Research on Online Visual Inspection System for Multi-Station Silicone Ring Defects
Silicone ring detection has the characteristics of multiple types,easy deformation,and high complexity.By proposing a vision-based electromechanical detection system that combines multiple stations,multiple angles,and different lighting meth-ods to solve the detection problems of such products.The image processing of the entire detection system is improved based on the OpenCV library in accordance with actual project requirements,using filter operators,threshold segmentation and morphologi-cal methods.The VGG16convolutional neural network training model is used to control the accuracy of the three defect classifica-tions and traditional visual inspection methods under small sample conditions.In a small sample,the detection accuracy of tradi-tional machine vision is>98%,which is better than deep learning methods.Complete the UIinterface design through Qt.The soft-ware system is written in a multi-threaded design model,and a new method of multi-station cyclic elimination is proposed,which has the functions of online detection,counting and elimination.The actual experiment and test on a specific silicone ring product test verified the reliability and practicability of the system.
万方数据
维普
