Research On The 2-D Size Measurement System Of Mechanical Parts Based On Machine Vision

Chinese has the title of "The World Manufacturing Factory" early, and the mechanical processing industry is an important basic industry in the manufacturing industry. Detecting is essential to ensure the quality of mechanical parts, and the detection of mechanical parts are often involve the 2-D contour size detection, but traditional detection tools such as calipers and gauge more and more difficult to meet the demand, at present, the size measurement of mechanical parts based on machine vision has become the mainstream of the development direction, this method has advantages of non-contact, high accuracy, fast speed and so on. Compare to the traditional detection method has obvious advantages. But in the image size measurement field, the development level between domestic and foreign has a big gap. Most of the image measurement system on the market are foreign company’s products, and the price is expensive, especially for the small company is difficult to buy, so there is an urgent need for the domestic universities research institutions and enterprises to develop relevant product with independent intellectual property rights and the low costs in order to catch up with foreign advanced level. It’s will promote the development of machine vision technology greatly and reduce the burden of payment for enterprises to purchase equipment.The research content of this paper mainly involves the following aspects: pixel and sub-pixel edge extraction algorithm; automatic location and recognition algorithm of line, circle and arc; realize automatic detection of batch mechanical parts. This paper use digital camera, telecentric lens and parallel back light source to build the hardware system, and software system is developed based on VC++ and Open CV. Firstly, based on the analysis of several classical pixel edge extraction method, because of based on morphological edge detection algorithm with speed, can obtain continuous and single pixel edge, this paper use morphological edge detection algorithm to obtain a pixel level edge, and then based on the fitting of orthogonal polynomials sub-pixel edge was detected. In view of the traditional methods can not well solve the problem about the identification and location of geometric primitives realize the location and recognition of geometric primitives accurately using the DXF file of parts and combined with the distance transformation image of edge contour.Finally in order to realize automatic detection of large quantities parts, the process of detection will be saved in the template file, then can realize automatic detection by manual selection or automatic search detection template. In the last chapter, analysis this measurement system, including bias, linear analysis, repeatability analysis, all performance indicators have reached the expected requirements. The results show that the mechanical part size detection system based on machine vision of this project has a good prospect of engineering application.