基于立体视觉匹配的采摘机器人目标定位方法  

Target location method of picking robot based on stereo vision matching

作　　者：金何[1] JIN He(College of Automotive and Electromechanical Engineering,Lu'an Vocational Technical College,Anhui Lu'an 237158,China)

机构地区：[1]六安职业技术学院汽车与机电工程学院,安徽六安237158

出　　处：《齐齐哈尔大学学报（自然科学版）》2023年第6期47-52,共6页Journal of Qiqihar University(Natural Science Edition)

基　　金：2021年安徽省高等学校自然科学研究项目“采摘机器人目标识别与定位研究”(KJ2021A1357)。

摘　　要：为了解决采摘机器人目标定位偏差大的缺陷,提出基于立体视觉匹配的采摘机器人目标定位方法。引入立体视觉模型获取目标图像,采用线性平滑滤波模板消除目标图像的噪声;应用Bouguet算法立体校正左、右图像对;应用Sobel算子检测采摘目标边缘,确定目标的采摘中心点,通过立体匹配确定左、右图像中采摘目标的对应关系,完成采摘中心点的正确匹配,实现采摘机器人目标精准定位。测试结果显示,应用提出方法获得的采摘中心点匹配精度更高,采摘机器人目标定位误差最小值为1.0%,降低了目标定位偏差,应用性能较佳。In order to solve the problem of large target positioning deviation in harvesting robots,a target positioning method for harvesting robots based on stereo vision matching is proposed.Introduce a stereo vision model to obtain the target image,and use a linear smoothing filter template to eliminate noise in the target image.Apply Bouguet algorithm for stereo correction of left and right image pairs.Apply the Sobel operator to detect the edge of the picking target,determine the picking center point of the target,and determine the corresponding relationship between the picking targets in the left and right images through stereo matching,completing the correct matching of the picking center point and achieving precise target positioning of the picking robot.The test results show that the proposed method achieves higher matching accuracy for the picking center point,with a minimum target positioning error of 1.0%for the picking robot,reducing the target positioning deviation and achieving better application performance.

关 键 词：立体匹配 空间定位 采摘机器人 立体视觉 目标识别 目标定位 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]