Defects Detection Of Knitted Fabric Using Computer Vision Based On Labview

During the processing of knitted fabric,quality control is very important while the most important thing is the detection of knitted fabric defects.The efficiency of the production in the textile industry has great improvement due to science and technology but the detection of knitting defects has no process.There are so much limitation such as low efficiency to use the manual work in the detection of fabric defects. The biggest problem is how to use computer vision to detect the fabric defects instead of manual work.This paper introduced a system and method for defects detection of knitted fabric using image processing and 2D discrete wavelet transform programming with Labview and Matlab were presented. Firstly median filter was used to remove noise and then binarization was worked on the image.After image preprocessing db wavelet transform was introduced to analyze the image,then the eigenvalues of image were extracted to recognize defects,finally part of the eigenvalues were used to train the BP networks from which could pick out the best one.The best BP network was used to test the rest eigenvalues.In the paper,how to choose the eigenvalues,which eigenvalues was used and how to sort the knitted fabric defects which were the most important was investigated.(1) The energy,the variance,the difference and the entropy were picked out to be the eigenvalues which could feedback the characters of the knitted fabric defects after many experimentations.(2) The knitted fabric image was divided up according to the texture of the knitted fabric.The warp and weft texture image was gotten after the db wavelet transform to analyze knitted fabric image.The eigenvalues which were extracted from the warp and weft texture image could feedback the characters of the knitted fabric defects.(3) The D-value between the max eigenvalues and the min one in every image was used to train the BP networks.The best BP networks which was picked out to test the rest engenvalues was used to calculate the rate of the right classification.The experiments indicated that the method could effectively detect common defects such as hole,drop stitch,fly waste and mark waste in plain,rib and interlock fabric.The right rate is 94%.