Oracle Bone Inscriptions Image Recognition and Classification Method Based on Deep Learning

Applying advanced image processing and deep learning techniques to the segmentation and recognition of oracle bone inscriptions can help accelerate the research process and improve the efficiency of cultural relic protection and historical and cultural research. Firstly, mathematical morphology methods are used to denoise the original oracle bone rubbings, which can better identify and process interfering elements. Subsequently, by applying one of the most advanced visual algorithm models, YOLOv8, to image recognition and segmentation of raw oracle bone rubbings, the recognition and segmentation tasks can be quickly and accurately completed. Finally, an improved ConvNeXt, which represents the most advanced convolutional neural network, was firstly applied to the classification experiment of oracle bone character images, achieving a validation accuracy of 94.5%. Through the aforementioned steps, a complete and effective method for pre-processing, image recognition, image segmentation, and image classification of oracle bone inscriptions is constructed in this paper.