Research And Application Of Bronze Bronze Image Recognition

Bronze inscriptions unearthed in archaeology are precious written materials.Accurate and rapid understanding of its interpretation and the evolution of character forms is of great significance to archaeology,history and linguistics research.In this paper,the data collection and preprocessing of the inscription rubbings are carried out,and the inscription image database is constructed;the recognition method of the word meaning classification of the inscription head and the accurate matching of the inscription variants is studied,and the scientific and technological support is provided for the cultural relics experts to study the inscription recognition..The main contents include:(1)Construction of the image database of bronze inscriptions.Using authoritative documents such as "New Inscriptions" as the original data source for inscription information collection,the automatic acquisition,segmentation and preprocessing methods of inscription images are studied,and the digital standard collection of original data is realized;inscriptions and bronzes are analyzed.,The relationship between the meanings of the characters,based on the principle of relational database,designed the ER relationship between Chinese characters-inscriptions-bronze data,and constructed an IB database for standardized storage of bronze inscription data.(2)Research on classification and recognition method of inscription image prefix based on hybrid features and SVM.Aiming at the problem that a single inscription feature cannot represent the inscription,the classification and recognition method of the prefixes of bronze inscriptions with mixed features is studied.Aiming at the problem that there are few inscription image samples that are not conducive to classifier training,based on the inscription image database,the training samples areexpanded by generating a confrontation network.Extract the orientation gradient histogram feature(HOG)and gray level co-occurrence matrix feature(GLCM)of the bronze inscription image,and train the SVM classifier by combining the structural feature and local feature to realize the recognition of the bronze inscription image prefix and improve the recognition accuracy Rate,reducing misjudgments.(3)Research on the matching and recognition method of Bronze inscription image variants based on dimensionality reduction SIFT-RANSAC.In order to realize the accurate recognition of a single Bronze inscription image variant,based on the classification and recognition of the prefix of the Bronze inscription image,a Bronze inscription image variant matching and recognition method based on dimensionality reduction SIFT-RANSAC is proposed.Aiming at the problem of low efficiency and easy mismatching of the classic SIFT algorithm,by reducing the dimension of the feature descriptor of the SIFT algorithm from a rectangular area to four concentric circles,the dimension reduction of the SIFT feature vector is realized,and the matching feature points are optimized by RANSAC.The mismatch rate of image recognition improves the efficiency and accuracy of recognition.