Single closed contour trademark classification based on support vector machine

Given an single closed contour trademark image, shape is one of the most important features in content-based trademark image retrieval and classification. So, we can extract the target image contour Fourier descriptor as feature vector. Fourier moments are not invariant to image scaling, rotation and translation, therefore Fourier moments are used as feature vector such that Classifier has better classification performance than traditional classification methods. The application of Support Vector Machine model solves the problems of poor generalization performance, local minimum and over fitting. In addition, kernel function applied in support vector machine maps data set linear inseparable to a higher dimensional space where the training set is separable. For this reason Support Vector Machine classifiers are widely used in pattern recognition.