Research On Automatic Sorting System With Smart Camera And Robot

Due to the great development of machine vision and robotics, the industrial robots with visual sensors play an important role in food packaging, pharmaceutical production, non-destructive testing and automatic production line field. As the “Made in China 2025” plan proposed, the technology, machine vision and industrial robotics highly integrated, will have a great influence on the development of fully-automated smart factories, especially in the manufacturing and sorting field. In view of the problems that may arise in the automatic sorting system, which is based on smart camera and industrial robot in the Eye-to-hand form, the objective of this paper is to have a research on the key technology and theory of automatic sorting system. The main contents and results are as follows:1. A new approach to the automatic sorting system calibration working with six DOF robot in Eye-to-hand form is proposed, which is based on photogrammetry. In the calibration process, the working camera takes one image of the planar target on the conveyor belt. The image taken is then used to calculate the homography between the working camera and the planar target. Then the robot rotates around its A4 axis once, while a photogrammetry camera takes a group of images of the visual markers scattered on the robot before and after the rotation respectively. Then the robot rotates around its A1 axis once, and the photogrammetry camera takes a group of images of the planar target and the visual markers scattered on the robot after the rotation respectively. Based on the 3D reconstruction results of the two groups of images, the coordinate system of the robot is established and the geometric relationship between the planar target and the robot is achieved. Combined with the homography calculated, the transformation matrix between the working camera image plane and the robot coordinate system is implemented, which completes the whole calibration. The superiorities of the proposed method lie in that only two rotating movement of the robot and some images are required, and only linear computation is involved in the whole process. Experiments show that the proposed method is of high calibration accuracy.2. Telecommunication between smart camera and industrial robot, with real-time and reliability is established, which is based on configuration file written into the robotic controlled PC and limited the format of communication data strictly. The telecommunication with real-time and synchronization is established on the EtherNet/IP and synchronous blocking mode, which is easy to communicate with other equipment with Ethernet interface. Experiments show that the telecommunication is with real-time and reliability.3. A combination of Eye-hand recognition and location to the moving objects on conveyor belt is implemented. The article discusses the algorithm of image processing and template matching. By run-length coding and freeman segmenting background and objects, the global features of object on the conveyor extracted by fast and effective image process method, have rotation and translation invariance. Then performing the matching step—matching the object with the template based on features, if the object is similar to the template, the smart camera sends the coordinates of the object to the robot by telecommunication, thus robot easily positions the target object. Besides, this paper also deeply discusses the solutions that there is occlusion among several objects. By means of local features and binary description and RANSAC, the method could match the key-points fast and accurately.4. Based on the key technology and the proposed techniques, the prototype sorting system based on a single VC smart camera and a KUKA industrial robot was constructed, and the relevant software, developed under VS2013 was supported, which includes template parameters and calibration result calculating and smart camera and robot tests. Experiments show that key technology is robust.