Abstract: In terms of the problems of applicability and poor antiinterference of visual navigation path extraction methods for agricultural machines in different scenes, an auxiliary linebased visual navigation path extraction algorithm for agricultural machines was proposed. First, the equalized farmland image was grayed out using the 1.8G-R-0.8B color model to obtain an image with an obvious distinction between the target and the background. Secondly, the image noise was removed using the OTSU method of threshold segmentation and the morphological processing method of opening and closing operations on the binary image. Finally, the corresponding auxiliary line processing was performed according to the vertical projection method and combined with the improved ROI method to extract the region of interest and determine the navigation localization points, and then the leastsquares method was used to fit the localization points to obtain the navigation path. The simulation results and comparison showed that the Euclidean distance of the extracted path was 1 001.9, the path extraction accuracy was improved by 47.9% compared with the traditional Hough method, and the extraction time of the highresolution image was shortened by 79.6%, which could meet the requirements of navigation path extraction of agricultural machinery and had higher universality.

Key words: machine vision, auxiliary lines, vertical projection, navigation path, ROI

摘要： 针对不同场景下的农机视觉导航路径提取方法适用性、抗干扰性差的问题，提出一种基于辅助线的农机视觉导航路径提取算法。首先，对均衡化处理后的农田图像采用1.8G-R-0.8B颜色模型进行灰度化，得到目标与背景区分明显的图像；其次，使用OTSU法阈值分割，对二值图像进行先开后闭运算的形态学处理方法去除图像噪声；最后，根据垂直投影法进行相应的辅助线处理，并结合改进的ROI方法提取感兴趣区域，确定导航定位点，进而最小二乘法将定位点拟合得到导航路径。仿真试验结果及对比表明：本文算法提取路径的欧式距离为1 001.9，路径提取精度相对于传统Hough方法提高47.9%，且对高分辨率图像提取时间缩短79.6%，满足农机具导航路径提取的要求的同时且具有更高的普适性。

关键词: 机器视觉, 辅助线, 垂直投影, 导航路径, ROI