Research On Key Techniques Of Card Printing Qua Ity Detection

This thesis presented a method based on computer vision and image processing technology aiming at the problem of card printing quality detection.Using industrial camera take photos of detected cards.The camera is calibrated by Zhang's calibration method.The distortion of the captured image is corrected by using the virtual image plane model.The principle is to construct a virtual image plane in the process of camera imaging,and restore the card image without distortion by using the projection points on the virtual image plane.When taking the card to be detected,the card is tilted at a certain angle by human influence,so it is necessary to correct the inclination of the detected image.We propose a method based on DBSCAN clustering for detecting the tilt angle of text in this thesis.The text pixels in the card image are regarded as the data sets to be clustered.With the DBSCAN based clustering method,pixels are classified according to word block.It is considered that pixels in each class can be fitted to a straight line parallel to the skew direction of text.The centers of gravity in every class in column direction are calculated.The line can be calculated using the centers through Least Squares Method.The angle of the line is considered the skew angle of the text.Descending sampling is used to improve the speed of algorithm.The standard template image is generated from the Excel data file.Unifying the text size of the source image and the template image According to translation between the tilt corrected source image and the template image after the tilt correction,image matching is done through template matching to find the best matching position.The difference image between template and tilt correction image is obtained by image difference method.The differential error is eliminated by open operation.The defects of the detected image are found by setting the threshold of the detected density,according to the characteristics that the density of the black pixels in the differential image is large at defects and small at non-defects.Morphological is used to mark defects in skew corrected images.The experimental results showed that the card defects can be detected fast and accurately by the proposed method for detecting card printing quality.It got a higher accuracy.The quality of the printed card is detected automatically by using computer vision technology in this thesis.It got a lower cost and higher detection efficiency.It can be widely applied to the industrial detection of card printing quality detection.