自动上料机器人视觉定位系统研究

发布时间：2018-05-10 20:39

本文选题：安全套 + 机器人　；参考：《沈阳工业大学》2017年硕士论文

【摘要】：自动上料机器人作为提高现代工业生产线自动化水平和解放生产力的有效途径,将其引入到安全套电干检环节,将对实现安全套出厂检测流程的自动化起到重要作用。机器视觉是当下自动上料机器人的一个热门研究领域,是自动上料机器人感知和获取外部环境信息的重要途径,同时也是指导自动上料机器人完成对物料的识别、定位、搬移等自主动作的关键技术。因此,本文将对安全套电干检自动上料机器人的视觉定位系统进行研究,着重对以下四个方面进行阐述。首先,介绍了现有安全套主流检测方法及电干检机工作流程。针对电干检机所存在的无法自动上料问题,提出了自动上料机器人+电干检机的新型自动电干检系统。其次,为了满足自动上料机器人生产工艺所需的实时性、定位精度要求,需对安全套特征区域的识别与定位算法进行研究,结合该区域的边界曲率特点,提出了一种预检测+精检测的两步检测算法和与之相匹配的形心定位算法。在预检测阶段,采用FAST算法中的Bresenham圆对已提取的安全套边界点集进行曲率分类,根据边界曲率趋势筛选出特定的边界点集并求出矩形掩膜区域。在精检测阶段,在矩形掩膜区域内生成ORB特征算子和BRISK描述子。采用最近邻域算法进行模板匹配,利用RANSAC算法剔除误匹配。通过对预检测后所提取边界点集中的一点以其为中心做它的Bresenham圆邻域,并以该圆邻域与边界的两个交叉点的中点作为该区域的形心,以实现对安全套特征区域的形心定位。再次,在VS2013平台上编写基于Opencv2.4.13视觉库的视觉定位软件,该软件融合了上述两步检测算法及形心定位算法,可以完成对于安全套的边缘检测、特征点检测与匹配、特征区域形心定位,同时兼具摄像头标定、校准以及通信功能。最后,将上述软件与实验装置进行组装和测试。结果表明,两步算法能够有效的规避在全局范围内进行费时的描述子生成过程,比单纯的ORB+BRISK、BRISK等算法快5-8倍;同时继承了ORB与BRISK算法的旋转不变形和尺度不变性,提升了对安全套顶部区域形变时的识别与定位精度。以两步检测算法为核心编写的视觉定位软件满足自动上料机器人的实际需求。
[Abstract]:As an effective way to improve the automation level and liberate the productivity of modern industrial production line, the automatic feeding robot will play an important role in the realization of the automation of condom testing process. Machine vision is a hot research field of automatic feeding robot at present. It is an important way for automatic feeding robot to perceive and obtain external environment information, and it is also a guide for automatic feeding robot to identify and locate materials. The key technology of autonomous movement, such as moving. Therefore, this paper will study the visual positioning system of the automatic feeding robot for electrical testing of condoms, focusing on the following four aspects. First of all, introduced the existing mainstream condom testing methods and electrical dry machine workflow. In order to solve the problem of automatic feeding of electric dry detector, a new type of automatic electric dry check system of automatic feeding robot is proposed. Secondly, in order to meet the requirements of real-time and positioning accuracy of automatic feeding robot, it is necessary to study the identification and location algorithm of condom feature area, combining with the boundary curvature characteristics of the region. A two-step detection algorithm and a matching centroid location algorithm are proposed. In the phase of pre-detection, the Bresenham circle of FAST algorithm is used to classify the extracted condom boundary points. According to the trend of boundary curvature, a specific set of boundary points is selected and the rectangular mask region is obtained. In the stage of fine detection, ORB feature operators and BRISK descriptors are generated in the rectangular mask region. The nearest neighborhood algorithm is used for template matching, and the RANSAC algorithm is used to eliminate the false matching. In order to realize the centroid orientation of condom characteristic region, the center of the two intersection points of the circle neighborhood and the boundary is taken as the center of the Bresenham circle neighborhood, and the center of the two intersection points of the circle neighborhood and the boundary is taken as the center of the point of the extracted boundary point set after the pre-detection. Thirdly, the visual location software based on Opencv2.4.13 vision library is written on VS2013 platform. The software combines the above two-step detection algorithm and centroid location algorithm, which can complete the edge detection, feature point detection and matching for condoms. Feature area centroid location, as well as camera calibration, calibration and communication functions. Finally, the software and the experimental device are assembled and tested. The results show that the two-step algorithm can effectively avoid the time-consuming descriptor generation process in the global scope, which is 5-8 times faster than the simple ORB BRISKBISK algorithm, and inherits the rotation invariance and scale invariance of the ORB and BRISK algorithms. Improved the identification and positioning accuracy of the top area deformation of condom. The visual positioning software based on the two-step detection algorithm can meet the practical requirements of the automatic feeding robot.
【学位授予单位】：沈阳工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】