Hyperspectral Image Spectral And Texture Feature Extraction Based On Improved Binary Ant Colony Algorithm

As the gradual development of sensor hardware,the application of hyperspectral remote sensing images has become increasingly common.As an important source of information for hyperspectral remote sensing images,spectral and texture features in hyperspectral remote sensing images have always been a research hotspot in remote sensing and surveying mapping.At the same time,hyperspectral remote sensing images have many problems,such as the number of bands,which can easily cause dimensional disasters.Therefore,it is of great significance to study and explore the feature extraction methods for hyperspectral remote sensing images.Mathematical and spatial geometrical features of the imagery were deeply researched around the target of extracting the features of hyperspectral remote sensing imagery.Focusing on the extraction of spectral features,texture feature extraction methods,and feature fusion classification methods were researched,experimented,and innovations were carried out.As a result,a spectral and texture feature extraction scheme based on an improved binary ant colony algorithm is developed for hyperspectral remote sensing images.The main work and results of the paper are as follows:(1)Aiming at the problem that there are many bands in hyperspectral remote sensing image and it is easy to generate redundant features,a band selection method based on improved binary ant colony algorithm is proposed.The band selection method such as basic ant colony algorithm,basic binary ant colony algorithm and genetic algorithm are studied,including its algorithm principle,related concepts and the existing problems.In the improved binary ant colony algorithm,the results of genetic algorithm are utilized as the initial heuristic information of binary ant colony algorithm to solve the random initial generation problem.To solve the problem that ant colony algorithm easily falls into local optimal solution,the ant path selection mechanism is improved to enhance the global optimization ability.Experimental results show that the method has better global search ability and can effectively extract spectral features.(2)In order to solve the problem of low efficiency of traditional Gabor filter,an improved texture feature extraction method of 2-D Gabor filter is proposed.On the basis of the 2-D complex Gabor transform,the corresponding function decomposition is improved,and the intermediate results in the running process are reused to reduce the operation time.The experimental results show that the efficiency of this method has been improved to some extent,and the texture features extracted by this method have good local and edge features,which can improve the classification accuracy of the image.(3)In order to further improve the classification accuracy of hyperspectral remote sensing images,a classification scheme combining spectral and texture features is proposed.In order to reduce the feature redundancy,the scheme firstly uses the improved binary ant colony algorithm to filter the extracted texture features corresponding to each band;then,the selected spectral features and texture features are integrated at the feature level;finally,the fused features are integrated to the lib-SVM classifier for classification.Through different data and different algorithms for verification,the experiment results verify the effectiveness of the improved binary ant colony algorithm,and the features extracted by this scheme have high classification accuracy and are suitable for hyperspectral remote sensing images.