Big Data Processing and Recognition of Oracle Topography Images Based on YOLOv8

The use of artificial intelligence and other technical means for oracle bone recognition is crucial for the inheritance of Chinese civilization and the promotion of language and writing research in the new era. This paper proposes a method to recognize oracle bone inscriptions based on YOLOv8 model processing. Firstly, median filtering and open morphological processing methods are used to reduce the noise, and then binarization and corrosion/expansion operations are carried out to highlight the fonts in order to reflect the morphology and characteristics of the fonts. And at the same time, bilinear interpolation is used to optimize the smoothness of the image and increase the curvature of the texture. Considering that some of the topos will be broken, this paper fills the topos with cracks. After that, the above processed images are used for character distribution recognition and automatic segmentation of oracle bone characters, and the obtained data are input into the YOLOv8 model for training and the adaptive optimizer is used to optimize the learning efficiency. After 100 iterations, a robust YOLOv8 model for automatic segmentation of oracle characters is obtained. The model can realize the functions of automatic segmentation of oracle text, data labeling and confidence level display.