Research On Key Techniques Of 3D Measurement Of Large Scale And Low Texture Surface

Large scale,low texture and smooth surface are widely used in all kinds of large high-speed vehicles.In this paper,the method of binocular vision and projectors is selected,and the method of texture image segmentation,dense binocular matching and surface 3D reconstruction is combined to achieve the 3D measurement of large scale smooth surface with low texture.Specific research contents and research results are as follows:(1)Establishment of binocular vision model and parameter calibration.This paper establishes the mathematical model of the binocular camera system,including the camera imaging model and the stereo model of the binocular camera.A camera parameter calibration method based on the absolute quadratic curve is proposed,and an intersecting coplanar ring calibration board is designed.Through the proposed calibration algorithm,the internal parameters of the camera and the external parameters between the binocular camera are calibrated.Based on the calibration results,epipolar correction was performed on the binocular images.Through repeated calibration experiments,the mean square error of multiple calibration results is obtained.The experimental results show that the calibration results of the proposed algorithm are stable.(2)Texture image segmentation.This paper proposes a texture image segmentation method for segmenting binocular images to improve the subsequent binocular matching speed and accuracy.In order to improve the recognition ability of different texture information,a new local histogram-based Texture descriptors model the distribution of textures in an image.At the same time,the texture descriptor can also obtain texture scale information.A boundary detector that calculates local multi-scale probability is designed to extract texture boundaries and use the texture boundaries for region merging.A large number of qualitative and quantitative experimental results show that the method in this paper is superior to other methods in color and texture segmentation,especially for images with rich color texture and gradual brightness.(3)Dense binocular matching.Based on the binocular image matching cost calculation method based on Census transformation,a new matching cost calculation method is proposed in this paper,which increases its perception range by modifying the shape of the Census template,and increases the matching cost by performing Gaussian weighting on the result of the Census transformation.The accuracy of the calculation.In order to achieve high-precision matching cost aggregation,this paper proposes a new image edge retention filtering method based on texture information and applies it to the matching cost aggregation operation.The proposed texture filtering method is a method based on bilateral filtering.By adding an adaptive scale kernel function and adding a new structure detection method,it can filter out the texture patterns while retaining accurate image edge information.On the basis of the preliminary left and right camera disparity maps obtained by cost aggregation,the binocular disparity map is subjected to matching consistency detection to obtain a high-precision disparity map.After experimental demonstration,the method in this paper can obtain dense disparity maps with high accuracy.(4)Surface 3D reconstruction.In order to obtain the final three-dimensional model of the target,the disparity map and the parameters of the camera system are used to calculate the dense point cloud on the target surface and the point cloud is stitched.The method calculates the vector field of the point cloud,and uses the method based on Poisson's equation to realize the 3D surface reconstruction of the measured object.(5)Zero-construction surface measurement experiment for large curved surfaces.This paper designs a large-scale three-dimensional measurement system,including the operation flow and hardware system of three-dimensional information collection.The projector is used to project the image of the target,and the binocular camera is used to capture the target to obtain the 3D data model of the target.The camera and the projector are alternately moved to expand the scanning range.A binocular camera simulation program has been developed to It is used to conveniently,intuitively and accurately select various parameters of the binocular imaging system,so as to provide a basis for the hardware selection of the camera system.The paper also designed a projection pattern based on the large size and low texture features of the measured object,and selected a projector equipped with appropriate parameters.In this paper,a white smooth spherical surface is used as a target to conduct a three-dimensional measurement experiment using a built three-dimensional measurement system.The experimental results show that the three-dimensional measurement method and system proposed in this paper can realize the three-dimensional measurement of large-size low-texture smooth surfaces,and the system measurement error is lower than 1mm.