| Today,all countries are actively seizing the heights of manufacturing.Similarly,China has put forward the ambitious goal of intelligent manufacturing 2025.Since the Revolution of 1911,the textile industry has been a vane of reform and progress,and an important industry to realize national prosperity and national rejuvenation.However,in order to face the impact of low-cost textile in Southeast Asia,Asia,Africa and Latin America,the combination of intelligent manufacturing and textile,reduce labor costs,improve industrial competitiveness,and accelerate economic transformation has become urgent.In traditional shoemaking process,the uppers need to be polished to increase the friction of cold sticking sole uppers.However,manual grinding is a laborintensive industry,and workers will produce and inhale large amounts of dust which is harmful to health.It takes a long time to acquire grinding track line by manual teaching.Based on the above background,this paper studies the automatic extraction of shoe upper grinding track of cold-stained line.After consulting the extraction technologies and solutions of upper grinding track at home and abroad,it is found that most studies use CAD/CAM technology or machine vision technology to obtain grinding or gluing curve,but only the processing track of sole or upper is extracted.In recent years,3d vision and non-rigid point cloud registration technology have been studied extensively.In this paper,a method of trajectory extraction using non-rigid point cloud registration algorithm is proposed.First,build the hardware platform.The hardware is properly selected and packaged as a camera box.The camera box photoelectric door and conveyor belt are used to build a 3d vision platform for sole to calibrate parameters.Steger algorithm was used to extract the center of the light strip,and Angle threshold method was used to extract the inner wall(indirect deletion of the middle support hole and rib plate point cloud)and put it together into a complete sole point cloud.Similarly,the camera box and manipulator were used to build the 3d visual platform of the upper and calibrate parameters,and the point cloud of the upper was obtained.Secondly,Bayesian Coherent point drift algorithm is used to register non-rigid point cloud between sole and uppers.The combination of Gaussian mixture model and EM algorithm can achieve a better fitting point cloud,so GMM model is used for point cloud registration in this paper.BCPD algorithm uses GMM model and Variable Bayesian inference to realize the absolute convergence of model parameters,and the actual fitting effect is good.Then,genetic algorithm is used to improve the normal vector solution and boundary extraction algorithm is used to obtain the grinding trajectory.In this paper,the common basic algorithms of point cloud are reviewed,and Genetic algorithm is used to improve the algorithm of weighted PCA normal vector.Experiments show that the algorithm has good accuracy under the condition of 5% to 25% error.Using normal vector features to solve the boundary points of point cloud and cluster segmentation,the boundary points will be offset to obtain 6-d OF machining points.In order to verify the trajectory accuracy,RANSAC and ICP algorithms were used to transform the manual teaching and polishing trajectory of finished shoes into the upper coordinate system and calculate the two trajectory errors.Experiments show that the method can effectively extract the trajectory curve of shoe upper,and the average error can be controlled within 2mm.Compared with the traditional method that relies on off-line sampling points to extract the trajectory,the acquisition time of trajectory can be greatly reduced.Finally,this paper realizes the architecture design and page design of related software.The software part of shoe upper grinding track acquisition system can be divided into hardware calibration,point cloud acquisition and point cloud processing.Using Open CV,PCL,NET controls to achieve the bottom of the three blocks and interface design. |
PDF Full Text Request