Moving Object Detection And Tracking Based Airborn Downward-looking Platform

Moving objecties detection and tracking in airborn downward-looking platform isone of the key technology in aerial reconnaissance systems, which has beed used innon-war military fields such as aerial reconnaissance, frontier defence, suppresssmuggling, frontier patrol, paroxysmal events management and antiterrorist. Airbornobject detection and tracking systems will be influenced by high dynamic platform andground scenes in condition of downward-looking, which will depress real-time abilityand reliability of airborne systems.The paper focuses on motion estimation and background compensatation, movingobject detection and tracking as follows:1) An improved coherent point drift feature matching is brought forward toestimate motion of airborne platform. Features points are extracted using Harrisalgorithm. Then the parameters of background global motion model are solved throughimproved CPD feature matching algorithm and robust RANSAC+LS parameterestimation algorithm. At last, actual data are used to validate and analysis the algorithm.2) A moving object detection method is brought forward on the ground ofbackground estimation. The improved Gaussian mixture model is used to updatebackground adaptly. Then segment background and former scenes in order to detect andorient objects.3) An effective tracking method based combined Kalman filter object trajectoryestimation and meanshift algorithm is brought forward to tracking moving object fromdownward-looking airborn platform after analysising and investigating Kalman filterestimation and meanshift algorithm. In comparison with traditional meanshift algorithm,the algorithm presented by the paper can solve tracking lag and occlusion objectmissing problem. At the same time, the improved algorithm can track moving objectstably in condition of airborn downward-looking platform.