Research and Application of a Method for Real Estate Document Image Classification

摘要

In order to quickly and accurately classify the massive real estate documents, a novel method of automatic classification for document image is presented. Based on the paragraph and local pixel feature, it is accomplished by SVM classifiers with Huffman tree architecture. This method, first, extracts the paragraph and local pixel features of the preprocessed document images, then constructs the SVM multiple classifiers based on the Huffman tree according to these features, finally, the feature vector set is extracted from the massive real estate document images to compare the classification abilities of multiple classifiers. The experimental results show that this method can achieve fast and accurate classification of the document images and has good application value on the automatically classification of the real estate archives.