Hyperspectral ground feature classification method and system of lightweight dynamic fusion convolutional network

Abstract

The invention discloses a hyperspectral ground feature classification method and system of a lightweight dynamic fusion convolutional network. The method comprises the following steps: carrying out the normalization processing of a hyperspectral image, and dividing the hyperspectral image into a training set, a verification set and a test set; setting a dual-path interconnection feature extraction module and a classification module, and constructing a network model; training a network model by using the training set, verifying the network model by using the verification set, and taking the weight of the network model of the first generation with the highest precision on the verification set as the final classification model weight; and inputting the test set into the network model of the first generation with the highest precision for testing to obtain a final classification result, performing hyperspectral image classification, and outputting a classification image according to the classification result. The network parameter quantity is further reduced, the network training time is shortened, and a better classification effect is obtained under the condition of fewer training samples.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a method and a system for classifying hyperspectral ground objects using a lightweight dynamic fusion convolutional network. Background technique [0002] Due to the unique nature of one object, one spectrum, hyperspectral images have been widely used in land use classification, military monitoring, object recognition and other fields. In recent years, with the continuous development and wide application of hyperspectral image acquisition equipment, the demand for hyperspectral image processing has also increased. Hyperspectral images have the characteristics of large data volume, multiple spectral dimensions, and high spectral similarity, which pose a great challenge to hyperspectral image classification. As deep learning has shown excellent results in traditional image classification tasks, the introduction of deep learning into hyperspectral image classif...

AI Technical Summary

Problems solved by technology

[0006] Although many lightweight studies have been applied to the field of hyperspectral image classification recently, the number of model parameters is still on the order of hundreds of thousands, and a large number of training samples is still required to achieve good classification accuracy

Images