Research And Implementation Of Image Enhancement And 3D Reconstruction Of Western Xia Auspicious Beast Cultural Relics

Cultural relics carry splendid civilization,inherit history and culture,and maintain the national spirit which are important resources to carry forward the excellent traditional Chinese culture.Auspicious animals unearthed in Western Xia are not only the heritage of Western Xia civilization,but also the excellent traditional cultural heritage of the Chinese nation.In recent years,with the rapid development of Internet cloud computing,big data artificial intelligence and multimedia technology,it has brought a profound impact on the protection and development of traditional excellent cultural heritage and provided broad space and technical support.In order to make the ancient cultural relics live,we should further strengthen the digital collection and reasonable protection and utilization of cultural relics information.This paper focuses on the three-dimensional reconstruction of Western Xia Rui animal cultural relics,and has carried out the research on image classification,low illumination image enhancement,image feature extraction and matching,and 3D reconstruction based on multi view,the main contents include:(1)In view of the fact that the image samples of Western Xia auspicious animals are few,this paper proposes a method of transferring large-scale pre training neural network to complete the training on Western Xia auspicious animals small-scale data set.This method not only achieves high recognition accuracy in image classification task,but also has fast training efficiency.(2)Aiming at the low illumination Western Xia auspicious animal cultural relic image,a lightweight image enhancement network model is constructed,which can enhance the image illumination.The network model has good enhancement effect for Western Xia auspicious animal image with low illumination,and has high processing efficiency,which lays a good foundation for further image feature extraction.(3)The accuracy of image feature extraction will directly affect the results of 3D reconstruction.Therefore,this paper makes a detailed study on image feature extraction and matching.By introducing the asymmetric convolution module to improve the SuperPoint network model,the accuracy of feature detection is improved to a certain extent.Through comparative experiments,not only the repeatability is improved,but also the correct matching rate is higher,It is more conducive to the follow-up study of 3D reconstruction in this paper.(4)In this paper,Zhang’s calibration method is improved by using ring calibration plate,eccentricity error algorithm and histogram equalization.To a certain extent,the calibration error caused by distortion parameters is corrected,and the calibration accuracy is improved,so that more accurate camera internal and external parameters can be obtained,which provides a strong data guarantee for the next three-dimensional reconstruction research in this paper.(5)In this paper,a complete set of 3D reconstruction process based on multi view is designed.According to the user’s input image of auspicious animal relics,the user’s accurate classification prompt is given,and the input low illumination image is enhanced and reconstructed.On this basis,the 3D point cloud model of Western Xia auspicious animal relics is generated by using SFM algorithm and PMVs algorithm.The comparative experiments show that the 3D reconstruction algorithm in this paper can obtain more accurate and dense 3D point cloud results compared with other methods,in order to make due contribution to the digital protection of Western Xia auspicious cultural relics.