Research On Landmark Recognition System Of UAV Based On ANN

Mini unmanned helicopter has been the study object of many institutions and organizations all over the world for its comprehensive applications in military and civilian. Landmark recognition system is a very important part of mini unmanned helicopter, because it can detect landmark and supplies the navigation computer with location information of the landmark and assist the helicopter with motion or pose information. In this thesis, the following are studied on the topic of landmark recognition for the vision system of UAV:(1) Firstly, the research background of UAV, the being research status of the vision systems, the difficulty of the UAV landmark recognition and some typical landmark recognition method for UAV are summarized;(2) Secondly, two kinds of recognition method, which one is the basic BP neural network classifier and the other one is Hu invariable moment classifier are designed and tested based on the principles of digital image procession and neural network;(3) Thirdly, A cascade landmark classifier are proposed on the basis of the basic BP neural network and Hu invariable moments aim at the difficulty of landmark recognition, which is composed of a BP network to recognize the rotation angle and another BP network to recognize the certain landmark.All of the three kinds of landmark recognition method are tested and the results show that the cascade classifier based on the neural network can not only recognize the landmark more effectively but also provide with the location information of the landmark.