A multi-function system based on inertial measurement unit (IMU) is introduced, which can fulfill navigation, attitude measurement of LOS in payload, platform stabilization and tracking control. The IMU is integrated with electro-optical sensors and a laser range finder on gimbals, which performs attitude calculation and navigation by constructing navigation coordinates in a mathematic platform, and the platform navigation information is obtained by transformation matrix between platform and gimbal coordinates. The platform comprising of gyros, electro-optical sensors and servo mechanism is capable of stabilizing line of sight and could be used to geo-tracking in the relevant field of view (FOV).The system can determine geography coordinates of the host platform and target only with navigation information and laser ranging data. The geo-tracking system always locked the target image at the center of FOV by calculating spatial geometry and adjusting LOS attitude. This tracking is different from TV tracking and geographical reference image tracking, which may be influenced by fog and obscurant. When the UAV is flying over urban or mountain areas for rescue missions, it can avoid the loss of targets due to strong maneuver or LOS obscuration, and reduce the operation load and improve rescue efficiency.
Global motion estimation within frames is very important in the UAV(unmanned aerial vehicle) image stabilization system. A fast algorithm based on phase correlation and image down-sampling in sub-pixel was proposed. First, down-sampling of the two frames to quantitatively reduce calculate data. Then, take the method based of phase correlation to realize the global motion estimation in integer-pixel. When it calculated out, chooses the overlapped area of the two frames and interpolated them with zero, then adopts the method based on phase correlation to achieve the global motion estimation in sub-pixel. At last, weighted calculate the result in integer-pixel and the result in sub-pixel, the global motion displacement in sub-pixel of the two images will be calculated out. Experimental results show that, using the proposed algorithm can not only achieve good robustness to the influence of noise, illumination and partially sheltered but also improve the accuracy of motion estimation and efficiency of computing significantly．