Hyperspectral Object Recognition Based on Optimized Gabor Spatial-spectral Feature

A hyperspectral image (HSI) feature recognition method based on Gabor filter bank parameters optimization is proposed. The spatial feature parameters of hyperspectral remote sensing features are optimized by the genetic algorithm to achieve accurate characterization and interpretation of features. On the basis of initialization, the Gabor filter parameters are updated and optimized through the genetic algorithm in genetic space, so as to achieve adaptive extraction of spatial features. The experiment results show that the classification effect of this method is better than that of the fixed filter classification algorithm. The overall accuracies of Indian Pines, Salinas, and Tea Farm hyperspectral datasets are improved by 0.65%, 1.61% and 2.92%. Thus, we can achieve a greater degree in spatial-spectral information mining.