| In recent years,with the rapid development of shoe industry,more and more enterprises manufacture sports shoes.The traditional shoemaking craft relies on the manpower for the most part,to the worker’s technical request is very high,the production sneaker quality is not stable.In the process of shoe making,the sole gluing occupies large workers,and the gluing is harmful to workers’ health.Therefore,it is of great significance to introduce machine vision and robot technology into the automatic production of sports shoes.First of all,combining with the technical requirements of cold-stick sneakers,the robot automation scheme of glue-spraying technology for sole is designed in detail.A 3d information reconstruction machine vision system based on a face array camera and a linear structured light sensor was built.Based on the principle of 3d reconstruction of linear structured light and the coplanar target method,the 3d coordinates of the center of the light bar in the camera coordinate system are extracted by using the method of intersection of linear planes,and the plane of the light is fitted by using the least square method.The calibration experiment of shoe sole visual reconstruction system is designed to verify that the optical plane calibration method proposed in this paper can reduce the calculation amount and still have high accuracy.Through point cloud reconstruction of standard block objects,the experimental results show that the accuracy of the shoe sole visual reconstruction system designed in this paper meets the actual requirements of the factory,and the accuracy in Z direction can reach 0.06 mm.Secondly,the characteristics of the light strip in the surface of the moving sole are analyzed.In order to solve the problems of the spot formed on the rough surface and the sharp change of the lateral width of the inner kernel of the sole,an algorithm for extracting the center of the smooth strip on the surface of the moving sole was proposed.On the basis of Otsu,the difference threshold between the gray mean of pixels in the template and the gray levels of other pixels is introduced to effectively segment the spot area of the light bar.Canny operator was used to extract the edge and width of the light bar,and the size of gaussian filter kernel in Steger algorithm was changed according to the width of the light bar,which solved the problem of missing the center extraction of the light bar caused by the change of light bar width.The effectiveness of the algorithm is verified by comparison experiments,which lays a foundation for the reconstruction of sole point cloud.After that,the analysis of the data structure of the point cloud and normal and curvature method to solve the point cloud,the sole point cloud background removal,discrete points filtering,then the sole point cloud plane projection,to extract the out-ofplane outline,and connecting with the out-of-plane extracted contour focus inward cut in outline,combining with the inner and outer contour extraction Ro I(Region of Interest)point cloud,Ro I point cloud amount is only 10% of the original point cloud.When calculating the curvature of point clouds,the optimized random sampling consistent algorithm(RANSAC)is used to solve the curvature.The point cloud normal direction was introduced to reduce the number of parameter estimation of the quadric surface model.Meanwhile,an adaptive sampling method was designed to reduce the iteration of the algorithm.Then,according to the curvature threshold,the sole point cloud boundary was determined.In order to obtain the preposition of sole spray,the pipe surface is constructed by using the die line,the point cloud area of the target range is searched by the radius,and the final preposition of spray is obtained by using the included Angle relation between vectors.Moreover,experiments are designed to verify that the algorithm presented in this paper can solve curvature faster than the traditional RANSAC algorithm,and can still better extract the edge of sole under 25% Gaussian noise,with good noise resistance.Finally,the software design of the motion sole information extraction system based on 3D vision is completed,and the function of three-dimensional reconstruction of the motion sole and the pre-point extraction based on the spray glue point cloud is realized.Combined with the hand-eye calibration of the robot,the experiment was carried out on the shoe sole spray preposition extracted in this paper,and the accuracy of the shoe sole spray preposition extracted in this paper was verified by comparing the distance between the point cloud spray preposition and the distance between the real spray preposition. |
