The Measurement Method Of The Sheet Metal Forming Based On Binocular Vision

The visual measurement using computer vision technology has completed the parts on-line non-contact detection. Binocular vision measurement technology by acquiring and processing the two images determined the geometry information of the target objects detected. It could complete the three-dimensional measurement of the targetobjects with some advantages of non-contact, integration and stable measurement and it had broad application in the field of industrial measurement.For the measurement of sheet metal forming, this paper improved a method forsheet metal forming based on binocular vision measurement, using the image correlation method to match the corresponding image, to complete the actual sheet metalforming measurement.First of all, according to Zhang’s camera model and the space triangulation measurement principle, the binocular vision measurement model was established. To improving the calibration accuracy of the measurement model of binocular vision, theinitial of the fundamental matrix value was estimated by using the linear method. Then according to the property of the fundamental matrix parameters, least squares optimization method was used to improve the accuracy of the fundamental matrix solution. Consider image noise, there was two target optimization functions for solvingthe fundamental matrix and the GRAD optimization functions got more stable result. By dealing with different types of calibration board image, two-plane calibrationwas determined the more precision and stability for estimate the fundamental, because the point to the corresponding epipolar distance could be more in line with the relationship of epipolar geometry. The focal point of the paper improved the methodfor solving the structure and motion parameters. The essential matrix which contains the structure and motion parameters was respected the fundamental matrix and thecamera internal parameters. Referenced to the nature of initial value matrix paramet ers, nonlinear least squares optimization was used for solving the essential matrix tomeet the geometric constraints. After optimization for the essential matrix, then decomposition of essential matrix to solve the structure and motion parameters improved the solution accuracy of the structure of the motion parameters. In addition, another method for solving the structure and motion parameters was assembling camera parameters. Compared the solution accuracy of two methods, mean errors and standard deviation of three-dimensional reconstruction size was smaller for the decomposition of essential matrix method.Binocular vision measurement was based on the parallax principle to restore thescene of three-dimensional object. The corresponding point on the matching two images was necessary. Image correlation matching method was suitable for smoothly and richly detailed image and matched more precision. In contrast to some image correlation method, the intensity gradient iterative algorithm was selected to calculate the parallax displacement of the corresponding image. The numerical simulation images and real speckle images experiment showed that the intensity gradient iterative algorithm for parallax displacement measurement accuracy could reach sub-pixel level.Finally, there were two experiments to prove the accuracy of binocular vision-based measurement of sheet metal forming. Static translation experiments for the depth displacement measurement of reconstruction completed to verify the accuracy. Sheet metal forming experiment completed to verify the proposed method to measure the feasibility of the geometric morphology of the surface deformation in the formingprocess, and then analyzed the factors affecting the measurement accuracy of binocular vision.This research has a role in promoting part visual inspection and developing mold optimized design.