Research On Online Defect Detection Algorithm For Package Printing Matter

With the development of printer detection technology, non-contact surface inspection system based on machine vision has become the main form of printer matter detection. It takes the template image as the standard, through the template matching method to determine whether the real time products are defective.In the modeling process, because the real-time acquisition of the product image contains the belt background, therefore, when generating the model image printed portion of the effective, we need to extract the foreground region and extract the ROI in the effective area of the extracted foreground. The traditional way of ROI extraction is all by-hand. This artificial division method is increasingly becoming a bottle neck of system performance with the detection system hardware upgrades and software system optimization.In the process of printing detection, for a class of detection accuracy requirements of the relatively high leakage defect detection, less considered to the algorithm. The majority is based on traditional detection methods, causes such defects are often missed or false.To the two kinds of problems of automatic modeling and color leak detection,This essay is based on digital image processing algorithms and uses the algorithms to printer surface inspection system of Focu Sight. This paper contains the following aspects:(1) The overall design structure of the hardware part of the system puts forward selection and configuration method based on the full analysis of the hardware structure parameters, the advantages and disadvantages of the machine vision. In the part of the software, the template matching process is summarized.(2) Combined with image processing algorithms to the automatic modeling process, starting from the actual detection of color printing, the color space conversion method, image segmentation, image feature extraction. Thereby effectively extract the foreground area of the printed matter for auto-modeling.(3) The accurate extraction of the Region of Interest(ROI) detection is the key to the success of product availability modeling(PAM). Automatic extraction and location calculation method based on multi features has been put forward.Considering the fact that the distribution of the parameters of the automatic extraction and location cannot be determined accurately, a method of determining the center point coordinate error method is proposed. The results show that the time is less than100 ms, the success rate is above 99.78%, which meets the needs of the actual automatic modeling.(4) Before defect detection,based on the type and characteristics of the image noise, the method of mean filter, median filter, adaptive median filter is analyzed. The adaptive median filter method can effectively remove the image noise. In the process of defect detection, against the color printed packaging images have different characteristics on defect contrast, color distortion, shape defects, image size,background complexity for drain color defects. Gauss transform and image enhancement method is been proposed to detect defect images, the detection success rate reached 99%. Meet the requirements of industrial batch testing.