Fake modern Chinese painting identification based on spectral–spatial feature fusion on hyperspectral image

摘要: Chinese painting is famous and valuable for special materials, skills final art effects used, yet this has resulted in many fake paintings being produced. Those were normally made by using modern high resolution scanning printing technology, thus it very hard to identify the ones human vision. To address challenging problem, paper, a hyperspectral image based features fusion method proposed. Firstly, we scan visual band camera with spectral frequency ranging from 400 900 nm. Then, spatial are extracted respectively principal component analysis convolution neural network. Finally, fuse these two kinds of input feature set into support vector machines classification. All samples real obtained local organization. The experimental result shows effectiveness proposed an accuracy achieved 84.6 %, which significantly higher than other approaches where only or used.