The Research About Register Detection In The Rotary Screen Printing Based On Machine Vision

The register accuracy detection of the rotary screen printing machine is one difficulty inthe research field of printing and dyeing control equipment. It is the main task of the newgeneration textile machinery equipment to improve the register accuracy，to reduce the patternerror or off-pattern phenomenon and increase on-line detection method for theequipment.Manual visual checking is a traditional way.Due to human physiological factors, itcan not achieve accurate register detection. Machine vision can replace human eyes to domeasurement and judgement.This method not only can replace manual detection, but also hasa number of advantages,such as high-speed, long working hours and high precisionmeasurement.With the aim to improve the register accuracy of rotary screen printing machine, theschemes of introducing machine vision technology into the register error detection wasproposed in this paper. To ensure the measuring precision,the linear array CCD was used tophotograph the fabric which follows the motion of the conduction band.And the fabric imagewas segmented and matched to achieve the register accuracy detection online. The majorwork of this paper is as follows:Fristly,the structure and technological process of the rotary screen printing machine wereintroduced, and the detailed analysis about register error was given.Then the automatic testprogram based on the machine vision was designed.Then,a detailed analysis about fabric image noise and the current image pre-processingalgorithms was made.The PGF nonlinear filtering method was chosen to do the imagefiltering preprocessing. As there is rich texture and color information in the fabric image,JSEG segmentation algorithm was used to do the segmentation in this paper.The JSEGalgorithm is able to take full advantage of texture and color information in image to do thesegmentation and thus it was particularly suitable for this color fabric image.Finally, to eliminate the shooting error, the edge of the target image was matched withthe standard edge image stored in the computer before,using Fourier-mellin curve matchingmethod. When it was finished, the curve matching method based on Fourier transform displacement theorem was used.The aim of the Fourier-mellin curve matching was to adjustoverall migration, and eliminate the shooting error.The aim of the second time matching wasto achieve accurate positioning of the register error and to calculate the size of the error. Thesimulation and experiment show that the system have high detection accuracy.