On-machine 3D Measurement Of Large-scale Workpiece Based On Binocular Vision

In numerical control machine and other automated manufacturing systems, using advanced on-machine 3d dimensional measuring technology to detect the size of large-scale workpiece quickly and accurately makes a great difference to the processing cost,quality and efficiency of metal-cutting. However, the traditional measuring methods and equipments cannot meet the most modern production enterprises measure requirements. In recent years, measuring technology based on vision, which has the advantages of non-contact, measure fast, high accuracy and real time, has developed rapidly and becomes to the most promising technology in solving measure problems of manufacturing industry.In this paper, according to the stereo vision principle and characteristics of on-machine measure of large-scale workpiece, some key techniques problems were studied, including image acquisition, image preprocessing, system calibration, feature extraction and matching, 3D information acquisition, dimensional measure and so on. Based on the above researchs,an on-machine measuring system which not only could obtain the 3d information but also could measure the critical dimensions of large-scale workpiece was built.Firstly, an image acquisition system was designed,including the cameras, lenses, image acquisition card, lighting system, mechanical structure and so on.Then a stereo vision system calibration method by using plane calibration plate was introduced, this method not only in laboratory measurement environment,and in factory workshop of CNC measurement environment also can obtain accurate calibration results.In image processing aspects,a method by using the optimal threshold value was studied to extract the target measure workpiece from complex background. In order to solve the high-reflective problem of metal surface, this paper proposed a algorithm which was based on the automatic threshold segmentation to detect high-reflective region,modify the gray value of detected region as same as non high-reflective region. As the workpiece is color, edge, contour and other characteristics of rare objects, and its surface is repetitive metal processing texture, so in the stereo matching process will find more than one corresponding feature local area, it is very easily to produce mismatch. According to the characteristics of workpiece and the measure objectives,three different feature extraction methods and two different feature match algorithms were in-depth studied. The three different feature extraction methods are intensive nature point extracting by using harris operator, the key edge feature points extracting according to the workpiece contour and the surface feature extracting by pasting marks. Oneintensity matching algorithm was based on the gray vavule of small setted area and another point matching algotithm was based on the epipolar constraint.Experiments show that the feature extraction and matching method could obtain more precise matching result. Finally, some other techniques were studied, such as the 3d information acquisition,3d mosaic and dimensional measurement.Based on the above research and analyzation the function of each module in the software system, we designed a user software interface in the development environment of Visual Basic 6.0 with Halcon image processing library. Selected three workpiece, sizes from small to large, structural parts from simple to complex, to experiment in the numerical control machine. Different methods to obtain 3d information was listed according to the shape characteristics and the measure goal requirements.Experiments show that this vision measuring system could accurately obtain depth information and the 3d coordinates of feature points,achieve the 3d dimension of large-scale workpiece on-machine.