Research on Identify Image of the Inscription at the Bottom of Purple Clay Teapot Based on Two-Classification Algorithm

We use an improvement to the traditional sift algorithm to extract some more stable feature points of the inscription image at the bottom of purple clay teapot. We found the appropriate features as input, and used two-classification algorithm to train the classifier, and obtained the classification function with better classification ability for the matching of the inscription image at bottom of the teapot. In this paper, we created a sample library containing 100 sets of images and 14968 data samples. The experimental results show that when we choose the first 100 singular values of X-Y (X and Y respectively refer to the feature descriptors of the two images to be matched, obtained by improved sift algorithm) as the input of the classifier, the classifier trained by the Gaussian radial basis kernel function of SVM has better classification ability. The classification accuracy rate reaches 83.47%.