【作者】 任烨；

【导师】 蒋焕煜；

【作者基本信息】 浙江大学 ， 生物系统工程， 2007， 硕士

【摘要】 基于机器视觉技术的机器人在设施农业中的应用已经越来越广泛。移栽机器人作为育苗工厂化生产的重要设备，在提高育苗工作效率上有着重要意义。本文应用机器视觉技术识别穴盘幼苗的生长情况，并利用多特征判断出适合移栽的幼苗，对其定位。同时设计了合理的末端执行器，利用计算机及PLC控制，实现对机械臂及末端执行器的定点控制。主要研究内容如下：1、综述了国内外设施农业中机器人的研究现状和机器视觉在机器人中的应用，特别阐述了国内外移栽机器人的研究现状，并指出了国内外研究中存在的问题。2、建立了适合本研究的机器视觉系统。该系统由CCD摄像头、图像采集卡、计算机、6支荧光灯，光照箱等组成。3、建立了适合本研究的试验平台。该试验平台由机器视觉系统、控制系统、传输系统及移栽系统四部分组成。4、基于机器视觉技术的番茄幼苗的识别和位置检测。通过对R、G、B三个分量的不同线性变换的研究，确定了适合本研究的图像分割的彩色特征变换，从背景中提出幼苗植株。然后采用单连通域分析算法提取每个穴孔中幼苗的叶片面积，并根据此特征对幼苗进行分类，判断出适合移栽的幼苗，并记录适合移栽幼苗的穴孔中心位置。实验验证，该机器视觉系统对于50孔穴盘番茄幼苗识别准确率为98.7％，对于72孔穴盘幼苗识别准确率为98.14％。5、研究了移栽机器人的控制系统，该系统由计算机、PLC、执行部件、反馈传感器等组成。计算机通过RS232串行通信器实现与PLC之间的通信，将视觉系统获取的位置信息传送至PLC，PLC发送信息驱动步进电机及电磁阀，运用电路控制和气路两路控制实现了对机械臂和末端执行器的运动控制。对移栽策略进行了分析，比较了几种不同的移栽路径，最后用贪心算法对路径进行了优化，缩短了总路径。6、根据穴盘幼苗的特性，开发了移栽机器人的末端执行器。末端执行器由气缸驱动，采用手指式抓取方式。用三种不同的手指分别作了比较试验，试验结果表明，采用两个铲式手指能更有效地抓取及释放幼苗块，完成移栽过程，移栽成功率为82.5％。更多还原

【Abstract】 Robots based on machine vision have been widely applied in facility agriculture. As a basilicequipment in the nursery factory, the transplanting robot plays an important role in improving thenursery efficiency. This research studied the application of machine vision to recognition and location ofplug seedlings, which are suitable to transplant decided by the characters of the area and perimeter ofleaves. Also an end-effector was designed to grasp, hold, insert, and release a plug. The movement ofthe manipulator and the end-effector was controlled by the computer and PLC. Main content and resultswere as follows:1. The research achievements in the field of facility agriculture robot and the application ofmachine vision in agriculture robot were reviewed, especially the development of the transplantingrobot, and the existing problems were put forward.2. A suitable machine vision system was set up for this research, which is composed of a grabber,a CCD camera, 6 fluorescent lamps and an industrial computer.3. A suitable experiment system was set up for this research, which is composed of four parts:machine vision system, control system, convection system and grasp system.4. The algorithm of recognition and location of tomato plug seedling based on machine visionwas developed. Through the analyses of linear transformation Of R, G and B in image shot by colorCDD, the color image was transformed to grayscale by proper color feature, and then was segmented bythreshold. The seedlings were recognized from background. Then the area of seedling leaves wasextracted using the blob analysis of each cell. The seedlings which need transplanted were recognizedand the position was recorded. The result shows the accuracy of machine vision system for the seedlingsof 50 cells was 98.7% while for the seedlings of 72 cells was 98.14%.5. The control system of transplanting robot was established. This system was composed of acomputer, a PLC, executive instrument, sensor and so on. RS-232 was used to communicate betweencomputer and PLC. Computer sends the position of seedling got from machine vision system to PLC,and then PLC sends the control information to the actuator of step motor and the electromagnetic. Themovement of the manipulator and the end-effector was controlled through the circuit and the air circuit. At last, the strategy of transplanting was analyzed, and different ways of transplanting were compared.The path was optimized based on greedy algorithm, and the total path became short.6. The end-effector was developed based on the character. The end-effector with fingeredgrasping was drove by two cylinders, and three kinds of fingers used to grasp, hold, insert, and release aplug were compared in the experiment, the result shows that the shovel fingers play the best behave, thesuccess rate is 82.5%.更多还原