A Hyperspectral Open Set Classification Method Based on Euclidean Distance and Deep Learning

Abstract

The invention belongs to the technical field of image processing, and discloses a hyperspectral open set classification method based on Euclidean distance and deep learning. The present invention first constructs the category center of each known class, the category prediction function based on Euclidean distance and the loss function based on Euclidean distance, and trains and optimizes the deep learning network model; secondly, combines the boxplot method and the Weibull distribution model of extreme value theory Combined, analyze and fit the distance between the input data and the category center of each known class, realize the constraints on each known classification boundary, and then realize the discrimination of unknown classes: the box plot is used to provide the fitting The number of abnormal points required by the Weibull distribution model, and according to the number of abnormal points, it is judged whether to use the Weibull model or use the upper edge of the boxplot to discriminate the unknown class. The invention aims at the open set classification problem in the hyperspectral field, and is a simple, practical method with high classification accuracy and strong robustness.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a hyperspectral open set classification method based on Euclidean distance and deep learning. In the testing process, the invention can not only accurately classify the known classes appearing in the training process, but also identify and reject the unknown classes not appearing in the training process. Background technique [0002] Hyperspectral classification technology is a technology that uses spectral information and spatial information to classify each pixel in hyperspectral images, and has great application value in national defense and civilian fields. In recent years, with the rise of deep learning technology, classification models based on convolutional neural networks and their derivative models (such as residual networks, recurrent networks, and densely connected networks) have gradually become important methods in the field of hyperspectral classification. ...

AI Technical Summary

Problems solved by technology

Therefore, in the actual use process, the robustness and generalization of comparison file 2 and comparison file 3 are poor, and the performance of unknown target detection fluctuates greatly

[0006] To sum up, the challenge of hyperspectral open set classification tasks still needs to be solved; secondly, the existing hyperspectral open set classification methods have the problems of difficult parameter adjustment and low accuracy, which need to be improved through further research Or propose a new open set classification method

Images