The Study Of Vision Based Landing Guidance System For Unmanned Helicopter

Miniature Unmanned Helicopter (MUH), which is widely used in both military and civil applications, has been one of the most hot research topics nationally and internationally in recent years. The autonomous landing of MUH is one of most significant procedure during flight, which requires high precision of the navigation, control and reliability. For traditional inertial navigation method is susceptible to the environment, it is necessary to introduce vision-based navigation method to autonomous landing of MUH. At present, vision-based autonomous landing has been widely studied by foreign universities. However, most domestic research in this field has only realized simulation or semi-physical simulation.In this dissertation, a Hirobo level-90model helicopter is utilized as the experimental platform for developing a novel vision-guided autonomous landing system for MUH. The mainframe of this dissertation is listed as below:(1) Considering the characteristics of the MUH platform, a method of vision-guiding autonomous landing system is proposed as below:attach four colored foam balls to the helicopter as targets; utilize a ground binocular vision sensor system to track the targets; calculate the pose and position of helicopter in the powerful computing ground station; finally, transmit the pose and position information using the wireless communication module to the flight control computer.(2) A hybrid particle filter and mean shift tracking algorithm is proposed after analyzing the characteristics and principle of mean shift tracking algorithm and particle filter tracking algorithm. Experiments are carried out using those three algorithms to track the targets. Experimental results show that the hybrid particle filter and mean shift tracking algorithm excels the other two in tracking accuracy and efficiency.(3) A monocular pose calculating algorithm and binocular pose calculating algorithm are used separately to estimate of the world coordinate and yaw angle of MUH, then the estimate result is compared to world coordinate from GPS and yaw angle from electrical compass and gyroscope. Experimental results show that: binocular pose calculating algorithm can effectively reduce the chances of outlier value, improve the reliability of the position and pose information; the estimate results from vision-guiding autonomous landing system complies with inertial data, which proves the method proposed in the dissertation is feasible.