Study On Trademark Image Classification Based On Support Vector Machine

It is the most important strategy to establish the brand for business development. The brand represents company's reputation, product quality and cultural value. Trademark as brand symbols of most intuitive performance, symbol of intellectual property rights of products, attracts more and more attention. For this reason, the number of trademark registration increases rapidly year by year, but it brings difficulties to trademark management. However, because of time-consuming, subjective weakness problems traditional manual classification management method can't meet the trademark image needs.For the classification of trademark image, the paper described classification method based on support vector machine. Firstly, the paper studied on support vector machine classification method in theory, analyzed the great performance of support vector machine in solving non-linear, the curse of dimensionality and local minimum problems, and discussed the key factors influencing the performance of classification.Secondly, the article introduced the feature extraction techniques of trademark image, analyzed the main features of the shape description, presented the method of trademark image shape feature in detail and elaborated the method of regional characteristics and boundary contour. The paper analyzed the boundary description of the method of Fourier descriptors.Finally, for single closed contour image classification, the paper constructed classification model based on support vector machine using Fourier descriptors as feature vectors. Also, construction process of classification model is given in article. The paper illustrated the feasibility of constructing classification model using support vector machine based on experimental data.