| The quality inspection of brake pipeline joints and ends is an important link in the production process of brake pipelines.With the improvement of the production automation level of enterprises,automatic production of brake pipelines has been realized.However,manual identification and sorting are still used for the detection of the dimensional parameters of brake pipeline joints and ends and the detection of surface defects of the ends.Due to the small size and high precision of brake pipeline products,manual detection has low efficiency and poor accuracy,and it is difficult to match with automatic production.Therefore,it is of great theoretical and practical significance to study the on-line automatic detection technology of brake pipeline joints and ends.In this thesis,combined with the existing brake pipeline production line,a set of on-line detection system for brake pipeline joints and ends based on machine vision technology is designed,and the key technologies involved are deeply studied.The main research contents are as follows:According to the inspection requirements and production environment of brake pipeline joints and ends,a visual online inspection platform for brake pipeline joints and ends based on machine vision technology is designed and built.The camera,lens,light source,conveying device and other hardware equipment in the inspection platform were selected,and the lighting mode of the joint and the end was configured.The key technologies in the detection of the dimension parameters of the brake pipeline joints are studied.An image tilt correction method is given to correct the tilted joint image to avoid the influence of the joint tilt on the detection accuracy;judging the installation direction of the joint based on the Graham scanning algorithm and the boundary rotation method;based on sub-pixel edge extraction technology,cubic Bspline curve fitting technology and least squares fitting straight line technology,the major diameter,middle diameter,minor diameter,thread pitch,tooth half angle and total length of the bolt joint are calculated.In order to ensure the measurement accuracy and avoid the loss of detection accuracy caused by lens distortion,the calibration method and lens distortion mechanism of the monocular camera are studied.Finally,experiments are designed to verify the versatility,reliability,detection accuracy and detection speed of the joint dimension parameter detection algorithm given in this paper.The key technologies in the detection of the dimension parameters of the the brake pipeline ends are studied.Based on Geometrical Product Specification and Verification(GPS)operation technology,the traditional end dimension parameter detection process is standardized,image processing methods such as grayscale transformation,image sharpening,threshold segmentation and morphological processing are used to complete the separation,edge extraction and filtering operations of the end image.Then,the extracted end outer circle and inner hole edge are fitted with the minimum circumscribed and maximum inscribed fitting criteria respectively,and the size parameters to be obtained are obtained by calculation.Finally,the versatility,detection accuracy and detection speed of the detection algorithm for the dimension parameters of the end in this paper are experimentally verified.The key technologies in the detection of typical surface defects of brake pipeline ends are studied,and two detection methods of end surface defects based on blob analysis and convolutional neural network are proposed.The blob analysis method mainly uses threshold segmentation,morphological processing,color space conversion and other means,which can detect defects such as unpierced head,flash,pit,pitting,scratch,rust,or copper leakage on the surface of the end,the experimental results show that the average detection accuracy of the blob analysis method for typical surface defects of six different types of ends reaches 92.8%.the given convolutional neural network model is based on Res Net-50,optimized and improved in combination with the characteristics of the end image,and integrated into the CBAM module,which is used to enhance useful feature information and suppress redundant information.The training and testing results on the self-built end data set show that the prediction accuracy of the model for typical surface defects of type I and type II ends reaches 98.1%and 97.8%,respectively,compared with other models and blob analysis methods,the accuracy rate is greatly improved.Based on software development platforms such as C#,Halcon and Python,the development of the software system for the online detection of brake pipeline joints and ends is completed,and the test is carried out in the actual production environment of the enterprise.The results show that the on-line detection system for brake pipeline joints and ends developed in this paper can meet the production needs of enterprises. |
