Based On Machine Vision Automotive Heat Exchanger Sizing And Visual Inspection Systems

Computer vision plays an irreplaceable role in automatic product inspection due to its characteristics of high speed, high precision and non-contaction. Abided by basic principles of machine vision and through referring large amounts of literatures, this thesis builds a real test platform and completes a design of a dimensional measurement and defect inspection system which is based on computer vision.The main contents of this thesis include:(1) A hardware design of the heat exchanger has been completed by selecting the hardware of the inspection system, including light source, cameras, lenses, Gigabit Ethernet and industrial control computers. A test platform is also built to acquire images of the heat exchangers.(2) To meet the needs of the actual measurement, some algorithms of the line and circle detection of heat exchangers are studied and proper algorithms of the dimensional measurement are determined, including edge detection, Hough transform detection of lines and circles, the circle fitting algorithm based on calibration points, the detection of small holes, the sub-pixel line detection method (improved Lagrange interpolation method and Gaussian curve fitting method) and so on. Also, standard computer-generated images are used to test accuracy of sub-pixel line detection methods while circular work-piece images are used to test accuracy of the circle fitting algorithm based on calibration points. At last, some lines, cylindrical holes and small holes of the heat exchangers are measured by relevant algorithms.(3) Three surface texture defects inspection methods for heat exchangers have been studied, including the Gray Level Co-occurrence Matrix (GLCM) method, the direct threshold algorithm and the threshold algorithm based on Fourier transform filtering. The GLCM method first calculates the eigenvalues of the GLCM, then it respectively uses BP neural network classifier and rule-based classifier to classify the defects. The threshold method first selects the proper threshold for binarization, then respectively uses different sizes of the structural element to do the morphological open operations. At last it divides the defects which are subjected to area threshold value. Finally the results of these three algorithms are analyzed and compared.(4) A set of dimensional measurement and defect inspection software based on computer vision is developed. In addition, the relevant software testing process is introduced. On the VS2010development platform, a software which uses C++programming language and open source computer vision library OpenCV is written to measure16sizes of the heat exchangers and detect the defects.