A SVM-based Image Classification Method in Document System of Personnel Archives

Information technology has deeply penetrated into the personnel archives management to improve its security, privacy and high efficiency. This paper comes up with a method to solve problems about image classification. It combines SVM with Huffman tree to construct a classifier HFM-SVM. Constructing HFM-SVM, it extracts paragraph and local pixel features of archive document images as training samples and test data and can classify all of personnel archive documents into five classes such as ID cards, application forms and labor contracts and so on. Comparing with multiple classifiers, the experimental results show that HFM-SVM does better in automatically fast and accurate classification of personnel archive document images.