Research On Detection And Tracking Technology Of Aircraft Targets In Complex Scenes

At present,with the rapid development of remote sensing technology,the application of remote sensing images and remote sensing video in military and civilian fields is becoming more widely.And as a special tool,aircraft plays a key role in military and civilian fields.Detection based on remote sensing image and tracking based on remote sensing video of aircraft targets play a key role in modern intelligence acquisition.However,with the improvement of the resolution of remote sensing images,the information contained in the image becomes richer and the image size increases significantly.It is difficult to deal with such changes.This paper studies detection and tracking technology of aircraft targets in complex scenarios,which refers to the scene with rich background information,large size and a wide gap with the spatial scale proportion of the target under the background of large remotes sensing image and video.Under the background of engineering practice,this paper applies target detection and tracking technology to solve the problem of aircraft moving target detection and multi-target tracking in complex scenes.The main research contents of this paper are as follows.Firstly,the Faster-RCNN target detection algorithm is used to detect aircraft moving targets in remote sensing images.The large-scale remote sensing image contains rich background information and the target is composed of a few pixels.When performing detection directly,it will lead to lower accuracy.So,the moving target detection method is used to detect the target in motion.Combined with morphological filtering,the target is finally detected.In this way,it is much easier to detect the aircraft in large scene images.The performance of the proposed method was verified by specific experiments.Secondly,the CN tracking algorithm based on kernel correlation filtering is used to study aircraft targets tracking in complex scenarios.The aircraft target tracking in remote sensing video is realized by using this algorithm based on color feature.For the problem that CN algorithm can be used in single-target tracking,a method of multi-target tracking is propoesd.Since,the studied object is a large-scene video of 4K,with complex environment,these targets that move in a slow speed are not opt to be detected.Therefore,the target detector designed in the previous part is used to detect the position of the target in the first frame of the video,and finally use CN algorithm for tracking.Experimental results show that the proposed method not only tracks moving targets in video,but also avoids the need to manually calibrate the moving targets in the first frame of video.Thereby lead to less error and superior tracking performance.