With the development of automation industry, machine vision technology has been widely used. For example, in the automatic manufacturing industry, with machine vision measurement, detection of various size parameters of the workpiece, such as length measurement, circle measurement, Angle measurement, arc measurement, area measurement, etc., not only can obtain the size parameters of online products, but also can make online real-time determination and sorting of products, the application is very common.

The detection system based on machine vision technology has the advantages of low cost, high precision and easy installation. Coupled with the powerful visual software, easy to adjust, flexible and convenient, and the measurement data can be stored, easy to establish statistical analysis, easy to quickly find problems, solve problems, is the industry detection technology of choice.

But in some machine vision inspection projects, many customers have encountered the problem of detection accuracy. For example: testing the external dimensions, inner and outer diameters of processed parts, requiring accuracy to 10UM. At present, many configurations choose 5 million cameras, which theoretically can achieve 10UM accuracy. But the accuracy of the actual customer test is far from the requirements of 10UM, 30UM is very good. The main reasons for the large error are as follows:

1. Choice of industrial camera: Many customers choose CMOS camera for cost consideration. When photographing the edge of the object, the contrast of the edge contour is poor and the noise is large, which brings large error to the software measurement. A CCD chip camera is usually recommended.

2. Selection of camera lens: Due to the difference in height of many mechanical parts, there is a large depth of field. However, because of the Angle of view, it is difficult to shoot the inner wall of parts and so on in ordinary lens, which leads to the difficulty of finding the most real and accurate edge contour in software processing, and puts forward higher requirements for image processing algorithm.

3. the choice of light source: many customers will choose ordinary back light source. When the visual detection accuracy is not high, the backlight source has the advantage of price. Under the very demanding premise, the principle of the back light source is to use light through the diffuse reflector plate to produce uniform and soft light, at a certain point the light will be generated to any Angle in space. That in the detection of circular, cylindrical objects, will produce diffraction phenomenon at the edge of the object being measured. At the same time, the brightness of the light source has a great influence on the brightness and darkness of the image.