Research On Online Defect Detection Method Of Printed Image Based On Machine Vision

Due to the printing technology,mechanical precision and other adverse factors,printing products often appear some defects,such as the wrong print,missing,blurred text,ink dirty,etc.,seriously affected the aesthetics of the print.The traditional method of printing defect detection is mainly done by manual off-line detection.The method of subjective evaluation is not uniform and inefficient,which can not meet the requirements of printing production.At present,most of the domestic small and medium-sized enterprises have an urgent demand for automatic online printing detection.It is of great significance to carry out a research on the automatic online printing detection technology to improve the printing quality of China’s printing industry.The main technologies involved in the online defect detection system consists of three parts:image acquisition,image registration and defect detection.This paper makes a targeted study on the existing problems,as follows:(1)The existing defect detection system mostly installs the synchronization device(rotary encoder)on the printing drum so that the collected images can be synchronized with the standard image.In order to improve the flexibility of image acquisition and the simplicity of user operation,modern enterprises began to use asynchronous acquisition method,the encoder installed in the printing defect detection production line.But this method can not determine the image acquisition position by the encoder,which requires the system to automatically calculate the acquisition position of detect image(also known as the acquisition of the start line).Few studies are done on this problem,the existing algorithm based on template matching is not suitable,this paper proposes an automatic location algorithm for acquisition start line based on row projection template matching for asynchronous acquisition,which can automatically calculate the acquisition start line with high accuracy and good applicability.(2)Image registration depends on the the feature area.However,most printing enterprises still choose the feature area manually which is very inefficient.In order to simplify the operation of the user and improve the automation and intelligence of the system,an algorithm based on contour profile direction and edge gradient direction for automatic feature location of printed image is proposed,which detects rectangles,shapes similar to rectangular or elliptical as the shape feature region in the image,simplify the operation and improve the degree of automation and intelligence of the system.Experimental results show that the method can extract the ideal shape as a feature region stably.(3)The key of shape matching is to describe the shape feature appropriately.A very popular shape matching algorithm called shape context(SC)is excellent.However,the dimension of shape context is too high,leading to its poor real-time performance.In this paper,a shape convexity matching algorithm with fewer features is proposed with reference to the sub-regional,multi-feature statistical characteristics of shape context.Compared with the standard shape context,the algorithm has high real-time performance and good matching effect.(4)An improved differential defect detection algorithm is proposed based on the research of defect detection algorithms,it uses a number of images without defect to train the detection model in order to determine a reasonable tolerance range.The experimental results show that the algorithm has a good detection effect.Based on the above algorithms,a defect detection software is designed with VC++6.0 and Hal con as the software development environment,algorithms proposed in this paper are validated.The experimental results show that the algorithms proposed in this paper are of good application,some algorithms have been applied in the enterprise,which has good practical application value.