Research And Implementation Of Shaft Parts Detection System Based On Machine Vision

With the rapid development of the modernization of China’s manufacturing industry,all kinds of mechanical equipment have higher and higher requirements for the accuracy of their parts.In the manufacturing process of parts,various quality problems and defects will inevitably form,such as dimensional deviation,crack and so on.However,the conventional manual inspection method is not only inefficient,but also has poor detection accuracy,and the contact measurement is easy to scratch the parts,so it is difficult to ensure the consistency of quality control.Therefore,aiming at the actual project requirements of dimension and quality inspection of shaft parts,this paper studies and implements a shaft parts inspection system based on machine vision,so as to improve the inspection accuracy and efficiency in the current actual production operation.The overall architecture of the detection scheme proposed in this paper includes two main steps: one is edge detection and the other is feature measurement.For the edge detection method of shaft parts,a part image edge detection scheme based on the improved HED deep learning network model is proposed.Firstly,the sub-pixel convolution technology is used to improve the up sampling method in the HED network model;Then,after the model training is completed,the non maximum suppression and double threshold processing methods are applied to complete the subsequent processing of the output image edge of HED model;In addition,in order to improve the accuracy of size measurement,the spline interpolation subpixel edge detection technology of improved morphological gradient technology is used to process the edge contour,so as to further obtain a finer edge image and realize the image edge at the sub-pixel resolution level.For the feature measurement method of shaft parts,the improved Hough circle transformation algorithm based on the principle of fixed chord and fixed angle and the sub-pixel corner detection algorithm are used to measure the feature size in the edge image,which further improves the measurement accuracy of the feature parameters of shaft parts.Based on the above detection scheme of shaft parts,the corresponding software and hardware design is given.In terms of hardware system design,according to the special shape and detection requirements of shaft parts,the overall structure of the hardware system is designed,the camera,lens,light source,lighting mode and motion control system are studied and selected,and the camera is calibrated to ensure that the detection system can obtain clear image resources;In terms of software system design,the visual inspection system adopts C/S structure,uses QT development software and SQLite database for development and design,and its main functions include station task setting,target part identification,size parameter detection,historical data query,etc.Based on the construction of the above detection experimental platform,the detection system of a small shaft part is verified.The experimental data show that the detection system in this paper can effectively detect the edge features and carry out the measurement work,and the measurement accuracy can be up to 0.001 mm.It provides a reference for solving the dimensional quality detection problems of relevant shaft parts and has certain application significance.