Research On The Technology Of Automatic Recognition For Position And Orientation Of Parts In Intelligent Coordinate Measuring Machine Base On Binocular Vision

Three Coordinate Measuring Machine（CMM） is one of accurate measuringequipment in modern manufacturing. The automation and intelligent of testing proceduresmake CMM having broad application space, and it can better fit fast development ofmanufacturing industry. The development trend of the new type of the three coordinatemeasurement is to achieve intelligent and automation. As the one of the key technology ofthe three coordinate measuring machines, the position and posture automatic identificationsystem for parts is the premises of intelligent detection. One of method to realize theposition and posture automatic identification is by capturing the image of measured partsuse CCD camera, and then recognize the posture, position and orientation.This dissertation is a sub-subject of "complex part measurement informationgenerated and obtain mechanism and intelligent three coordinates measuring keytechnology research"（project number51075119）, which is financially supported by theNational Natural Science Foundation of China. The major research and creative work ofthe thesis includes the following aspects:1) Single camera stereo vision measurement system is built up by advantage theprecision movement of CMM. Then define the main mission of recognition system andillustrate the principle and merit why adopt this system model to realize the position andposture automatic identification system for parts.2) Camera calibration. Camera calibration is one of the most important parts incomputer vision. The calibration of single camera stereo vision sensor has been researched.In the process of the calibration, extrinsic and intrinsic parameters of camera can becalibrated by only one time, which has simplified the calibration process and reducedcalibration workload.3) Image processing. Using image filtering, binarization, edge detection and so on todeal with the parts images by camera, finally obtain the complete edge of part image.4) Recognition of part pattern based on combined invariant moments and BP neuralnetwork. The paper use combined invariant moment features, including the HU invariantmoment, the affine invariant moments, the normalized inertia moment, and principalcomponent analysis are used for feature selection. Finally, the BP neural network was usedto recognize the way of the part.5) Stereo vision matching. A new method of stereo vision matching on the basis of displacement of image centroid is proposed. The center of edge image should be picked upbefore stereo vision matching, and then the offset distance between two centroids is usedas a restriction to accomplish stereo vision matching.6) Automatic locating the position and orientation of the part. With thethree-dimension information of part by stereo vision and the information from CADmodel, through calculating transformation relation between two coordinate systems， theorientation and position of the object can be determined.Some new methods are proposed through theoretical analysis and experiments onstructure of the position and posture identification system, camera calibration, imageprocessing, recognition of part pattern, calculation of position and orientation. The worklays a foundation for further improving the accuracy and simplifying the process.