Research On Methods Of Generating Automatic Spraying Trajectory Of Shoe Sole Based On Stereo Vision

In recent years,with the continuous upgrading and innovation of the manufacturing industry,the application of robots is not only in the area of heavy industry,but also gradually replace manual work in light industry.Although its efficiency in some repetitive work is obviously higher than that of manual labor,it has great difficulties in meeting the personalized needs and flexible processing needs of the footwear Industries.At present,part of the automation system for the production of shoes is put into use,but in the sole gluing still requires manual participation and in the automation of the sole inner edge gluing still lack of stable and efficient program.The combination of non-contact measurement technology in machine vision and automatic robot machining can effectively solve the customized machining problem,especially in the production process with high accuracy and high efficiency requirements,such as accurate 3D measurement can plan the path for robot machining or generate control vertices.In this thesis,the production method of sole spraying trajectory required in the processing of flexible manufacturing of sole was studied.This thesis presents a method of automatic spraying trajectory generation of soles based on 3D laser vision,and studies the application of traditional image processing methods and deep learning methods in 3D vision.The main work is as follows:The 3D database of soles was established.According to the requirements of precision and efficiency of the set application scene,a 3d laser scanner and a servo translation platform are used to build a working platform for collecting 3D data.The platform mainly carries out visual measurement for irregular sole,so as to obtain 3D data of the measured sole.To process the depth data of sole,use the block maximum searching method to extract discrete points of the edge of the sole and the least square method is applied to the data optimization of toe cap and heel.And B-Spline control points were calculated.Finally,curve fitting was carried out on the edge of the highest point of the sole to generate a 3D trajectory curve and the effect of curve generation is compared.The application of point cloud information processing and deep learning to the extraction of spraying trajectory of sole was studied.The 3D semantic segmentation network is trained by using the 3D coordinates of point cloud and the differential information of point cloud as input data.The effects of different training methods were compared,and a set of annular strip points of the sole was extracted from the test data.And the median point in the direction of spatial height was used as the input point of 3D curve fitting.Finally,the 3D curve fitting method is used to fit the trajectory curve.