Research On Detection Algorithm Of Ground Time Sensitive Small Target In Visible Light Image With Complex Background

In the modern military war,attack the enemy's ground fast maneuvering time sensitive target has become the primary task of attacking,such kind of target include variouskinds of vehicles,such as armored vehicles,missile launchers,tanks,command vehicles,mobile radar and so on.The ground environment of this kind of time sensitive small targets includes urban road,desert Gobi,country roadand so on.In addition,in the process of imaging,the missile-borne motion platform detects inthedirection of forward and down,and the target size is small and the detail information is insufficient in the image when the missile is far from target.Besides,the characteristics of the vehicle targets in the image are diverse,the scale of target size is large and the motion state is uncertain.Facing the above complex background and target characteristics,the traditional small target detection technology can't be applied,it is urgent to carry out the corresponding technical research.Firstly,this paper deeply analyzes the characteristics of the targets in visible image,and designs pre-treatment algorithm which can be applied to long distance vehicle small target in complex background,and combined with the experimental data to verify the validity of the algorithm.The author also migrates the image pre-treatment algorithm to the GPU platform for a large scale data processing by adopting a GPU parallel computing technology named C++ AMP to shorten the image pre-treatment time and improve efficiency.Then,on the basis of analyzing the characteristics of target and its neighborhood background,this paper adopts the box filter based on the target size and shape information to extract the suspected vehicle targets.At the same time,this paper presents a road edge detection algorithm used for assistance of detecting the target vehicles.Finally,this paper gives a false target suppression technique based on the measurement parameters of the flight platform.The experimental results show that the new algorithm and method caneffectively extract the small vehicle target from complex scene.