农机自动导航技术是实现智能农机装备和智能农业机器人的一项基础性支撑技术，是实现农业机械智能化和自动化的重要保障；基于机器视觉的农机智能导航技术的广泛应用仍需在多个方面展开研发。其中，导航线提取算法是基于机器视觉导航的关键技术之一。本文针对插秧机实际工作情况设计了两种导航策略，并提出了相应的导航线提取方法。当农田中没有秧苗时，采用大津法分割图像、边界拓扑分析方法提取农田边界轮廓；当农田中有秧苗时，采用改进的超绿法提取秧苗，利用算法自动选取合适的秧苗中心点。利用最小二乘法和随机采样一致性算法拟合导航线，最终计算导航参数（横向偏差和航向偏差）。试验结果表明，在横向距离为100 cm时，不同角度下的横向偏差的平均误差为2 cm，标准差为3 cm，角度误差测量均值为0.14°，标准差为0.7°，该方法可以用于插秧机的视觉导航。
Agricultural machinery automatic navigation technology is a basic support technology to realize intelligent agricultural machinery equipment and intelligent agricultural robot, and is an important guarantee to realize agricultural machinery intelligence and automation; the wide application of agricultural machinery intelligent navigation technology based on vision still needs to be developed in many aspects. Among them, navigation line extraction algorithm is one of the key links of vision-based navigation. To address this question, two navigation strategies with corresponding guiding line extraction methods are proposed for rice transplanter. When no transplanted rice seedling is observed in the sight field, images are binarized by Otsu method, and ridge contours are found according to topological analyzation. Otherwise, rice seedlings are located by modified excess-green transformation, and subsequently centre points are found by proposed algorithm. With basic points of ridge contours or rice seedlings, a guiding line is computed using least square and RANdom SAmple Consensus （RANSAC）, and finally, navigation parameters, i.e., lateral deviation and course deviation, are obtained. Experiments proved its considerable applicability in visual navigation of rice transplanters. When true lateral deviation was 100 cm, the mean error of lateral deviation under different course deviation was 2 cm, with a standard deviation of 3 cm. Meanwhile, the mean error of course deviation under different lateral deviation was 0.14°, with a standard deviation of 0.7°.
[1] 张漫，季宇寒，李世超，等. 农业机械导航技术研究进展[J]. 农业机械学报，2020，51（4）：1-18.
[2] ENGLISH A, ROSS P, BALL D, et al. Vision based guidance for robot navigation in agriculture[A]//proceedings of the 2014 IEEE International Conference on Robotics and Automation （ICRA）, F, 2014 [C]. IEEE.
[3] BENGOCHEA-GUEVARA J M, Conesa-Munoz J, Andújar D, et al. Merge fuzzy visual servoing and GPS-based planning to obtain a proper navigation behavior for a small crop-inspection robot[J]. Sensors, 2016, 16（3）: 276.
[4] 乔智利，崔彦平，邹新光. 农田视觉导航基准线的识别与提取方法研究[J]. 农机化研究，2016，38（7）：124-128.
[5] 李勇，丁伟利. 基于暗原色的农机具视觉导航线提取算法[J]. 光学学报，2015，35（2）：229-236.
[6] Rovira-Más F, ZHANG Q, REID J F, et al. Hough-transform-based vision algorithm for crop row detection of an automated agricultural vehicle[J]. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, 2005, 219（8）: 999-1010.
[7] Sogaard H T, OLSEN H J. Determination of crop rows by image analysis without segmentation[J]. Computers and Electronics in Agriculture, 2003, 38（2）: 141-158.
[8] POYNTON C A. Rehabilitation of gamma[A]//proceedings of the Human Vision and Electronic Imaging III, F, 1998 [C]. International Society for Optics and Photonics.
[9] OTSU N. A threshold selection method from gray-level histograms[J]. IEEE transactions on systems, man, and cybernetics, 1979, 9（1）: 62-66.
[10] SUZUKI S. Topological structural analysis of digitized binary images by border following[J]. Computer vision, graphics, and image processing, 1985, 30（1）: 32-46.
