Three-dimensional Reconstruction Of Grid Structure And Bending Rod Discrimination Based On Computer Vision

Computer vision is a method to achieve the recognition,understanding and reconstruction of the three-dimensional scenes by using cameras,computers and other equipments.With the development of hardware devices and deep learning theory,computer vision has become the focus of many studies in recent years and been gradually applied to various professional fields.Based on the binocular stereo vision technology and image recognition technology in the field of computer vision,this paper proposes a three-dimensional reconstruction method for the spherical joints and rod members in grid structures,and uses the image processing technology to initially realize the judgment of bending rods under simple backgrounds.The binocular stereo vision technology is applied to capture the measured structure.Through the relative positional relationship of the binocular camera and the position of the ball node in the image,the center coordinates of spherical joints can be calculated,which further realizes the three-dimensional coordinate reconstruction of the spherical joints.The view of camera is limited by its angle and the distance from the camera to the object being measured.Whereas,the grid structure is usually in large-span,which means the requirement of multiple times of binocular vision measurement.Therefore,the three-dimensional coordinate transformation model under arbitrary rotation angle shall be applied to unify the coordinate system of multiple measurements,and further achieve the measurement of large-span grid structures.Then,the deep learning network is introduced to perform the identification of spherical joints and rod members in the reconstruction process.For the automatic identification of the rod,the Mask R-CNN is adopted to obtain the position information of the rod in the image and its mask information too.The mask information of the rod and the image coordinates of the nodes are used to judge the connection relationships between every two nodes.The clip image obtained by the Mask R-CNN is taken as the research object to compare a variety of image enhancement algorithms and edge detection algorithms The contrast limited adaptive histogram equalization is used to enhance the contrast ratio of the image.The edge of the rod is extracted by the Canny edge detector,and the quadratic curve fitting is used to approximate the axis of the rod,and the maximum curvature is calculated to realize the judgment of the curved rod.The three-dimensional reconstruction of the grid model was realized by the test of an orthogonal square space truss,and the feasibility of the three-dimensional reconstruction method of the grid structure by computer vision technology was verified.