基于机器视觉的蒜种识别夹取试验台设计与试验  

Design and Experiment of Garlic Seed Recognition Clamping Test-bed Based on Machine Vision

作　　者：罗昌志 邸志峰 向阳 赵男 周进 朱正波 Luo Changzhi;Di Zhifeng;Xiang Yang;Zhao Nan;Zhou Jin;Zhu Zhengbo(College of Agricultural Engineering and Food Sicences,Shandong University of Technology,Zibo 255090,China;Shandong Academy of Agricultural Machinery Science,Jinan 250214,China)

机构地区：[1]山东理工大学农业工程与食品科学学院,山东淄博255090 [2]山东省农业机械科学研究院,济南250214

出　　处：《农机化研究》2024年第9期72-81,共10页Journal of Agricultural Mechanization Research

基　　金：山东省农科院科技创新工程项目(CXGC2022D08);国家自然科学基金青年基金项目(52005309)。

摘　　要：为实现大蒜的定向精确播种,提出一种基于机器视觉的蒜种精准识别与夹取方法,并设计蒜种识别夹取试验台,采用机器视觉方式确定蒜种的位置、朝向,使用机械臂夹取蒜种。采用阈值分割的方法分割蒜种和背景,对图像进行去噪,提取蒜种轮廓,根据识别像素点计算蒜种质心、朝向,经过识别训练和参数优化,确定识别阈值为110,识别成功率达到95%以上,平均识别时间0.06s。以金乡白皮蒜蒜种为试验对象,分析确定装置性能的指标为夹取成功率,影响因素为传送带速度、蒜种尺寸、种子夹取高度,通过三因素三水平正交试验分析,运用Design Expert软件对装置的关键结构参数进行优化,建立夹取成功率和各因素的回归模型,模型系数R_(2)^(2)为0.9433。试验结果表明:当传送带速度0.11m/s、种子夹取高度3.5mm、选取投影尺寸510~540mm2时,夹取成功率为93.73%。研究结果可为蒜种分级和大蒜的定向播种提供一定的参考。In order to realize the directional and accurate planting of garlic,a method of accurate identification and clamping of garlic seeds based on machine vision was proposed.The garlic seed identification and clamping test-bed was designed,the position and orientation of garlic seeds were determined by machine vision,and the garlic seeds were clamped by mechanical arm.The method of threshold segmentation is used to segment the garlic species and the background,denoise the image,extract the contour of the garlic species,calculate the centroid and orientation of the garlic species according to the recognized pixel points,and determine the recognition threshold of 110 after recognition training and parameter optimization.The recognition success rate is over 95%,and the average recognition time is 0.06s.Taking Jinxiang white skin garlic as the test object,the performance index of the device is determined to be the clamping success rate.The influencing factors are the speed of the conveyor belt,the size of garlic seeds and the clamping height of seeds.The key structural parameters of the device are optimized through three factor three level orthogonal test analysis and design expert software,and the regression model of the clamping success rate and each factor is established.The model coefficient R_(2)^(2)is 0.9433.The test results showed that:the speed of the conveyor belt was 0.11m/s,the height of the seed clamping was 3.5mm,and the projection size was 510-540mm^(2).At this time,the clamping success rate was 93.73%.The test bench realizes the functions of accurate identification and automatic grasping of garlic seeds.The results of this paper provide some reference for garlic seed classification and garlic directional sowing.

关 键 词：机器视觉 蒜种 机械臂 朝向识别 精准抓取 

分 类 号：S237[农业科学—农业机械化工程]