Research On The Technology Of Measurement Of UAV's Attitude Based On Image Registration

The attitude estimation of UAV based on image processing technology is a new assisted method for the traditional pose measurement. Due to the advantages of passive and independent, it has become a hot research topic at home and abroad. In this thesis, the novelty is that a continuous image acquisition of the ground target is carried out by means of UAV imaging facility. Combined with computer vision, the image matched feature points extracted by image processing algorithm and the coordinate geometry transformation to finish the attitude estimation of UAV.In order to complete the attitude measurement, the UAV should be placed in a suitable coordinate system, the coordinate system can describe the state when the attitude have changed and establish the relationship between the coordinate system. The process of conversion also need to consider the UAV,and platform and three of the earth's coordinate system, complete the matrix conversion between each other and simplify the process.To extract feature points from the image, we compared with the classical Harris feature point extraction algorithm, SIFT feature points extraction algorithm and multi-scale wavelet transform energy accumulation feature points detected method based on wavelet transform.With the change of image scale and rotation, the performance of the three methods is compared with that of the feature point detection, and the performance of multi-scale wavelet transform energy accumulation feature points detected method is more excellent.The similarity measurement technique is introduced to measure the feature points between the two images which have geometry transformation. Compared the common features of similarity measurement technique based on the normalized cross correlation with the feature point similarity measurement technique based on Fourier-Mellin transform, the performance of the feature point similarity measurement technique based on Fourier-Mellin transform is better when images have complex geometry transformation. In order to further improve the computation speed of the similarity measure method of scale and rotation invariant feature points, the image trace transform is introduced.The pair feature points of the image acquired by UAV's camera are found by image processing methods, and the transformation parameters are obtained according to the affine mathematical model.The attitude measurement is accomplished by the fusion of UAV attitude model and the parameters. In the end,we analyze the error of attitude measurement.