Research On Enhancement And Recognition Of Mongolian Furniture Patterns Based On DWT-AMSR And YOLO Algorithm

Mongolian furniture is an important carrier of Mongolian culture.In recent years,the quantity of Mongolian furniture has decreased dramatically and its quality has been damaged to different degrees,so the identification and classification of furniture patterns are considered in order to provide an effective reference for the conservation study of Mongolian furniture patterns.Mongolian furniture is decorated with various forms of patterns,mainly animal,plant and geometric patterns.Most of the current studies on Mongolian motifs rely on photographic data and cultural relics in collections,but the environment,equipment,history and other conditions make the obtained furniture motifs not clear enough and affect the research results.Therefore,this thesis enhances the fuzzy and distorted patterns and then identifies and classifies the three types of patterns.The main research contents are as follows:(1)The principles of several enhancement algorithms and their enhancement effects on Mongolian furniture patterns were introduced: Histogram Equalization algorithm(HE),Artificial Gamma Correction algorithm(AGC),Homomorphic Filtering,Single-scale Retinex algorithm(SSR),Multi-scale Retinex algorithm(MSR),and Multi-scale Retinex algorithm(MSRCR)with color recovery.The DWT-AMSR based on Mongolian furniture pattern enhancement algorithm was proposed.The adaptive Multi-scale Retinex algorithm processes the low frequency components after discrete wavelets,and the improved threestage threshold denoising processes the high frequency components.The experimental results show that the contrast,brightness and detail of the pattern processed by DWT-AMSR algorithm ware improved,and the color deviation was reduced.(2)The enhanced grain pattern images were evaluated,and the enhancement effect was analyzed using five objective evaluation indexes such as Peak Signal-to-noise Ratio(PSNR),Mean Square Error(MSE),Structural Similarity Index(SSIM),Information Entropy(IE),and Enhancement Mesure Evaluation(EME).The results show that the DWT-AMSR algorithm has better performance in the seven algorithms and enhances the quality of the pattern to a greater extent.(3)The YOLOv5 model was used to recognize Mongolian furniture patterns.Mongolian furniture patterns were classified into three types: animal patterns,plant patterns,and geometric patterns,and the recognition effect of the model on Mongolian furniture patterns was evaluated using Precision rate(P),Recall rate(R),and mean Average Precision(m AP).The overall recognition accuracy of the three patterns was 77.9 %,the recall rate was 92.9 %,and the average accuracy was 95 %,which can effectively identify the patterns.