Hyperspectral band selection algorithm based on artificial bee colony fusion genetic idea

Hyperspectral images occupy a very important position when it comes to remote sensing imaging due to their multi band and high precision characteristics. However, due to the redundancy caused by various bands and the large amount of calculation, Reducing the size of hyperspectral photographs is important. This article proposes a program that combines cross genetic thinking with artificial bee colony algorithm (hereinafter referred to as Ag) to select the features of hyperspectral images, In the preprocessing stage, the initial feature set of the image is screened by using the outlier detection method based on the percentage level, and the feature set after the initial screening is selected by Ag algorithm. Finally, the selected band combination is sent to the neural network combining three-dimensional convolution and two-dimensional convolution for training, Experimental results on the Pavia University and Indian Pines datasets show that the algorithm has good convergence, and the band combination obtained with low computational complexity can achieve better classification accuracy and consistency in image classification.