Large-scale Binocular Stereo Vision Measurement Technology

With the development of the economy and the level of mechanical manufacture,the measurement of the geometrical dimension of the large-scale workpieces is needed in the fields of aerospace,port machinery,metallurgy and so on.It is difficult for the conventional measurement methods to measure the sizes of the large dimensional workpieces due to the large size.In this thesis,analyzing the status of large-size workpieces measurement methodology at home and abroad,the large-scale binocular stereoscopic vision measurement system was developed.The measurement system is composed of two CCD cameras,a projector,an acquisition card,a computer and the other components.According to the binocular stereoscopic principle,the three-dimensional spatial point cloud is obtained by calculating the parallax formed between the left and right views of the spatial points.Therefore,it is necessary to match the two corresponding points on the left and right views.So the operation principle of the system is determined: projecting binary coded black and white stripes through the projector to the measured workpiece surfaces in the order of certain time sequence,capturing stripes images by the two cameras,transmitting the images to the computer for being processed,finding the points with the same code value in the left and right camera images namely the matching points,and obtaining the three-dimensional coordinates of the corresponding spatial points by combining the intrinsic parameters of the cameras acquired during calibrating process with the relative position relationship of the cameras.Finally the three-dimension point cloud of the measured workpieces is obtained using a large number of matching points and the geometrical parameters of the measured target are obtained.The main research contents of this thesis are as follows:To calibrate the cameras a self-calibration algorithm is proposed according to the calibration template based on two coplanar intersecting circles.By solving the coordinates of the ring points,the internal parameters of the two cameras and the relationship between the two cameras are linearly solved,that is rotating matrix and translation matrix.Compared with the traditional Zhang Zhengyou calibration algorithm,the developed algorithm in the thesis do not have to match the two-dimensional image points with the three-dimensional space points,it can automatically calibrate without manual intervention.In order to solve the problem of stereo matching,this thesis found the matching points in the left and right camera image by projecting binary coded black and white stripes to the measured workpiece surfaces and decoding the coded images.When decoding the threshold segmentation is implemented,the feature points in the images are extracted,the point matching is achieved and the code values are obtained in the coded images,the points with the same code value in the left and right camera images are matching points.After calibrating the cameras,image preprocessing,stereo matching,employing this system the three-dimensional point cloud is obtained to measure the workpiece.The experiment results verified that the system meets the needs of large-scale and non-contact measurement.