The Position Attitude Estimation And Path Planning Of MAV Based On Distributed Vision

As the UAV has a increasingly wider application range in military, civil and inbusiness, it has also put forward to new requirements for UAV navigation control. Howto plan a reasonable and shortest path fast and effective in the flying area has become agreat challenge. On the other hand, the selection of sensors plays an important role inthe fast accurate navigation field. Visual navigation becomes a key point which takesthe advantage of wider detection range and more accurate information acquisition. Theprocess of realizing the navigation means to plan a collisionless path through theObstacles area, which has been the main research topic for UAV over these years.In this research, after the distributed vision system is built on the basis of visualnavigation theory, target identification which is based on color image is achieved byusing its characteristics. And in accordance with camera model, the measurement modelof3d reconstruction is built, by which3d gesture position information of air vehicle isobtained. Based on the information, it needs to plan a collisionless flying path in theflying area, namely to visual navigation range, and the path planning of air vehicle isachieved.Firstly, this article builds a distributed vision system platform based on“ShinyView” SG-100-D1C/M camera and Nuvo-1000industrial personal computer. Inaddition, the hardware distributed structure and software modularization programmingconstruct are also projected.Secondly, according to distributed vision system platform, projection algorithmand recognition algorithm based on OpenCV contour tree are applied to identify thelandmarks of air vehicle. Through the actual experimental results, the effectiveness ofthe two algorithms is verified, and with the information, two algorithm’s stability andaccuracy are contrasted. In order to reconstruct the3d model of the pose and positionfor air vehicle model, binocular ranging model is built. And real-time gesture of airvehicle is demonstrated better with the use of processes technology and modeling byOpenGL.Lastly, in order to plan the path of the aircraft movement area, the rapid exploration random tree algorithm is applied. Through the MATLAB simulation, its validity isverified. Then it is transplanted to the actual project successfully. In allusion toproblems in real situation, such as local minimum area, the algorithm is modified andimproved.