基于机器视觉的悬链线上吊钩的识别研究

发布时间：2019-04-13 16:19

【摘要】：悬链生产线在工业生产中应用广泛，如涂装、抛丸、烘干、牲畜屠宰等生产线上大量使用了悬链输送。目前悬链线上工件装卸仍然采用人工完成，依靠工人的体力搬上、卸下，工人劳动强度大，生产效率低，另外涂装等作业还会对人体健康产生严重的影响，所以悬链线上只有采用自动装卸设备上、下件才能满足产品质量和生产效率高的快速高效要求。为保证自动装卸工件，需要找到悬链线上吊钩和放置工件的位置，才能将工件放置到吊钩上，针对这一特点，本文将机器视觉引入悬链线上吊钩的识别中，，采用机器视觉确定吊钩的位置，自动装卸设备调整运动轨迹准确放置工件。本文研究内容包括以下几个方面：第一，完成了吊钩图像采集硬件系统的设计，包括照明系统、摄像机、采集卡等硬件。选用加拿大Point Grey公司的Bumblebee2平行双目摄像机对悬链线上吊钩进行图像采集工作，采用张正友平面标定方法完成了本课题的双目摄像机内外参数的标定工作，并对得到的参数进行了误差分析。第二，以家禽屠宰线上的吊钩为研究对象，针对此类吊钩的特点，首先采用中值滤波算法对图像进行了去噪处理。在比较各种特征匹配方法的基础上，采用SIFT特征匹配算法并结合计算投影变换矩阵的方法来求得两幅图像之间的匹配点，得到了匹配点之间的对应关系，最终根据三角形法则得到了吊钩的空间三维信息，为机器人自动识别吊钩提供了可靠的识别图像。第三，以熔模铸造车间悬链线上的吊钩为研究对象，对所采集的图像进行预处理后采用Hough变换将吊钩的识别区域缩小到目标范围，通过水平与竖直方向的灰度投影情况清晰地识别出目标区域。将两幅图像进行匹配，得到了吊钩目标识别区域的空间信息，并与吊钩的实际尺寸进行了比较。
[Abstract]:Catenary production line is widely used in industrial production, such as painting, shot, drying, livestock slaughtering and so on. At present, the loading and unloading of workpieces on catenary lines is still done manually, relying on the physical strength of the workers to carry them up and unload them. The workers have great labor intensity and low production efficiency. In addition, painting and other operations will also have a serious impact on human health. So only by using automatic loading and unloading equipment on catenary line can the next parts meet the requirements of high-speed and high efficiency of product quality and production efficiency. In order to ensure the automatic loading and unloading of workpiece, it is necessary to find the hooks on the catenary and the position of the workpiece so that the workpiece can be placed on the hooks. In view of this characteristic, this paper introduces machine vision into the identification of hooks on catenary. Machine vision is used to determine the position of the hook, and the automatic loading and unloading equipment adjusts the movement track to place the workpiece accurately. The research contents of this paper include the following aspects: firstly, the design of hook image acquisition hardware system, including lighting system, camera, acquisition card and other hardware, is completed. The Bumblebee2 parallel binocular camera from Point Grey Company of Canada was used to collect the images of the hooks on the catenary line, and the calibration of the internal and external parameters of the binocular camera was completed by using Zhang Zhengyou plane calibration method. The error analysis of the obtained parameters is also given. Secondly, taking the hooks on the poultry slaughtering line as the research object, according to the characteristics of this kind of hooks, the median filtering algorithm is first used to de-noising the image. On the basis of comparing various feature matching methods, the matching points between two images are obtained by using the SIFT feature matching algorithm and the method of calculating projection transformation matrix, and the corresponding relation between the matching points is obtained. Finally, the spatial three-dimensional information of the hooks is obtained according to the triangle rule, which provides a reliable recognition image for automatic recognition of the hooks by the robot. Thirdly, taking the hooks on the catenary of the investment casting workshop as the research object, the recognition area of the hooks is narrowed to the target range by Hough transform after pre-processing the collected images. The target area is clearly identified by horizontal and vertical grayscale projection. By matching the two images, the spatial information of the target recognition area of the hook is obtained and compared with the actual size of the hook.
【学位授予单位】：济南大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH22

【参考文献】