Hyperspectral remote sensing image dimensionality reduction method based on adaptive filtering

Abstract

With the rapid development of hyperspectral image technology, remote sensing technology has ushered in an innovation in theory and application, and hyperspectral remote sensing images have come into being. However, due to its high data dimensionality, it is difficult for statistical classifiers to work on it, making the technology face development difficulties. Therefore, how to effectively reduce the dimensionality of hyperspectral remote sensing images has gradually become a research hotspot in this field. The study clusters bands by K-means algorithm, and then combines the least mean square algorithm in adaptive filtering and recursive least squares method, and uses this as the basis for band selection. Finally, the dimension reduction effect is verified. The experimental results show that the improved band selection method achieves an overall accuracy of over 80% and 90% in the hyperspectral datasets of Pavia University and Idian Pine respectively, with the Kappa coefficient reaching 0.9. In the overall dimensionality reduction classification of the Indianan data, the accuracy also reaches 90% and can be maintained consistently, indicating that the method has high accuracy and can effectively reduce the dimensionality of hyperspectral remote sensing images.