Research On 3D Measurement Technology Of Water And Heat Bending Based On Binocular Vision

Currently,domestic shipbuilding companies mainly rely on experienced workers to complete the processing of ship hull outer plates.The traditional water and fire bending plate detection method has low efficiency,large errors,and cannot meet the requirements of modern shipbuilding industry for automated processing and production.This is also a bottleneck that restricts the production cycle and quality of ships.Automated detection of hull outer plate forming is a pressing problem.New and effective measurement methods are needed for highly curved and large-area curved plates.Binocular vision measurement technology can achieve contactless and efficient measurement of curved plates.Therefore,this thesis designs a three-dimensional measurement system for water and fire bending plates based on binocular vision.The research work of this thesis is as follows:(1)On the basis of Zhang’s calibration method,the problem of high camera accuracy requirement and large reprojection error of left and right images is solved.The calibration experiment only needs 3 non-collinear points and 2 images to obtain the camera’s intrinsic and extrinsic parameters without complex matrix calculation.(2)The gray wolf algorithm is used to perform Otsu multi-threshold segmentation on the experimental images of water and fire bending plates,solving the problem of uneven illumination of the plates under natural light affecting the quality of image preprocessing.The optimal threshold is obtained using the gray wolf intelligent optimization algorithm,and a comparative experiment of Otsu multi-threshold segmentation is conducted.(3)The Steger algorithm is selected for image edge detection.The Steger algorithm requires calculating the Hessian matrix for each point in the entire image and performing five Gaussian convolutions.By utilizing the separability of the Gaussian function,the two-dimensional Gaussian convolution is simplified into a one-dimensional Gaussian row convolution and a one-dimensional Gaussian column convolution.Based on this,a recursive algorithm is applied to perform recursive convolution on the image in the order of rows and columns,thereby obtaining the same result as two-dimensional convolution.(4)For the stereo matching problem caused by the weak texture characteristics of the hull outer plate,SAD operator,SSD operator,and NCC operator are selected to measure the similarity of the left and right images of the curved plate,and the suitable operator for the curved plate is obtained.For the stereo matching problem caused by the large size of the hull outer plate,different window sizes are selected for comparative experiments to determine the suitable window size for the curved plate.(5)In order to meet the requirements of automated measurement of water and fire bending plates,a three-dimensional measurement system for water and fire bending plates based on binocular vision is designed.The system mainly consists of image import,two-dimensional image preprocessing,three-dimensional reconstruction,and reset modules.The binocular cameras capture real-time images of the curved plate,which are imported into the system to obtain the three-dimensional measurement data of the curved plate.The deviation value is fed back to the system to adjust the processing parameters and perform reprocessing.Experiments show that compared to the traditional binocular calibration method,the improved algorithm proposed in this thesis reduces the calibration error to within 0.06.The Otsu multi-threshold segmentation method used in this thesis demonstrates good segmentation results,with a 50% reduction in noise points compared to the original image.It has good performance in terms of problem-solving accuracy and convergence speed.Compared to the original Steger image edge detection,the method proposed in this thesis introduces recursive thinking,resulting in a increase in algorithm running speed,and this improvement is independent of the choice of Gaussian parameters.Selecting the SAD operator and a matching window size of 9x9 significantly improves the quality of the disparity map.By comparing the deviation of rib position point clouds,the forming accuracy of the curved plate can be judged to meet the requirements of shipbuilding.If the forming accuracy of the curved plate does not meet the requirements,the deviation value is fed back to the plate forming system to adjust the processing parameters and perform reprocessing,realizing automated measurement of the curved plate.