The Research On The On-line Inspection System Of Industrial Product

Qualified product is very important to the product quality.Manual testing method of products has low efficiency and high labor cost.With the wide application of machine vision detection in industrial products,it greatly improves the production efficiency and reduces the labor cost.In this paper,a kind of industrial product called Remote Radio Unit produced by a factory still need to manually detect whether the screws on this product was missing or not.It brings a low efficiency,fatigue workers and missing detection.Aiming at these problems,an on-line inspection system for industrial products was set up on the production line to detect whether the screw on the product was missing or not and OCR character recognition.The main contents,results and innovations of this paper are showed as follows.(1)An industrial product on-line detection system was designed and built.The hardware system is mainly composed of a computers,a SLR cameras,a acrylic panels,a strip LED lights,a rotating platforms and a cabinets and other components.It realized automatic photographing of the different detection surfaces of the industrial products.And it provided a hardware platform for automatic detection of missing screws and OCR character recognition.(2)The algorithm of detecting the missing screw on the industrial product was studied.Three algorithms based on gray histogram,LBP and phash were used to detect the missing of screws.Comparing the results of three methods,it was found that the detection algorithm based on phash was better.(3)The OCR character recognition algorithm of the industrial product was studied.Firstly,the preprocessing operations such as gray-scale,binarization,denoising,and correction were performed on the acquired image.Then we used an custom character library training method which was provided by the Tesseract-OCR to train the industrial product character library.Finally,we used the Tesseract-OCR engine to identify the characters.The recognition results showed that self-training character recognition accuracy rate of 96.2%.(4)On Visual Studio 2010 platform,the online detection system software ofindustrial products was developed by using C # language,which mainly includes two parts: product testing parameter setting and product online testing.First of all,the industrial products required to set the detection parameters,and then automatically taked pictures of the product of each test area,and judged whether the screws were missing and recognised OCR character for each detection surface detection area,the test results would be saved in the document for reference.In this paper,the industrial products defect on-line inspection system in the actual operation of the plant results showed that the system is running stable,high precision detection.