智能码垛机器人关键技术的研究及开发

发布时间：2018-06-23 23:18

本文选题：智能码垛机器人 + 双目视觉　；参考：《河南科技大学》2015年硕士论文

【摘要】：当今时代,机器人的应用越来越普遍,尤其是随着制造业与物流行业的发展,机器人码垛技术受到越来越多人的关注。但是,国内码垛机器人技术的研究和应用程度较低,智能化程度有限,相关技术不够成熟。造成这些问题的原因一方面是由于国外机器人对我国机器人市场的冲击,另一方面是因为国内相关技术研究力量薄弱。为提高国内机器人码垛的智能化程度,本文对码垛机器人的视觉系统、多传感器信息融合等关键技术展开研究。分析国内外码垛机器人的研究进展和应用状况,总结国外码垛机器人应用特点和国内码垛机器人在研究和应用上存在的不足,提出本文研究的重点内容;根据实验与研究的要求,选择合适的摄像机、镜头和图像采集卡等组成本文的双目视觉系统;分析视觉系统坐标系和摄像机成像模型,研究双目视觉系统的测量原理以及双目视觉系统的标定方法;分析双目视觉测量数学模型,得出双目视觉系统空间坐标获取方法;为获得较高测量精度,建立双目视觉测量精度分析模型,以几何学的方式对焦距和基线与光轴夹角等因素对测量精度的影响进行分析,得出减小误差的方法;采用MATLAB标定工具箱和平面圆点靶标定的方法分别对双目视觉系统进行标定,对比两者的标定的精度,选择两者中精度较高的标定方法;为了更加精确的获取码垛对象测量关键点的坐标,提出边缘点拟合直线求交点方法进行测量关键点定位,提高匹配点定位精度;对ABB IRB2400型机器人进行运动学正解、逆解和工作空间求解,并对其进行路径规划;根据本文实验要求选择合适的传感器,构成多传感器信息融合码垛对象识别系统;对比多传感器信息融合方法的特点,采用BP神经网络对视觉信息、力觉信息和热觉信息进行融合的方法对码垛对象进行识别与判断。搭建码垛机器人视觉系统与抓取实验平台,对视觉测量、轨迹规划和多传感器信息融合的理论方法进行实验验证。实验结果表明双目测量相对误差为0.87%~2.51%,传感器融合系统判断准确率可达91.7%,验证了本文方法的可行性。
[Abstract]:Nowadays, the application of robot is becoming more and more common, especially with the development of manufacturing and logistics, more and more people pay attention to the technology of robot palletizing. However, the research and application of palletizing robot technology in China is low, the intelligence degree is limited, and the related technology is not mature enough. The causes of these problems are due to the impact of foreign robots on the robot market in China on the one hand, and the weakness of research on related technologies in China on the other hand. In order to improve the intelligence of palletizing in China, the key technologies such as vision system of palletizing robot and multi-sensor information fusion are studied in this paper. This paper analyzes the research progress and application status of palletizing robot at home and abroad, summarizes the application characteristics of foreign palletizing robot and the shortcomings in research and application of domestic palletizing robot, and puts forward the key content of this paper. According to the requirements of experiment and research, the binocular vision system is composed of suitable camera, lens and image acquisition card, and the coordinate system of vision system and camera imaging model are analyzed. The measurement principle of binocular vision system and the calibration method of binocular vision system are studied, the mathematical model of binocular vision measurement is analyzed, and the method of obtaining the spatial coordinate of binocular vision system is obtained. The precision analysis model of binocular vision measurement is established, and the effect of focal length and angle between baseline and optical axis on measurement accuracy is analyzed by geometry, and the method of reducing error is obtained. The binocular vision system is calibrated by using MATLAB calibration toolbox and the method of plane dot target. The calibration methods with higher accuracy are selected by comparing the accuracy of the two calibration methods. In order to obtain the coordinates of the key points in the measurement of palletizing objects more accurately, the method of edge point fitting and straight line intersection is proposed to locate the key points in order to improve the accuracy of matching points, and the kinematics forward solution of ABB IRB2400 robot is presented. Inverse solution and workspace solution, and path planning; according to the experimental requirements of this paper, select the appropriate sensor, constitute multi-sensor information fusion palletizing object recognition system; compare the characteristics of multi-sensor information fusion method, BP neural network is used to identify and judge palletizing objects by fusion of visual information, force information and thermal information. The vision system and grab experiment platform of palletizing robot are built to verify the theory and method of vision measurement, trajectory planning and multi-sensor information fusion. The experimental results show that the relative error of binocular measurement is 0.87 and the accuracy of sensor fusion system can reach 91.7. The feasibility of this method is verified.
【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41;TP242

【参考文献】