| In recent years,people’s cultural literacy has been continuously improved,the cultural life needs has become more and more diversified with the rapid development of social economy,the trend of cultural relics collection has been extremely popular.Among,ancient ceramic relics are the symbol of the splendid civilization of ancient China,the historical witness of the continuous development of ancient Chinese history,and the important carrier of the dialogue cross time and space between modern social skills and ancient culture and art.Thus,ancient ceramics have high art collection value and historical data value.However,with the activeness of the antique market,driven by huge economic interests,the order of the antique market has been destroyed to a certain extent,which is extremely unfavorable to the protection and inheritance of ancient ceramics.Therefore,the accurate identification of ancient ceramics is particularly important.At present,the identification of ancient ceramics is still at the stage of relying on manual experience and scientific and technological instrument.A large number of archaeological identification tasks have been completed through these two methods.However,there are still great limitations.The key point of artificial experience identification is to judge by experience,which is greatly affected by subjective factors,so it is difficult to form an objective scientific basis for identification.While the scientific and technological instrument identification needs to be based on an authoritative and complete database of standard artifacts,supported by advanced instruments and equipment,which makes it difficult to carry out research.In addition,it will cause a certain degree of damage to the samples,which is extremely unfavorable to the protection of ancient ceramics.Therefore,none of the methods can meet the urgent need for accurate identification of ancient ceramics under the background of rapid development of modern society.On the basis of studying the existing identification methods of ancient ceramics,machine learning technology is taken as the core in this paper,a dating method of ancient ceramics based on multi-visual-features fusion is proposed,and the software system is developed.The main research contents of this paper are as follows:(1)Research and digital modeling of appearance characteristics for ancient ceramics.In order to bridge the gap between the abstract semantic description of the appearance of ancient ceramics and the extraction of visual features,the laws of time development contained in the pattern and shape visual features of ancient ceramics images is emphatically explored in this paper based on the analysis and research of the characteristics of ancient ceramics images and the visual feature information.Through the construction of the characteristic digital model,the evolution process of the appearance of ancient ceramics with the development of time is further discovered,and the mapping relationship between the visual characteristics and the dating of ancient ceramics is established.(2)Research on intelligent dating algorithms for ancient ceramicsThe ancient ceramic dating identification algorithm includes prep-rocessing of the ancient ceramic images,extraction of visual features,and the construction of fusion classifier model.Firstly,a pre-processing algorithm including key technologies such as image enhancement,main perspective correction and restoration is designed.Secondly,on the basis of studying the appearance feature distribution of ancient ceramics,the LBP-HOG fusion features and geometrical parameter features of the shape are extracted.Finally,with multi-visual-features fusion technology as the core,the accurate dating of ancient ceramics can be achieved by building a generalized regression neural network(GRNN)based on particle swarm optimization algorithm.(3)Design and realization of the intelligent auxiliary dating software system of ancient ceramics.The intelligent auxiliary dating identification software system of ancient ceramics is designed by C++language,and is on the basis of VS2010 software development platform,which is combined with MFC framework,OpenCV visual development library and other auxiliary technologies and tools.The system is mainly including the construction of software system interface,the transplantation of ancient ceramics dating algorithm,the call of database and so on.Finally,an intelligent system with dating,information storage,display and query has been built.To verify the feasibility and accuracy of the method,the ancient ceramics containing banana leaf patterns as auxiliary decorations(hereinafter referred to as banana leaf pattern ancient ceramics)and banana leaf pattern jade pot spring vases are used as display objects for experiments.Comparative experiments were carried out on the influence of different texture feature quantities,different feature categories,and different classifier models on dating of ancient ceramic.The experimental results demonstrate that the LBP-HOG fusion features can digitally describe the pattern information of ancient ceramics,and the evolution law of pattern for ancient ceramic with the development of the times can be mined and verified from the digital quantities.The dating method based on the multi-visual-features fusion has higher classification accuracy than the currently commonly used dating methods with a single feature,and the accuracy of dating identification can reach 98.18%.In addition,the GRNN model has excellent expressive power in the processing of small samples of ancient ceramic data.In short,the intelligent dating system of ancient ceramics in this paper can be used as an effective means and auxiliary tool to assist manual experience identification and scientific and technological identification methods. |
PDF Full Text Request