基于单目视觉里程计的甘蔗收割导航线提取研究

打开文本图片集

中图分类号：S233.4 文献标识码：A 文章编号：2095-5553（2026）02-0196-07

Research on sugarcane harvesting navigation line extraction based on monocular visual odometry

Li Longqin，Lu Jingping，Zhang Yongxu， Feng Wukai， Lin Yundong （School of Mechanical Engineering，Guangxi University，Nanning， 53ooO4，China）

Abstract：To develop a self-navigation system suitable for sugarcane harvesters，a method for extracting navigation lines of sugarcaneharvesters basedonmonocularvisualodometry isproposed.Themoving imagesof theharvesterwere collected byusingamonocularcamera，and thethree-dimensional point cloud informationof thesugarcanefieldterrain was extracted byutlizinga monocularvisualodometry.Subsequently，thepositions ofsugarcane stem lines andsugarcane ridgesurfaces were extracted respectivelythrough theRANSACalgorithm，withtheir intersectionserving asanestimation point forthesugarcaneroot position.Finall，linearregresionwas usedtofitthestraight linesof the harvesting navigation lines.Experimental results demonstrate that themonocular visual odometry can track the pose of the image capturevehicle ata speedof25 frames persecondandrefresh thethree-dimensional point cloudat arateof 3frames per second.Themaximumangulardeviationbetweenthenavigation linefitedbythis methodand themanuallyextracted navigation line is 3.933∘ ，and the maximum distance from the sugarcane root to the navigation line is 0.164m .The maximum processing time per point cloud frame is 127.4ms

Keywords：sugarcaneharvester;monocular visualodometry；navigation line extraction;sugarcane field；point cloud

0 引言

我国是世界第三大甘蔗生产国，但机械化收割的相关技术仍不完善，导致收割效果不理想。（剩余11361字）