Motion Control Of UAV Based On The Guidance Of Binocular Vision

At present,the application of UAV is becoming more and more widespread.It can sense the external environment and its own state through the information collected by various sensors above,and can fly independently according to the information processed,the research on autonomous navigation of UAV is always in the hot development.Inertial Navigation System(INS)is the main system to realize autonomous navigation.Nowadays,there are many integrated navigation or integrated navigation systems,among which INS/GPS system is the most widely used.However,GPS has the limitation of weak local signal or no indoor signal.Binocular stereo vision is more and more widely used.It obtains disparity by stereo matching,so it obtains depth information based on disparity.Binocular system is a good positioning method with low cost and high accuracy.Visual navigation is also one of the hottest navigation systems currently studied,so this paper considers replacing GPS with binocular vision,and studies a vision navigation system based on inertial navigation system and binocular vision technology to obtain positioning information,the main work of this paper is as follows:(1)Based on the consideration of memory and processing speed,the hardware and software platform of binocular vision was built on the CM3 board of raspberry pie,on this platform,the calibration,binocular correction,image acquisition in various synchronization modes and stereo matching of binocular cameras are realized,the target area is selected and the location of the UAV is realized based on the binocular vision technology on the platform.(2)Because stereo matching is not compatible with matching speed and matching effect,a method based on detection of left and right target points and then disparity is designed.This method can shorten the time under the premise of better matching effect.In the experiment,the region to be measured according to the target color model is extracted based on the color model,then the BP artificial neural network algorithm is used to design the classifier to exclude the interference according to the feature vectors extracted from the region to be measured in the training samples,and finally the parallax is obtained based on the center of mass of the right and left views.(3)Location based only on binocular vision system requires depth calculation for multiple target points.Here,choosing the attitude angle information obtained from the inertial measurement unit to assist the location of UAV by coordinate transformation.In order to achieve better synchronization effect between image acquisition and IMU data acquisition,an IMU is designed on the CM3 board of raspberry pie instead of the IMU on the flight control system to obtain attitude angle information.(4)A binocular vision/INS integrated navigation system is designed for data fusion,and a 10-order state variable system model is designed based on Extend Kalman Filter algorithm to provide more accurate data for the realization of double closed-loop PID controller of inner-loop attitude and outer-loop position.(5)The Socket program based on UDP protocol is compiled to realize the communication between CM3 board of raspberry dispatch and ground station on UAV,so that the ground station system can track the navigation data information of UAV.