Environment Perception Method Of UAV In Low-altitude Urban Area Based On Stereo Vision

Unmanned aerial vehicle(UAV)is widely applied for military and civil use because of its advantages such as the low cost,good mobility,high cost-effectiveness,strong viability,and zero risk of casualties.However,the activity environment has a great impact on the working status of a UAV.Therefore,current UAVs work mainly in high-altitude or open areas,and the low-altitude urban environment is still an obstacle to the autonomous flying of UAVs.For instance,a UAV’s safety tends to be threatened by the most common buildings in a city,since the UAV itself does not have the capability of intelligent perception.Thus,the ability to sense the surrounding environment has become a prerequisite for the autonomous obstacle avoidance of UAVs.So far,relevant researches on the environment perception of UAV have been taken both domestically and internationally.In order to realize "the Environment Perception in low-altitude urban area by UAV",the main research contents in this dissertation are divided into two parts.In the first part,a three-dimensional environment is constructed on the basis of the flying platform.Then the three-dimensional information on surroundings is obtained by taking advantage of the principles of binocular stereo vision,and the detection for obstacles is achieved via related data processing.In the second part,an orthogonal invariance method based on template matching is presented with the purpose of achieving the object identification of typical urban obstacles—buildings.Firstly,the purposes and the overview of researches on UAV environment perception are introduced in this dissertation.Then the relevant principles and processes of obtaining three-dimensional information by using stereo vision are explained.After that,specific introductions of the type of environmental sensor—binocular camera that is used in obstacle detection in urban environment perception are proposed.After obtaining the three-dimensional scatter plot,Mean-Shift clustering method based on depth information is used to achieve the obstacle detection and the region of interest.Thereafter,the object identification for buildings within the interested areas is worked out by making use of the orthogonal property and HOG characteristic of the buildings.Finally,relevant cases of flight experiments are introduced.The final results show that the low-altitude urban environment perception has been realized by the method based on stereo vision that is adopted in this dissertation,and the effectiveness of the proposed algorithm for building identification has been proved.