background perception related filtering algorithm.The background perception algorithm only uses the gradient histogram feature,which cannot effectively deal with the problem of target rotation and scaling.The method of naming features with features and colors makes the algorithm more universal.Background perception related filtering algorithm The position filter model update method is linear interpolation,the update rate is fixed,and it is easy to fail tracking.In response to this disadvantage,this paper proposes an adaptive model update method to improve the robustness of the algorithm.According to the experimental results of OTB2013,the improved background perception related filtering algorithm can track dynamic targets more efficiently.