Research Of Target’s Surface Defects Detection System Based On Machine Vision

In the industrial production line, the detection of products’ surface defects is an important part. With the rapid development of the electronics industry, components production is increasing. Traditional manual detection is low-efficient and costs a lot, unable to meet the growing production capacity of producers. Machine vision inspection technology has become a popular and been applied in various fields. To reduce costs and improve production efficiency, automated production line needs to develop suitable machine vision technology.The main study objects of this subject are the surface defects of CBB capacitor(bubbles and threadbare) and cone loudspeaker(watermarks and spots). Based on the basic structure of machine vision technology and the needs of detection system, the targets’ surface defects detection system mainly includes the transfer module, image acquisition module, image processing module, and control execution module. The main study points are image acquisition module, image processing module and the design of detection software system(namely the application software of camera). In this subject, the main completed work is as follows:(1) Design and implement the image acquisition system, mainly including the diffused LED light source, the programmable CCD industrial camera, and the signal trigger system. Among them, the main focus is the way of lighting: using annular and coaxial light to detect the cone loudspeaker; using foreground light to detect the capacitance, what’s more, polarizes the lens and lighting source to decrease the reflection. The industrial camera needs to work in the external trigger mode(using the trigger signal to trigger the camera), and set the appropriate and fixed parameters in a fixed environment to ensure image quality.(2) The image processing algorithms of surface defects detection designed include preprocessing and feature analysis. In the preprocessing stage, use threshold segmentation, vertical and horizontal projection to extract foreground(cone images needn’t to extract foreground) and target segmentation. In the feature analysis phase, corresponding algorithms is designed to different defects. For the detection of capacitance, the bubble blemishes are discriminated by comparing the threshold with total shadow pixels in the R channel; for the threadbare defects, use regional consistency detection algorithm or the gradient-based detection algorithm(the latter with higher efficiency is been applied). For cones’ watermark defects detected by adaptive regional consistency detection algorithm; for spot defects, use the algorithm based on wavelet transform, or the gradient-based detection algorithm(the latter with higher efficiency is been applied). Through experimental simulation, the algorithms applied can complete the detection of target surface defects, with high efficiency and meet the needs of the detection system.(3) The detection system software platform based on OpenCV and MFC is designed in VS2010 environment. Transplanting the image processing algorithms designed in MATLAB into C++ programs based on the OpenCV library, improves the detection efficiency. The software is designed and developed based on the overall workflow, acquisition and control processes of the industrial camera. It can control camera’s normal work, set up appropriate and fixed parameters of the camera on the basis of different environments, detect the defects of target images, and display the original and processing result in the interface.The target surface defect detection system designed in this subject, through experimental test analysis, has reached the requirements of the system. With simple user interface and comprehensive function, it can be effectively auto-detect the surface defects of CBB capacitor and cone loudspeaker, and has laid a good foundation for the final procedure of excluding the substandard products. It is hopeful that this system will have strong market competitiveness in the machine vision product areas.