A method of Jia Gu Wen recognition based on a two-level classification

Jia Gu Wen is the earliest Chinese character. Since it has many of the properties of drawings, common recognition methods cannot be efficiently applied. We provide a recognition method using a two-level classification. First, we regard the character as a non-directional graph and extract its topological properties. Thus we separate the JGW characters into different classes. Then, based on the definition of an extensive stroke, we extract the relevant properties of the strokes and make a second-level recognition. Experiments have shown that the recognition rate is above 95%.