Research On Workpiece Positioning Algorithm Based On Machine Vision

With the continuous development of industrial processing to the industrial 4.0 era and industrial flexible processing,the methods of manual clamping and clamp clamping in traditional industrial processing process affect high-precision machining accuracy and time efficiency,and workpiece clamping accuracy directly affects the workpiece.Precision of finishing.When the workpiece requiring secondary finishing for the spline is subjected to secondary clamping,the conventional clamping method cannot meet the industrial production requirements.Machine vision-based inspection methods are widely used in all walks of life,providing great convenience for people’s lives and improving the level of industrial processing intelligence in the industrial field.Therefore,this paper studies the positioning method of the surface-distributed workpiece based on machine vision,and completes the high-precision positioning of the workpiece.The main research contents are as follows:The research status of workpiece clamping and positioning at home and abroad is analyzed,the system solution of visual positioning system is determined,and the design of optical system and software system is completed.The motion state of the workpiece and the motion of the workpiece feature point during the motion are analyzed.Based on the traditional image processing and localization algorithm,the positioning measurement accuracy with the error of±0.01~°is realized in about 25 seconds.Industrial field tests were conducted to meet the accuracy and time efficiency requirements of industrial processing.A positioning algorithm based on image depth learning is designed.A 50-layer ResNet network is used as the image feature extractor,and the extracted features are predicted by linear regression method.This method only needs to collect 2 images in the whole process,which can greatly improve the time efficiency of workpiece positioning,and complete the positioning progress of±0.03~°in 3 seconds.Although there is still a gap between industrial application requirements,there are still many spaces that can be optimized in the current network model.Finally,the advantages and disadvantages of the two positioning algorithms are compared,and different application scenarios are given.