Agricultural Uav Auto-Visual Navigation Based On Line Matching

For visual navigation algorithms, there’s no need to send and receive signals, establish the ground stations, and worry about the navigation precision affected by the area where the agricultural UAV are used around obstructions, so it is a relatively reliable navigation method that suitable for agricultural in UAV. In this paper, based on farmland crops points of ridge formation is rich in texture, field road, we can use the texture to extract the linear for auto-visual navigation. As follows are the main content of the paper:First of all, to extract the texture in the farm pictures made by camera edge, we use straight line extraction algorithm, synthesize to extract the texture edge fitting straight line. The matching thought of scale-invariant feature point descriptor are used to describe local characteristics of the extracted linear. And the method of nearest neighbor distance ratio are used to match.Secondly, in order to meet the navigation requirement of different agricultural UAVs that work at different areas, we propose agricultural UAV auto-visual navigation algorithm based on plane line matching and agricultural UAV autonomous navigation algorithm based on space line matching. We take advantage of the geometric relationship gained from matching the straight lines and plane projection lines to build constraint equation. In this paper, we make use of least square method and singular value decomposition when we solve the constraint equations. In order to reduce the noise disturbance and improve the precision of calculation. In this paper, we utilize the points on lines to adjust rotating matrix and the translation vector for secondly and obtain the exact pose parameters values of UAV eventually. We apply 3σ Guidelines to eliminate gross errors. Analyze the advantages and disadvantages of two algorithms by the simulation, verify the high precision, robustness and high real-time performance of this algorithm.Thirdly, in the case of space line vectors unknown, in this paper, we propose the agricultural UAV relative navigation based on image lines. We extract and match corresponding the lines in three sequential photographs in this algorithm, make use of the traits that one line has different projections in different images and there are many geometrical relationship between them, we can build constraint equation based on them. We analyze the precision robustness of the relative algorithm to test the feasibility of this algorithm.At last, we summarize the absolute navigation and relative navigation algorithm of agricultural UAV, and propose the outlook.