A Workpiece Positioning Technique Using Region-based Obiect Segmentation And Tracking

Workpiece moving is a big usage scenario for robotic application.Nowadays,there is a great demand for intelligent workpiece load-unload robots,especially for those small enterprises.The workpiece load-unload robots can be divided into two types:without vision systems and with vision systems.Robots without vision systems usually suffer from time-consuming artificial calibration and low accuracy.Nonetheless,robots with vision system and conveyor belt has high requirements for environment and equipment.This thesis proposes an Eye-in-Hand based coordinate robot system.With the proposed system,a novel workpiece positioning technique based on monocular vision system is demonstrated and investigated.The major work of this thesis is as follows:1.The definitions of four image-related coordinate systems,as well as the relationships between different coordinates systems are analyzed.The mutual mapping between two-dimensional image pixels and three-dimensional space points are derived.2.Firstly,the color and geometric features of the stacking boxes(which is used for holding workpieces)are studied.Secondly,Hue and saturation(HS)value are utilized for distinguishing the stacking box and the background.Then,the approximate area of the stacking box are calculated with Hough transformation and the line cluster algorithm.Finally,the ROI(region of interest)of the stacking box can be obtained based on inversed mapping three-dimensional coordinates.3.Based on the ROI of the stacking box,the accurate location of the workpiece can be obtained with the threshold method.Firstly,a channel-histogram-based K-means algorithm is proposed for clustering the image pixels.Then,a comprehensive distance is defined for evaluating the effectiveness of the K-means algorithm.Finally,an approximate position of the workpiece is calculated.A more accurate positioning result is obtained by the proposed system with Watershed algorithm.4.The resolution image is further optimized with a tracking operator.The tracking operator can make the stacking box fill the frame of the picture as much as possible and lie in the center of the picture.