Research On Monitoring And Diagnosis Of Robot Palletizing Process Based On Vision

The research on state monitoring and fault diagnosis of robot operation process based on vision is one of the latest fields, and it has a great theoretical significance and comprehensive foreground of applications. This paper has done a detailed study about robot operation process based on the analysis of robot palletizing process and applicably of the theory and methodology of tracking moving targets and machine vision.The main content of research and result includes:Firstly, aimed at the box's recognition and the problem of environmental illumination changes, this paper has researched into the recognition theory based on characters of color and shape. The hue of color is steady when the environmental illumination varies, and at the same time box's color differs greatly from the ambient color, and the box's shape is rectangular. According on those, for one thing, the system separates the moving box from environment by hue, and then the system distinguishes the box and moving states based on image moment.Secondly, aimed at the fast or real-time, this paper has studied the visual tracking arithmetic of Polynomial fitting and Kalman filter. The box that is palletized has been tracked on the tracking window. When processing tracking image, the part image in the window is processed. So this means can shorten the time of processing irrelevant information.Thirdly, this paper has analyzes some dangers and faults which occur during robot palletizing process. And this paper has fitted the track curve of boxes moving by the theory of least squares. On this basis, it can be judged whether the position and speed of moving box is within the range of permission.Finally, this paper has designed and developed real-time software used graphic language LabVIEW and has experiments for a confirmation test. This paper uses a modularization procedure process methods and optimization techniques. So the system of monitoring and diagnosis has achieved optimal performance and image processing time for each frame is only about 20 milliseconds.A great number of simulation tests prove that the system of monitoring and diagnosis can recognize truly and fast the position of moving target, and dispose the shadow and parts of block. The processing speed is about 15 frames per-second to the color image with 768×576 resolution . Moreover, this system satisfies real-time requirements and can be used in actual industrial production.