Aircraft Detection Algorithm Research Based On Structured Features In Images

With the rapid development of computer science, artificial intelligence and computer vision are paying more attention on the analysis of targets action and status. Aircraft target detection is the base of aircraft behavior and state analysis, and is very crucial to automatic airport management, aircraft flight condition monitoring, aircraft security and so on. At the same time, the complex background, the huge difference between aircraft and backgrounds, the variation of aircraft poses are difficult problems for aircraft detection to solve, and they are also problems for ting target detection method.Aiming at the difficulties of aircraft detection, this paper researches the structural features of aircraft, and proposes the aircraft detection algorithm based on corners and HOG feature. The main work is as follows:(1) The various types, different appearances and posture changes of aircraft are unsolved problems in aircraft target detection. Thus dynamics, design principles of aircraft and its manifestations in plane image are analyzed in this paper; Relying on corners and slender body edge symmetrical features of aircraft, this paper applies the corners, edge of aircraft and their structural relation to research the detection of aircraft targets under different backgrounds.(2) The occlusion and incomplete contour extraction of targets because of complex backgrounds and low contrast of grey-scale are also difficult to solve in aircraft targets detection. According to the structural relationship between corners of aircraft, based on the theory of convex hull in the computational geometry, the aircraft ROI(Regions of Interest) extraction algorithm is designed. Experimental result shows that the algorithm can effectively extract the regions of interests for aircraft target under different backgrounds.(3) The aircraft structural extraction feature of aircraft edges is hard to accomplish in complex background. This paper applies HOG features to extract the aircraft edge; Adaboost learning algorithm based on cascade are applied to avoid the high dimension of HOG features, and an aircraft classifier is got as a result. The result shows that the classifier has good performance to the aircraft detection in different backgrounds.The result shows that proposed algorithm in this paper can get good performance in aircraft detection and overcomes the influence of background occlusion, various appearance and different posture. This algorithm gets higher detection rate and lower false detection rate compared to traditional aircraft targets detection methods, and lays a solid foundation for the intelligent analysis of aircraft state and behavior.