Research On Application Of Machine Vision In Robot Palletizing System

With the rapid development of industry, the palletizing technology of domestic andforeign has realized the leaping progress. The early artificial palletizing, low throughput,low load, high labor cost, low handling efficiency, which can not meet the productionneeds. In the industrial production, the stacking robot widely used for automatedproduction in essence is a kind of common industrial robot, which mainly responsible forloading and unloading tasks, and use the method of teaching generally, preseting catchstarting point and placing point. This way of working is not able to the production lines,such as not being able to distinguish between the size of the workpieces, not qualifing theworkpieces, not carrying out sorting to the workpieces. The machine vision based andpalletizing robot, which has the function of human recognition, is very necessary toguarantee the product quality, reduce the labor cost, optimize work layout, improveproduction efficiency, increase economic benefit, realize production automation etc..Machine vision system is according to the image pixel color, brightness and otherinformation, the processing software system analysis the images and control the equipmentmove by the judgment result. With the infiltration of the application of machine vision toproduction line, large machinery manufacturing, precision assembly industry in areas suchas machine vision, leads to increased demand, also determines the machine vision will befrom the past simple data acquisition, transmission, judge action, gradually towards theopen direction, this trend will indicates that the machine vision and artificial intelligencetechnology, robot technology, automation technology, intelligent control technology forfurther combination, promote the process of industrial production, automation, intelligentrobot is the inevitable trend of the development of the times.This paper built the grasping intelligently platform on the industrial production linebased on the SIASUN SRH66axis industrial robot. Firstly, in order to improve theprecision and accelerate the speed of calibration algorithm, this paper presents an improvedcamera calibration algorithm. This method first uses the improved method to detect corners, finds location of the corner points of the calibration plate quickly, and then improves theaccuracy of corner subpixel according to the minimum error function, and selects centerregional calibration template results as the initial value to calculate the distortioncoefficient, last uses Levenberg-Marquardt algorithm to optimize the global parameters.Secondly, the experimental system recognized the workpieces by sift or hough transform.Thirdly, calculating the geometric center coordinates of the model and finally guiding themanipulator grab the model under-control the robot.Experimental shows that two algorithms can identify and match well for in the light ofdifferent workpieces which have different features. can obtain the position of theworkpieces, determine the pose by the angle between workpiece axis and the X axis of theimage plane，finally guid the manipulator grab the workpieces.