| Ancient ceramic relics have a long history and carry rich historical,cultural and artistic connotations.In recent years,with the great abundance of people’s material life and the continuous enrichment of spiritual civilization,the research,appreciation and collection of ancient ceramics and cultural relics have become more and more dynamic.As the market for ancient ceramics and cultural relics continues to expand,the research on ancient ceramics and cultural relics has gradually shifted from traditional expert experience appraisal to modern scientific appraisal.Nowadays,the scientific identification of ancient ceramics based on the fields of materials,physics and optics is gradually improving,but the above-mentioned methods generally have the problems of long detection cycle,easy damage to cultural relics and high detection costs,which are difficult to popularize.Therefore,how to design lightweight and high-standard modern scientific identification methods to assist in the development of scientific identification of ancient ceramics and cultural relics is extremely urgent.Since ancient times,the appraisal of ancient ceramics has usually relied on experienced experts to make inferences and judgments based on their own rich knowledge.However,such artificial appraisal is highly subjective and the appraisal results are easily affected by the cultural literacy and empirical knowledge of experts and scholars.In recent years,ancient ceramic identification methods relying on modern scientific methods have flourished,but the current scientific identification mostly focuses on the analysis and research of cultural relics’ own materials,and the identification process is cumbersome and easy to cause cultural relics to be damaged.It can be seen that the research on the identification of ancient ceramics is quite limited.Nowadays,with the continuous development of digital research on cultural relics,the research and application of ancient ceramics and cultural relics images are gradually enriched.Therefore,this research combines the in-depth application of modern machine vision technology and deep learning technology in the field of image processing,and proposes a convolutional neural network(CNN)based image recognition method for ancient ceramic artifacts.By building a CNN dating model,collecting and processing a large number of ancient ceramic artifacts from different dynasties for training and testing,establish a CNN-based ancient ceramic dating classification model.Automatically extract features based on the CNN network,avoid the tedious process of manually extracting features blindly in the traditional image processing and classification process,and establish an efficient and convenient ancient ceramic dating identification algorithm to assist manual identification or other scientific research methods to accurately identify ancient ceramic cultural relics.In addition,the type of ancient ceramics has always been a prominent feature in the identification of cultural relics.Not only that,the three-dimensional model of cultural relics is also of great significance to the study of the evolution of cultural relics,historical verification,and the characteristics of the times.It is an important carrier to promote the inheritance of Chinese culture and art.In a market where the authenticity of ancient ceramics is incomplete and scarce,the realization of the restoration of the three-dimensional model based on the ceramic image is of special significance for the protection of cultural relics and the inheritance of historical culture.Therefore,on the basis of studying and solving a series of technical problems such as contour edge extraction,image distortion correction,and contour nonlinear modeling of rotating ancient ceramic images under complex background,this paper has researched and realized a set of ancient ceramic three-dimensional Reduction algorithm.The specific research content of this article includes:(1)Research on the date and classification algorithm of ancient ceramic imagesTaking the ancient ceramic image dating and classification algorithm based on the convolutional neural network as the core idea,the CNN dating model is established through the process of building the convolutional neural network dating model structure,collecting ancient ceramic image samples,and training model optimization parameters.Aiming at the problem of small training samples of ancient ceramics,a high-precision dating model was established by enhancing image samples,adjusting network structure parameters and iterative strategies,and completing the accurate dating of ancient ceramics by testing the samples to be tested.(2)Research on three-dimensional restoration algorithm of ancient ceramic two-dimensional imagerThe three-dimensional restoration of ancient ceramic imagery is mainly aimed at the two-dimensional image of rotating ancient ceramics obtained from imaging equipment such as cameras,through normalization,grayscale,image filtering and noise reduction,wavelet sharpening enhancement,binarization,and contour The edge extraction and other preprocessing processes extract the contours of the edges of the ancient ceramic images,and then extract the upper and lower edge ellipses according to the ancient ceramics imaging of the rotating body,which can be used to optimize the imaging distortion effect,and then use the neural network modeler-type side edge contours.According to the intersection point and the central axis of the extracted contour,the three-dimensional model of the ancient ceramics of the rotating body is generated by the rotation restoration.(3)Design and implementation of ancient ceramic digital software systemCombining with the needs of the system,establish a digital software platform with friendly interaction,access operations,and process visualization.Design software based on App Designer and build each module of the software system.It mainly includes software interactive interface design,image acquisition and storage,algorithm migration module,database module,etc.Finally realize the three-dimensional modeling of the ancient ceramic images of the rotating body and the dating and recognition of the ancient ceramic images,providing the access operation of the image data and the data of each process model,the visual interaction of the various stages of the restoration of the model,and the query record of the dating results and other functions.Through the research of the ancient ceramics image dating and the overall test of the three-dimensional restoration of the ancient ceramics of the rotating body,the results show that the ancient ceramics three-dimensional restoration algorithm can accurately obtain the three-dimensional model of the ancient ceramics of the rotating body,when the shooting angle is controlled In the range of-10°～10°,the relative error between the restored three-dimensional model and the standard three-dimensional model is less than 1.7%,which greatly facilitates the analysis and research and restoration of ancient ceramics.The established CNN ancient ceramic dating model breaks the technical bottleneck of"determination of ancient ceramics’ characteristics depend on manual completion",and the accuracy rate of classification and dating of ancient ceramics reaches 96.37%,which can be used as an effective auxiliary means for ancient ceramics identification.It is of certain significance to explore the new development direction of ancient ceramics identification research. |
PDF Full Text Request