Hyperspectral position target detection method based on EVM and deep learning

Abstract

The invention belongs to the field of hyperspectral intelligent perception, and discloses a hyperspectral position target detection method based on EVM and deep learning, and the method comprises thesteps: S1, carrying out the preprocessing of sample division and PCA of a hyperspectral data set, randomly setting a known class and an unknown class of a sample class, and dividing the data set intoa training set and a test set; S2, inputting the preprocessed test sample data into a 3DCNN network for training, storing a training model, and outputting a classification feature vector; S3, using anEVM algorithm to establish an EVM model based on Weibull distribution for the feature vectors of all the training data and the corresponding category labels; and S4, inputting the preprocessed test sample data into a 3DCNN model and an EVM model, and calculating the probability of belonging to a known class or an unknown class. The method is clear in structure and easy to implement, and effectively improves the detection precision of the unknown target on the premise of ensuring the classification precision of the known class.

Description

technical field [0001] The invention mainly relates to the field of hyperspectral intelligent perception, in particular to a hyperspectral unknown target detection method based on EVM and deep learning. Background technique [0002] A hyperspectral image is a three-dimensional spectral image with a spectral resolution in the range of 10l, including two-dimensional geometric space and one-dimensional spectral information of ground objects, realizing the integration of maps and spectra. Hyperspectral images have a spectral resolution of nanometers, which can accurately describe the reflection spectra of different ground objects, and effectively improve the ability to process hyperspectral image classification and recognition. Object classification and recognition is one of the research focuses of hyperspectral data processing. Hyperspectral image classification is based on the classification of remote sensing images, combined with the characteristics of hyperspectral images, ...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the problem of hyperspectral unknown target detection that needs to be solved and the shortcomings of the current proposed method, the present invention proposes a hyperspectral unknown target detection method based on EVM and 3DCNN

Images