Research On Image Target Tracking Algorithm For Aircraft Platform

Image object tracking,as a way to detect objects,has low cost and is easy for user to interact with machine.Because of these two advantages,it has a good application prospect on small aircrafts.In addition,because of the motion of camera and the long tracking duration,object tracking on the aircraft has its own characteristics and difficulties.With the goal of applying them on aircraft,a number of classical image objects tracking algorithms and related computer vision methods are discussed in two respects:recognition of objects and tracking method,and the machine learning is also introduced.Based on the introduction of the fundamental of geodesic active contour model and Kalman filtering,a basic tracking algorithm based on Kalman filtering is presented,which can track the object's position and area in image using the geodesic active contour of object as the observation of object.The failure of basic algorithm observed in experiments,which is caused by the nonlinear variation of object's states,is solved by matching the SURF key points between two images and solving the geodesic active contour again on the basis of matched result.The occlusion problem is a widespread issue in object tracking task.In the object tracking task on aircraft,the object is more likely to be hidden by background,because of the motion of camera and the long tracking time.Hence,an accessional method using key points matching and color histogram back projection is proposed to address the occlusion issue in the object tracking.The validation of algorithm is showed by some experiments.