Study On The Method Of Stereo Vision Measurement Of Large Size Geometric Quantities

The measurement of large dimension geometry is the key and difficult point of product quality inspection in the fields of aviation, spaceflight, shipbuilding, wind power, radar and other advanced manufacturing fields. Aiming at the common measurement methods for measurement of low efficiency, high cost, complicated operation and other issues, considering the advantages of machine vision measurement, this paper makes a research on large dimension geometry measurement, which is based on binocular stereo vision and transfer station.Firstly according to the camera perspective model and the characteristics of large dimension measure in practical application, this paper proposes a binocular camera calibration method based on Zhang Zhengyou plane calibration method and active vision calibration method. Using this method, the internal parameters and the structural parameters of the cameras are accurately calibrated, and the calibration results are verified by experiments. At the same time, in order to overcome the disadvantage of poor robustness of visual measurement, the paper pastes marking point on the surface of the workpiece to help measurement. The circular degree threshold method is used to recognize circular target, center coordinates are getted by contour fitting method, a comparative analysis of various methods was carried out by experiments.Aiming at the shortcoming of the low match correct rate of traditional dense similar target, according to the cross ratio invariance, a new method which based on the cross ratio invariance properties and polar line constraints is proposed. The experiment verifies the advantages of the new method. Next, this paper compares the differences on solving the coordinate transformation matrix of anti-interference ability and solution precision between 3points method, SVD method, unit quaternion method through the analysis of the simulation experiment, then determine use distance constraint method to obtain spatial matching point pairs and SVD method to solve the transformation matrix to unify local point cloud which are in coordinate system.Finally, this paper through large workpiece surface morphology experiment and large size length measurement experiment verifies the feasibility and accuracy of the mehiods proposed in this article. The measurement error of the system is also analyzed.