Automatic recognition of textile texture using back-propagation neural network

This study proposed to use wavelet transfer to acquire image features, and use back-propagation neural network to classify type of textile texture. Firstly, wavelet transfer is applied to obtain vertical, horizontal and diagonal images of original image, and compute its wavelet energy to take them as texture features of this image. Finally, the back-propagation neural network is adopted to recognize texture feature of this image. As indicated by experimental result, this system can recognize accurately texture in woven fabric. Among 350 test samples in total, the general recognition rate amounts to 96%. Therefore, this study succeeded in building the automatic computer visual inspection system to recognize textile texture type, which can greatly improve and avoid current low efficiency, non-objective judgment and labor waste due to human inspection.