Design And Realization Of A Pick-And-Place System Based On 3D Vision System And Six - Axis Robotic Arm

• Main Authors: GAO,WEI-XUN, 高偉勛
• Other Authors: Wu, Hsien-Huang
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/2cx63v

碩士 === 國立雲林科技大學 === 電機工程系 === 105 === In recent years, among automation exhibitions around the world, more and more applications are using 3D computer vision with a mechanical arm for automatic picking and placement. They can be found in the factory automation with growing popularity, for example, in the automotive industry, textile, casting ... and so on. If the machines are used to replace the human manpower, they not only can improve the speed, but also can reduce the human error. Given an automation application, regardless of the type of production, the first step is to feed the material for processing. Raw materials are usually contained in the vehicle and carried to the factory, there are sometimes neatly arranged, but mostly are free to be placed. The purpose of this study is investigating the use of mechanical arm with stereo vision to achieve automatic loading to improve the production capacity and reduce the accident. In this thesis, the goal is to accomplish the pick and place of a pile of sliders for heat processing. Due to different load, the type and size of the slider is also relatively different, the maximum weight can reach 10kg. Obtaining the depth data of each individual slider by using TOF technology, we can then send data to the robot arm, so that the arm automatically clamp (pick) the slider. In order to achieve 1mm accuracy for two-dimensional coordinates, an area camera is attached in front of the robot arm to obtain a higher resolution image before grasping. In contrast, the depth acquired by the ToF camera has an accuracy of 3mm. The overall time for the processing is about 2 minutes and 10 seconds, but more than half of the time is used to wait for the heating procedure. The actual on-line test used 1600 sliders for verifying the pick and place process. Because of the texture and roughness of the slider surface, the current overall success rate is 97.25%.