Research On Estimation Of UAV’s Landing Parameters Based On Vision Information

Vision navigation system, as one of navigation technology, has become a hot research topic due to its advantages of passivity and complete autonomy. In this paper, some image processing algorithms are employed to extract the vision features based on image processing algorithms, then combination with computer vision and photography geometry technology, the position and attitude parameters of UAV(Unmanned Aerial Vehicle)can be estimated.Considering that image quality is easy to be effected by random noise, and noise not only reduce the quality of image but also destroyed the useful information image carries, a modified median filtering algorithm is introduced, which has the ability to overcome some defect of traditional median filtering algorithm, such as fixed window shape and unconditional filter. Besides, the results of experiments show that the modified median filtering algorithm owns better performance than that of traditional median filtering algorithm.After analyzing the characteristics of horizon of landing ground, a new method, based on gray projection theory, for searching horizon region is proposed. In the characteristic region, the Otsu criterion firstly is used to roughly extract possible feature points of horizon, then, the final feature points of horizon can be obtained using some screening rule.Besides, the horizon detection can be finished by using LMS algorithm. At last, compared with traditional methods of extracting the horizon, the method has been improved in terms of time.The next part is to employ the texture template matching algorithm to detect and track the runway. And then combination with the Sobel and Hough transformation algorithm, the extraction of two edge lines of runway can be finished when UAV enters landing field completely. Compared to the traditional Hough algorithm which extracts the line by detecting the peak, the Hough algorithm used in this paper has been modified to detect the two edge lines of runway based on known features.The transfer relation of world coordinate system, image coordinate system and camera coordinate system are introduced. Combination with the characteristic of horizon and edges of runway, the position and attitude parameters are estimated. The algorithm is finished using VC++ environment, and the simulation results show that the proposed method is feasible,and the quantitative analysis of error is finished based on the theory of Taylor series.