Application of The Multi-feature Splicing Technology Based on Residual Network Identification
Xiangzhen Rui, Xuchao Zhang, Runying Wang · 2021 International Conference on Electronic Information Technology and Smart Agriculture (ICEITSA) · 2021
The use of computer technology to assist the restoration of cultural relics can improve the efficiency of practical work and avoid secondary damage during the research and management of cultural relics. Especially for paper recognition of debris region, the rational use of deep learning algorithm can obtain better recognition effect. However, the accuracy of shallow network identification is too low, and deep network degradation will occur. In fragment matching and splicing analysis, it is difficult to obtain excellent results by using the simplified features if the geometric contour is too complex. Therefore, on the basis of understanding the current research status of residual network identification, according to the features of multi-feature splicing technology for cultural relics restoration, this paper analyzes the splicing method based on multi-feature constraint of contour line, and selects the Terra Cotta Warriors fragment data set for empirical analysis, so as to clarify the effectiveness of multi-feature splicing technology for cultural relics restoration.