基于单目视觉的田间菠萝果实识别  

In-field pineapple recognition based on monocular vision

作　　者：李斌[1] Ning Wang 汪懋华[1] 李莉[1] 

机构地区：[1]中国农业大学“现代精细农业系统集成研究”教育部重点实验室,北京100083 [2]Oklahoma State University, Stillwater, Oklahoma 74078, USA

出　　处：《农业工程学报》2010年第10期345-349,共5页Transactions of the Chinese Society of Agricultural Engineering

基　　金：广东省农业机械装备技术公共实验室开放基金课题

摘　　要：菠萝果实的准确识别是菠萝采摘机器人视觉系统的关键。针对田间复杂环境中的青色菠萝,采用图像处理技术、数学形态学方法,识别菠萝果眼,并获取果眼的中心点信息;引入层次聚类分类方法,对多个中心点进行聚类分析;求取最多点集的中心坐标,作为菠萝区域的近似形心。选用在广东湛江菠萝田间拍摄的35幅图像作为评价样本,经过运算,迎光条件下形心识别正确率达到85%。本研究为菠萝采摘机器人的田间果实识别提供了有效的技术方案。Recognition of pineapple fruit is key to the vision system of pineapple harvesting robots.In this paper,image processing and morphological technologies were employed to recognize the fruit eyes and get the central points information of the eyes on the green pineapples in field;then,hierarchical clustering algorithm was used to cluster the central points.By getting the central coordinates of the dataset containing the most fruit-eye central points,the approximate shape center of the pineapple was calculated.35 images taken in pineapple field in Guangdong province were selected as the validation dataset.Experimental results showed that,the fruit correct-recognition rate could reach 85% in light-facing conditions.The research can provide effective technical solutions for the in-field fruit recognition to pineapple harvesting robots.

关 键 词：图像处理 视觉 形态学 菠萝 识别 层次聚类 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置] TP391.41[自动化与计算机技术—控制科学与工程]