Abstract:

In order to solve the problem of low-accuracy classification in objective automatic evaluation of seam-puckering, a novel method based on gray level co-occurrence matrix, wavelet analysis and back propagation (BP) network was proposed for automatic seam-puckering evaluation. Firstly, a standard seam image was captured and the gray level of the seam image was reduced to 16 level, the gray level co-occurrence matrix, and the mean values of the characteristic parameters were obtained in the direction of 0° and 90°, respectively. Meanwhile, the standard deviation of the horizontal detail coefficients of the seam image was extracted and calculated by using Haar wavelet of the sixth anslysis scales. After that, five extracted characteristic parameters were input to the BP neural network for training and recognizing, and the standerd seam image was exaluated. The evaluation results show that the proposed algorithm, compared with one adopting gray level co-occurrence matrix characteristic or wavelet characteristic alone, has higher correct classification rate and stable classification effect.

Key words: seam-puckering, gray level co-occurrence matrix, wavelet analysis, BP network