Preliminary study of vision system for the colorfastness rate assessment on woven fabrics

Commonly, subjective method base on human visual perception is the main method in textile colorfastness assessment. However, this method is time consuming and relatively inaccurate. This research aims to develop an objective method using vision system as an alternative method to perform woven fabrics colorfastness assessment. This vision system method relied on texture analysis, which is based on woven fabrics samples color homogeneity quantification. Firstly, image pre-processing method is conducted to get minimum fraction channel of the image, followed by feature extraction process using histogram method image and Grey Level Co-occurrence Matrix (GLCM). The pixel space used in GLCM method was 4 pixels with 0°, 45°, 90° and 135° observed orientation angle. Variance and bucket histogram are used as the parameters on the feature's extraction through histogram analysis while GLCM use contrast, correlation, energy and homogeneity aspect as its parameters. Output of the feature extraction became input for classification system using artificial neural network. The artificial neural network analysis results shown that a system using a parallel input combination of contrast, homogeneity and variance can be used as a proper input for woven fabrics colorfastness test. This vision system results showed the validity of the system which can determine the colorfastness rate of the woven fabrics with accuracy of 100% for red, purple and yellow, 83.33% for green and 91.67% for blue.