Vision Positioning On Basis Of Robot Feeding System

Computer vision as one of the ways of getting environmental information,if it is usedto robot feeding system to get the location of object,it can improve the automation of thesystem,meet the flexibility of the robot,enhance its own ability.Machine vision how to getthe perfect target image information;how to identify the target from a complexbankground，extract feature，track the location and get the location and orientation;how topass the location and oritentation to robot;how to coordinate image acquisition、visionpositioning and robot. These are the key technologies in robot feeding system.With theseproblems for discussing in-depth.First on the basis of the whole of program of robot feeding system,building hardwareplatform.According to integrated control software coordinate image acquisition、visionpositioning and robot’s work and realize the robot intelligent crawl.Through experimentverify accuracy.On the basis of edge detection, propose a corner detection method, as the basis toform the edge of the object to form a closed curve. Use the methords of removing tooshort and isolated edge and merging adjacent curve that have similar curvature. Removeexcess corner with automatic serch. To verify the accuracy of the algorithm detectedthrough physical corner.On the basis of extraction coner detection flat workpiece centerpoint extraction methods, propose an irregular extraction method of the center point of theplane of the irregular object.According to examples to verify its accuracy.A positioning for flat objects without the calibration plane is proposed.This combinedwith movement of the plane of the target object,established the moving distance of thetarget object within the same time interval with distance relationship of the imagecoordinate system in the corner point moves. A method of positioning of a tracking of themovement of the workpiece is proposed.To validate its accuracy, Measuring the spatialdistance and calculating distance with the distance from the corner points.Based ontracking and positioning,it can judge these special cases such as overlap with the of theplane、tilt of the plane.Propose a new method of acquiring depth information.Finally, these algorithms are programmed, exploit the feeding system in visual positioningsoftware,debug it and embed it to integrated control software.The software accuracy of theverification can use robot crawling on material objects...