Method And Implementation Of Sketch Extraction Of Painted Cultural Relics

Painted cultural relics have a long history and rich colors,which represent historical changes and the replacement of civilizations.They are an indispensable part of the development of civilization in China,and have high historical and scientific research value.The lines of painted cultural relics vividly reflect the times style and painting techniques of artists in different periods.The lossless line sketch extraction can help us understand the intention and method of artists' painting more clearly,and at the same time,painted cultural relics can be better inherited and protected in this way.This paper studies and implements an intelligent and automatic extraction system for the lines of painted cultural relics,which can accurately and completely extract painted cultural relic lines,and the system can be used for the images collected by general cameras and hyperspectral imagers at the same time,which brings convenience for the research and inheritance of painted cultural relics.The main contents are as follows:(1)Make the data set of painted cultural relics.As there is currently no special painted cultural relics data set for line extraction,this paper has investigated and verified the existing methods to design the production process of colored cultural relics line data,and has created a data set specifically for the research of painted cultural relics line extraction.(2)Introduced a Bi-Directional Cascade Network.Line extraction was performed on painted artifacts from the perspective of edge detection,and the number of key parameters Incremental Detection Block and loss function weights were studied.Experiments were performed on data collected by general digital cameras and hyperspectral data.The results show that this method can obtain clearer and more complete lines than the existing methods.(3)For some painted artifacts that are severely damaged or noisy,in order to obtain more complete line drafts,a generation adversarial network was introduced to perform humanmachine interactive repair on the extracted lines,and more complete lines were obtained for subsequent copying and further protection and repair.Provide convenience.(4)Designed a deep learning-based line drawing extraction system for cultural relics.The system includes four modules: image visualization,hyperspectral image processing,line sketch extraction and line sketch repair.The system integrates the above line sketch extraction and line repair methods,which can process general digital Camera images can also process hyperspectral images,providing a processing platform for multi-source data line sketch extraction.