Study On Classification Technology Of Wood Species Based On Fusion Of Texture And Spectrum

The identification technology of wood species based on image analysis and spectral analysis is becoming more and more mature.In order to make better use of these two technologies to improve the accuracy of wood classification,this paper proposes a method combining the two technologies of texture and spectrum to identify and classify wood.This article selects the common 21 kinds of timber on the market as the research object,based on the texture characteristics of timber species recognition are studied respectively,and based on the spectral characteristics of timber species recognition,and then established the tree species of wood texture information and spectral information model of feature level fusion and decision level fusion model,and explores the different texture feature extraction method and spectral preprocessing methods,different methods of pattern recognition methods,the fusion of different effects on the results of the identification model.The main work and research of this paper are as follows:(1)Research on Wood Species Identification Technology Based on Texture Features.Traditional timber species recognition study usually use statistical methods of gray level co-occurrence matrix to extract the texture characteristics of wood,and this article tried to put forward in recent years new and efficient texture feature extraction methods are improved basic gray level aura matrix and fuzzy gray level aura matrix used in the field of wood species recognition.This paper compared three kinds of texture feature extraction is used to establish the model of the highest classification accuracy,gray level co-occurrence matrix,improved basic gray level auramatrix,fuzzy gray level aura matrix is 80.95%,87.62%,and 92.86%,respectively,can be seen from the results,this article uses two kinds of texture extraction can realize efficient recognition of 21 kinds of wood.At the same time,it also provides a good foundation for the subsequent establishment of the fusion model.(2)Research on Identification Technology of Wood Species Based on Spectral Features.It is also a common method in the field of wood species identification to classify wood by NIR spectroscopy.In this paper,the use of near-infrared spectrometer and high SOV710 spectrometer instrument respectively to extract the 21 kinds of spectral information of wood transverse section,and through different pretreatment methods for processing,SNV,smooth and differential of the S-G respectively to get the highest classification accuracy is 90.48%,87.62%,84.76%,behind the results good pretreatment method for fusion model is established to provide a higher classification accuracy.(3)Research on Fusion Recognition Model Based on Texture and Spectral Feature Levels.In this paper,texture features and spectral features are combined to establish a feature classification model of wood species,and different classifiers are used for classification prediction.The results show that the extreme learning machine used in this paper is more effective in classification.At the same time,the fusion model,single texture model,and single spectral model are compared,and the results show that the feature level fusion model is more efficient than the single feature model in recognition and classification.This provides a new model for the identification of wood species.(4)Research on Decision-level Fusion Recognition Model Based on Texture and Spectrum.This paper mainly studies the texture and spectral characteristics of the fusion model,not only puts forward the feature level fusion,but also puts forward the decision level fusion model,experiments using decision-level fusion model of 21 kinds of timber to classify the highest accuracy can reach 99.05%,this shows that the fusion model in the field of timber species recognition is a research method.