An Indoor Environment Perception And Obstacles Warning Technology For UAV Flight

When unmanned aerial vehicles fly inside buildings or other environments which may shield the connection signals,it's hard for UAVs to acquire their pose information as a consequence of the failure to receive GPS signals.What's worse is that the complexed indoor environments threaten their flight safety.In order to solve these problems,we should make use of ground station to develop an autonomous positioning functionality for UAVs.On the other hand,we need to design a functionality with which UAVs are able to percept the complexed environments nearby in real-time and give alarms to obstacles threatening the safety of their flight as soon as possible.The main contributions and innovations are as follow:(1)Make a introduction to the techniques solving SLAM problems using vision in detail.A detailed description of key methods involved in realizing the visual SLAM has also been made.Analyze most of the current visual SLAM systems and compare the characteristics among them.By consulting a great sum of literature provides many research ideas to perceive the environments and give warnings to obstacles using UAVs with vision inside a building.(2)Present main algorithms involved in strcture from motion method and analyze visual odemetry techniques based on this.After these preparations,then analyze a classic algorithms relative to monocular visual SLAM and apply this algorithms to a quadcopter AR.Drone2.0 under ROS.By making analysis and combining the requirement UAVs need when fly inside a building,design an adaptive feature points generation technique to inprove the PTAM algorithm.The results of flight tests prove that the improvement in PTAM can increase the efficiency and flexibility when used inside a room.(3)Based on the improved PTAM algorithm,select an appropriate cluster algorithm to classify landmarks distributed on the map generated by SLAM,aiming at the requirement to avoid obstacles when UAVs fly in the complex indoor environments in real time.Design an auxiliary mechanism which will make UAVs warn operators of obstacles which threaten the safety of their flight actively.The efficiency of this improvement has been proved by flight tests.