A Multi-feature SVM Classification of Thangka Headdress

The classification of Thangka headdress has a wide application in the semantic retrieval and semantic annotation of Thangka. The existing classification methods are facing difficulties of segmentation, which affects the practical application. In order to improve classification efficiency, this paper proposes a new method using support vector machine (SVM) for the classification of Thangka headdress with combination of multi-features. The new classification adopts the following steps: firstly, segment the Thangka headdress using Kirsh segmentation method, secondly, extract the features of Hu moments, Fourier moments, and Zernike moments, thirdly, extract the color features, the total being 35 features, finally, verify the classification accuracy using SVM, BP neural network and the random forest respectively. The classification result shows that SVM classification can achieve the best accuracy and the requirements of practical application.