Research and Application of Chinese Paints Classification and Recognition for Identifying Hyperspectral Imaging Technology

Chinese painting is famous for its special painting materials, techniques and final artistic effects. However, there also have been many fake paintings. These fake paintings are usually produced by using modern high-resolution scanning and printing techniques, making it challenging to identify fake paintings through human vision. To solve this tough problem, we propose a method for pigment classification and recognition based on hyperspectral imaging technology in this paper. Firstly, a video with a hyperspectral camera is used to scan Chinese paintings with spectral frequencies ranging from 400 to 900 nm. Subsequently, spectral and spatial features are extracted through principal component analysis and convolutional neural networks, respectively. Finally, the two functions are identified, and the feature set is entered into a support vector machine for classification. All Chinese paint samples come from Chinese local painting organizations. The experimental results show that the effectiveness of the method is 84.6%, which is significantly higher than other methods where only spectral or spatial features are used.