基于标记分水岭算法的插秧机器人导航路径检测  

Navigation path detection of rice transplanting robot based on marked watershed algorithm

作　　者：白云龙 傅彬[1] 史振华[1] 王健[1] Bai Yunlong;Fu Bin;Shi Zhenhua;Wang Jian(Shaoxing Vocational and Technical College,Shaoxing,312000,China)

机构地区：[1]绍兴职业技术学院,浙江绍兴312000

出　　处：《中国农机化学报》2021年第7期142-147,共6页Journal of Chinese Agricultural Mechanization

基　　金：浙江省2018年度公益技术项目(GN18E09000);2020年绍兴职业技术学院科研项目(SZK01)。

摘　　要：针对插秧机器人机器视觉导航路径检测鲁棒性差、受杂草和翻土影响严重的问题,提出一种基于标记分水岭算法的视觉导航路径检测方法。首先,采用灰度化处理、直方图均衡化和中值滤波对目标秧苗列和目标田埂进行预处理;然后,利用标记分水岭算法对识别目标进行图像分割,并通过均值法采集导航路径特征点集;最后,使用最小二乘法将特征点拟合成导航路径。试验结果表明,相比传统分水岭法和区域生长法,本文的导航路径检测方法具有最好的识别效果,在秧苗列和田埂上的检测精度分别达到93.4%和96.6%。Aiming at the problems of poor robustness of machine vision navigation path detection of rice transplanting robot and serious effects of weeds and soil turning,a visual navigation path detection method based on the marked watershed algorithm was proposed.First,grayscale processing,histogram equalization,and median filtering were used to preprocess the target seedlings and target ridges.Then the marked watershed algorithm was used to segment the image of the identified target,and the navigation path feature point set was collected by the mean value method.Finally,the least square method was used to fit the feature points into the navigation path.The experimental results showed that compared with the traditional watershed method and the regional growth method,the navigation path detection method used in this paper had the best recognition and detection accuracy on the seedling row and field ridge,reaching 93.4%and 96.6%,respectively.

关 键 词：插秧机器人 导航路径 机器视觉 标记分水岭算法 

分 类 号：S223.91[农业科学—农业机械化工程] TP242[农业科学—农业工程]