Research On The Technology Of Measurement Of UAV’s Attitude Based On Image Processing

The UAV has the advantages of light weight,small size,strong adaptability and flexible maneuver.In recent years,it is more and more widely used in military,civil,scientific research and other fields,attitude angle,as an important flight control parameter of UAV,plays an important role in the function development of UAV.Compared to traditional navigation methods based on inertial sensors and global positioning system,the measurement of UAV attitude based on image processing technology is a new attitude measurement method,and it has become the focus of research at home and abroad.The innovation of this thesis is modeling the UAV attitude measurement model and the attitude angle error analysis model,and then combine the image processing technology,complete the calculation of three attitude angles of the pitching angle,roll angle and heading angle of the UAV at all times.Finally analyze the theoretical error of the attitude angle calculated by the solution.In view of the above,the following work is done in this thesis.(1)Introduce the coordinate system of UAV attitude measurement model,such as geographic coordinate system,body heading coordinate system and cloud platform coordinate system,and calculate their transformation matrixes.(2)Adopt the Harris feature point detection algorithm and the SIFT scale invariant feature detection and matching algorithm to complete the extraction and matching of the image features,but in the comparison of these two algorithms in image rotation and scale change,their existence the flaw,Finally we use the matching algorithm of SURF feature point detection based on RANSAC image feature extraction and matching,the feature extraction speed of the algorithm is faster than SIFT 3~5 times,and the matching performance is still a good performance.(3)Complete the attitude measurement model of UAV and the modeling of the error model in theory,and the simulation analysis isalso completed.The RANSAC based SURF algorithm is used to extract the feature points of 24 images and match the feature points to obtain the change of the azimuth and high angle of the image.We extract any 3 pairs of matching points from two matching graphs to calculate the azimuth and high and low angles of the image,combined with the UAV attitude measurement model,the three attitude angles of the UAV at each time are calculated.Because of any 3 pairs of feature points,we can calculate the attitude angle of the UAV can,so there is a great error in the results.Based on this,the thesis uses the least square method.from the imageto which to be registered extracte 18 pairs of feature points,and calculate 6 attitude angles by the solution.The 6 attitude angles are used to fit a value that is minimal to its residual,and it is used as the final attitude angle.the theoretical error of the optimized attitude angle is compared with the actual error of the measurement result.It is found that the actual error of pitch angle and roll angle all fluctuates around the theoretical error,and the error range is less than 3,while the theoretical error of side slip angle is 1.7,so it is verified that the model is available.