Aiming at the matter for the detection of fabric defect, this paper uses the strategy of Fast Fourier Transform and basic Image processing operations to find the material defect, the algorithmic rule planned during this paper will considerably shorten the time of detection on the idea of guaranteeing the proper detection rate.
Fabric defect detection could be a key part of the textile business, the strategy of detection of material defects square measure principally focused within the artificial check for surface defects. This methodology can cause labor redundancy, not solely reducing the proletariat however conjointly increasing the chance of miscarriage of justice, the false detection rate, and false negative rate.
Therefore, to attain quick and correct detection of material defects has become a hot subject in recent years. At present, there are few kinds of research of material defect detection, generally are: FFT (Fast Fourier Transform) is Associate in Nursing algorithmic program that computes the separate discrete Fourier transform (DFT) of a sequence, or its inverse (IDFT). analysis converts a symbol from its original domain (often time or space) to a illustration within the frequency domain and contrariwise, this technique incorporates a detection rate is comparatively high will satisfy the period of time detection.
Fabric Image may be a 2-D image, the material is captured within the type of image exploitation camera or the other image capturing devices , then material pictures is regenerate to HSV color channel exploitation HSV associate object with bound color may be detected and scale back the influence of sunshine intensity from the surface. Gaussian may be a type of noise that is typically found within the captured pictures and this noise might misunderstanding with the important defects present on the material texture. when changing the initial material image to HSV image, Gaussian by exploitation Gaussian blur that results of blurring the pictures, Gaussian blur may be a filtering technique used to sleek the digital image or blur the digital image, Gaussian blur is additionally called Gaussian smoothing. 2-D image contains 3 color channels, ripping of color channels and apply the fast Fourier transform to every color channels then applying of OTSU threshold operation on the fast Fourier transform output allows to identifies the defected region, next we’d like to differentiate the various between the noise and also the defected half supported the intensity values of all the 3 color channels , when obtaining the ultimate image we discover the contours to get the sides coordinate of the defected region and draw the contours to focus on the defected region
Split the color channel to R,G,B
Convert to HSV color channels
Read The Image
Step 1: Scan the input image of 512*512 pixels
Step 2: Convert the input image to HSV color channel
Step 3: Split the color channels R,G,B
Step 4: Apply the Fast Fourier Transform to every color channels
Step 5: OTSU Threshold for every output of Fast Fourier Transform
Step 6: Defect Extraction
SIMULATION AND ANALYSIS
The software environment of this paper is as follows: Win7 64 bit OS, anaconda integrated development atmosphere, and spyder compiler
Experiment hardware environment:
central processor Core i7 930 ?Frequency 2.8GHz?
Memory 2GB GPU NVIDIA GeForce GTX650
The simulation experiment, the jump flower, hook broken wire material defects are the foremost common sort of textile business. material defect image to be designated to find the scale of the 512*512.
Input Image Output Images
Quality control Fabric business is significant for reducing price of merchandise and rising quality of the products. exploitation digital image method technology can supply less price of labor, fast and proper method. investigation defects and classifying the reason of inflicting defects supported the image processing offers the advantage in control and speedy production. exploitation the mixture of various ways of defects detecting can supply less complicated defect investigation technique and higher rate of identification This paper achieves 2-D FFT material defect detection, once the material image is reworked, the high speed and parallel methodology is adopted to satisfy the high period characteristics of the particular industrial production and detection of FFT. algorithm projected during this paper can even effectively discover the common defects. within the detection time potency, compared with the other algorithms, the speedup is concerning four times.
 Chi-ho Chan, Grantham K. H. Pang, “Fabric Defect Detection by Fourier Analysis,” IEEE Trans on Industry Application., vol. 36, pp. 513-518, 2000.
 S.Q. Zhang, D.S. Yu, Design and implementation of a parallel real-time FFT processor, Proceeding of ICSICT2004, pp. 1665-1668,2004.
 D.F. Zhang, MATLAB digital image processing. Beijing, Mechanical Industry Press, 2009.
 Y.L. Jiang, C.F. Xu, Fast Fourier transform FFT and its application, Photoelectric Engineering Press , 2004.