Research on the Extraction Technology of Earth Remote Sensing Images Based on Computer BP Neural Network

Abstract

Due to its huge number of bands, hyperspectral remote sensing images directly lead to high redundancy of information and complex data processing, which not only brings a huge amount of calculation, but also damages the classification accuracy. Therefore, dimensionality reduction becomes necessary before processing and analyzing hyperspectral images. Neural network sensitivity analysis can be used to simplify the dimensionality reduction of the model. In this paper, the method is applied to the dimensionality reduction of hyperspectral remote sensing images, and the correlation between bands is weakened by subspace division. The experimental results show that the overall classification accuracy of BP neural network is 87.85%, and the Kappa coefficient is 0.84, which are 5.53 percentage points and 0.07 higher than the minimum distance method classification respectively. Experiments show that the BP neural network classification method is an effective and more accurate classification method.