The Research Of Automated Detection Based On Machine Vision For Rotary Screen Printing

In the pringting process of Rotary screen printing machine, there is inevitably "errorpringting" phenomenon due to mechanical equipment and restrictions of the process.Present detection and adjustment can be only made by manual observation. the way ofmanual observation are labor-intensive, low printing precision and low printing efficiency.With the rapid development of computer science, image processing, pattern recognitionand other areas, the technology of machine vision becomes more well. Therefore, inindustrial production, it is widely used in automated inspection of product quality.Machine vision technology will be introduced to the automatic detection of rotary screenprinting, it is currently a hot research of the textile industry.To improve the pattering precision of rotary screen printing machine, this paperproposes a program by introducing machine vision technology into the pattering errordetection of rotary screen printing. Through the image analysis and processing of printingand dyeing fabrics, and curve matching of image regions, the detection of patteringprecision can be achieved.This paper is divided into following five parts:Firstly, to begin with technical process of rotary printing machine, and after theanalysis of pattering error formation, this paper proposes the program of introducingmachine vision technology into the pattering error detection of rotary screen printing,combining the technical characteristics.Secondly, describes the image pre-processing-related technologies, including theknowledge of color space, image noise, image filtering and image enhancement.Thirdly, the relevant knowledge of image segmentation is described, focuses on theK-mean algorithm, mean shift algorithm and the EM algorithm for Gaussian mixturemodel. The K-mean algorithm has been improved, K-mean clustering operation for theimage from RGB to Lab color space conversion. EM algorithm for Gaussian mixturemodel is also improved, the operation of image segmentation from RGB to YCbCr colorspace conversion. These improvements achieve good segmentation results. The validationof image segmentation for mean shift algorithm, also have good segmentation results.Fourth, the introduction of a quarter of the field edge detection algorithm and Fourier-Mellin curve matching algorithm, and the algorithm has been verified. Accordingto the matching information to calculate the actual wrong distance of fabric printing.Fifth, the experimental automatic detection platform of a rotary screen printing isbuilt based on machine vision. The system’s detection principle is described. A systematicsoftware is designed and developed. The Self-testing process of fabric printing in thesystem software is simulated. The effectiveness of the detection system is verified.