西红柿采摘机器人视觉系统的研究  

Research on Tomato-harvesting Robot Visual System

作　　者：郭凯敏[1] 崔天时[1] 张桢[1] 郭志强[1] 朱铁欣 谢学刚[1] 

机构地区：[1]东北农业大学,哈尔滨150030

出　　处：《农机化研究》2016年第12期217-221 226,共6页Journal of Agricultural Mechanization Research

基　　金：国家"863计划"项目(810028)

摘　　要：针对果实采摘机器人果实识别率低的问题,设计了一组用于西红柿识别和定位的双目立体视觉系统,为机器人的采摘作业提供更有利的条件。为此,采用Bumblebee双目立体视觉系统,基于成熟果实与植株颜色特征的差异进行图像分割,来识别成熟的西红柿;在完成相机标定、特征点提取和特征点匹配的基础上,通过三维空间定位获取果实的三维坐标。实验结果表明:该系统果实识别的整个过程平均耗时150ms,对成熟西红柿的识别率达到99%,测试误差在10mm以内,能够较好地满足西红柿采摘工作的要求。Aimed at the problem of low recognition rate of fruit picking robots,design a set of binocular stereo vision system is used to identify and locate tomato,provide more favorable conditions of picking robot operation. The Bumblebee binocular stereo vision system,the identification of visual system is developed through the difference between the color characteristics of ripe tomatoes and plant. after camera calibration,feature point extraction and feature points matching,the space coordinates is obtained through 3D reconstruction. The experiment results show that the identification rate of the vision system can reach 99%,and identification average time is about 0. 15 s,the absolute error of distance- measuring can be controlled within 10 mm. It can meet the needs of practical work very well.

关 键 词：农业机器人 双目立体视觉 相机标定 特征点提取 特征点匹配 三维重建 定位 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP242[自动化与计算机技术—计算机科学与技术]