Research on Oracle Bone Script Image Segmentation and Automatic Recognition Based on Machine Learning

This article is based on the problem of segmentation and recognition of oracle bone script images, using digital image processing and machine learning methods to study and discuss the key technologies of oracle bone script digitalization engineering. Three original oracle bone script rubbings were preprocessed and a model was established to achieve preliminary judgment and processing of interference elements in oracle bone script images. Before preprocessing, it is necessary to identify the interfering elements in the image, so that we can “prescribe the right medicine” accordingly; At the same time, there are many ways to deal with image interference elements, and the advantages and disadvantages of different processing methods can also be compared. In order to achieve automatic single character segmentation of different original oracle bone images, this paper needs to establish corresponding oracle bone image segmentation models through machine learning. Building a model requires a test set and a training set. Preprocessing the annotated and segmented dataset ensures that JPG files correspond one-to-one with JSON files; After the model is established, its performance can be evaluated through the corresponding parameters during the model building process or by comparing different oracle bone image segmentation models. This article proposes the problem of automatic text segmentation and storing the segmentation results in XLSX files. Firstly, the oracle bone images need to be preprocessed. Through preprocessing, the clarity and recognizability of the images can be increased, which can improve the efficiency of machine learning. Finally, we need to output the obtained results to the XLSX table.