The Research On Machine Vision Detection System Of Industrial Components' Shape And Dimension

In industrial production, the traditional detection technique requires a large number of workers. It not only affects the production efficiency, but also brings unreliable factor. Visual detection is widely used because of detecting safety, reliability and high automation, which overcomes the shortcomings of the traditional detection technique. It has become the hotspot of the detection research .In order to resolve the traditional detection problems of industrial component, the machine vision detection system of industrial component's shape and dimension is designed in this paper, which is based on machine vision technology, Sobel edge detection algorithm is used in this system.Compared with the traditional machine vision system elements, hardware selection required by the system to achieve the goals and tasks is completed in this paper, such as a light source, lens, CCD cameras, frame grabbers, and so on. The software flow of the machine vision detection is designed. Then, the part of image acquisition and processing is analysised. A proportionate calibration method of linear camera is presented in this thesis, which is simple and practical. The standard compenent as calibration object and proportionate algorithm to realize the camera calibration is used in the calibration method. The cumbersome process of standard grid calibration is avoided in this method. In the condition of the background measured by the imaging contrast and good lighting, sobel edge detection operator and the sub-pixel location algorithm of the Gaussian curve fitting is selected to fit the system. It improves the edge detection accuracy.An industrial compenent is describled as an example, the process of the compenent's machine vision detection is introduced in this paper. The measurement errors are also analyzed. The experimental results show that the program feasible, it can achieve higher accuracy and meet the industrial requirements.