CLOTHING PATTERN RECOGNITION FOR VISUALLY IMPAIRED PEOPLE

Abstract

It is very difficult for visually impaired people to choose clothes with different patterns and colors. Pattern Recognition method can be implemented to carry out automatic cloth pattern recognition with machine learning approach. For researchers, automatic clothing pattern recognition as a challenging research problem due to the factors of large intra class pattern variations, scaling, illumination and rotation. The system proposed here considers a PC based system that can be used to recognize the clothing patterns in four different categories (stripped, plaid, irregular and pattern less) and also identify color in terms of percentage of 15 basic colors. It is implemented using MATLAB tools for image processes. A database of cloth images present in wardrobe of a visually impaired person is prepared. Features of the test image can be extracted using special functioned descriptors. After identifying the features, nearest neighbor method classifies the images in order to their categories. By normalizing the histogram of image in HIS color space, colors in the image are identified. The pattern and color identification result is given as audio output.