A combined recognition system for oracle bone inscriptions

摘要

>Oracle bone inscriptions (OBIs) are figures inscribed onto cattle bone or turtle shells 3,000 years ago. Understanding these inscriptions can provide a lot of insight into world history. A recognition system that combines two recognition methods for recognising OBIs, line features recognition and template matching, is developed in this paper. First, the system reduces the noises in the OBIs using Gaussian filtering and labelling and then extracts the skeleton using affine transformation and thinning. Next, the line feature points of inscriptions are extracted by clustering. The system then calculates the minimum distance of the line feature points between the original image and templates in the OBIs database and lists the five most minimum templates as the candidate templates. Finally, the template matching is applied for recognising the affine transformed inscriptions from the five candidates. Experimental results show that 89% of inscriptions are recognised on the third-most similar templates.