Research On Hyperspectral Image Classification Algorithm Based On Fusion Model

With the rapid development of hyperspectral imaging technology and big data technology,the resolution of hyperspectral images is getting higher and higher,and the information contained is becoming more and more abundant.At present,hyperspectral images have been widely used in agriculture,forestry,medicine and other fields.Image classification is one of the most important methods to obtain image information.Therefore,hyperspectral image classification methods have high research value.Hyperspectral images have the characteristics of high data dimensionality and high correlation between data,which makes the classification of hyperspectral images face huge challenges.In recent years,the fusion model has shown good results in dealing with the problem of hyperspectral image classification.Therefore,this paper introduces multi-classifier fusion and feature grouping fusion algorithms,and proposes the following two hyperspectral image classification models:Aiming at the problems of large amount of hyperspectral image data,high data dimension,large data correlation and unsatisfactory classification effect of a single classifier,this paper proposes a hyperspectral image classification model based on the fusion of multiple classifiers.The model is composed of three heterogeneous base classifier models.First,the bilateral filtering algorithm is used for denoising.Secondly,the three base classifier models use linear discriminant analysis(LDA)algorithm combined with principal component analysis(PCA)algorithm,separate PCA algorithm,Gabor filter combined with PCA algorithm to reduce the dimensionality of the data.With feature extraction,three feature sets are obtained,and the three feature sets are respectively classified using Support Vector Machine(SVM)classifier,Light GBM classifier,and Ada Boost classifier.Finally,an AHP-voting method is proposed,which combines the results of three classifiers.Experimental verification shows that the model has improved classification effect due to the advantages of fusion of three classifiers.Aiming at the problem of insufficient feature extraction in the process of hyperspectral image classification,based on the multi-classifier fusion model,this paper proposes a hyperspectral image classification model based on feature grouping fusion.The model first improves the bilateral filter algorithm,and uses the improved bilateral filter algorithm and the PCA algorithm combination,the PCA algorithm and the LDA algorithm combination,the Gabor filter algorithm and the PCA algorithm combination to extract the features of the hyperspectral image data set,and obtain three feature sets.Secondly,a feature grouping fusion algorithm based on Bootstrap algorithm is proposed to fuse feature sets,and the decision tree algorithm is used to pre-classify the new feature sets,increasing the number of error-prone samples,and obtaining nine new feature sets.Finally,group the nine new feature sets into three groups,group them into the SVM classifier,the Light GBM classifier,and the Ada Boost classifier to classify,and use the AHP-voting method to fuse the classification results.Experiments have verified that the model achieves more effective feature extraction and improves the performance of the classification model through feature grouping and fusion.