Ground Vehicle Target Recognition Based On Video Stream Image Feature Extraction

How to classify and identify various wheeled vehicles and tracked vehicles is of great value to ground reconnaissance,battlefield situation awareness,threat assessment,command decision-making and precision attack in modern war.Video stream feature is an important feature of ground vehicle target.Choosing an appropriate video stream feature extraction method is the key to ground vehicle target recognition.Based on the analysis of various video stream feature extraction algorithms and the characteristics of ground vehicle targets,this paper proposes a dynamic modal decomposition（DL-DMD）feature extraction algorithm based on dictionary learning.Firstly,various video stream feature extraction algorithms are analyzed,and the dynamic modal decomposition（DL-DMD）feature extraction algorithm based on dictionary learning is introduced.The principle and characteristics of the algorithm are introduced,and its application in video surveillance is analyzed;Then,based on the algorithm,a ground vehicle video foreground segmentation platform is proposed.The platform is designed from five aspects: the construction of experimental environment,the overall design of video segmentation and recognition system,the design and testing of user interface and the analysis of results.Through the platform,the video stream features of ground target vehicles are extracted and recognized;Finally,the development trend of ground vehicle target recognition technology is prospected,which is expected to provide reference for the classification and recognition of ground vehicle targets.