摘要
针对花椒市场需求量大、采摘困难的现状。笔者设计了自动识别花椒系统,通过对比4种识别算法的性能,采用识别效果最好的K-means聚类算法对花椒果实目标进行提取,针对花椒串生长具有离散性特点提出用平面内质点系模型求出花椒串质心。提出了Otsu算法与K-means算法结合图像相减的方法识别出花椒的结果母枝,然后根据小孔成像的基本原理和凸包理论得出单目视觉的花椒深度信息。最后以到花椒质心最短距离为限定条件确定母枝上采摘点所在的直线段,经过坐标转换求出直线段上采摘点的三维世界坐标。
In view of the current situation that the market demand of prickly ash is large and it is difficult to pick prickly ash.This research designs an automatic recognition system of prickly ash. Four recognition algorithms are compared in their performance, and K-means clustering algorithm is used to extract the Zanthoxylum bungeanum fruit. Considering the discrete characteristics of the growth of Zanthoxylum bungeanum fruit bunch, the in-plane particle system is used to obtain the centroid of Zanthoxylum bungeanum bunch. A combination of Otsu algorithm and K-means algorithm is proposed to identify the fruiting mother branch of Zanthoxylum bungeanum. According to the basic principle of pinhole imaging and convex hull theory, the depth information of prickly ash is obtained with monocular vision. Finally, the shortest distance to the centroid of prickly ash is taken as the limited condition to determine the three-dimensional world coordinates of the picking point on the mother branch. The monocular vision system designed in this research is installed at the end of the manipulator, which can effectively avoid the occlusion problem in the picking process with the movement of the manipulator.
作者
杨萍
郭志成
YANG Ping;GUO Zhicheng(College of Mechanical Engineering,Lanzhou University of Technology,Lanzhou 730050,China)
出处
《河北农业大学学报》
CAS
CSCD
北大核心
2020年第3期121-129,共9页
Journal of Hebei Agricultural University
基金
兰州市科技局(2019-1-126).
关键词
K-MEANS聚类算法
花椒
视觉定位
特征点匹配
凸包面积
K-means clustering algorithm
Zanthoxylum bungeanum
visual positioning
feature point matching
convex hull area
作者简介
第一作者:杨萍(1964-),女,甘肃兰州人,博士,教授,主要从事机器人研究.E-mail:1059342855@qq.com。
