Abstract
Recently, deep learning-based algorithms have been widely used for classification of hyperspectral images (HSIs) by extracting invariant and abstract features. In our conference paper presented at IEEE International Geoscience and Remote Sensing Symposium 2018, 1-D-capsule network (CapsNet) and 2-D-CapsNet were proposed and validated for HSI feature extraction and classification. To further improve the classification performance, the robust 3-D-CapsNet architecture is proposed in this article by following our previous work, which introduces the maximum correntropy criterion to address the noise and outliers problem, generating a robust and better generalization model. As such, discriminative features can be extracted even if some samples are corrupted more or less. In addition, a novel dual channel framework based on robust CapsNet is further proposed to fuse the hyperspectral data and light detection and ranging-derived elevation data for classification. Three widely used hyperspectral datasets are employed to demonstrate the superiority of our proposed deep learning models.
Highlights
W ITH advanced technologies on sensors and imaging systems, a hyperspectral image (HSI) contains abundant spectral information in hundreds of narrow and contiguous bands, and presents rich contextual structure information of imaged scenes
In our previous conference paper [58], we proposed the 1-DCapsNet and 2-D-capsule network (CapsNet) models for HSI feature extraction (FE) and classification
A robust 3-D-CapsNet architecture has been proposed for HSI classification, which introduces the maximum correntropy criterion (MCC) to address the noise and outliers problem, generating a robust and strong generalization model
Summary
W ITH advanced technologies on sensors and imaging systems, a hyperspectral image (HSI) contains abundant spectral information in hundreds of narrow and contiguous bands, and presents rich contextual structure information of imaged scenes. In order to improve the discriminative ability of extracted features, a supervised deep CNN architecture containing five layers was developed for HSI classification [40]. Following this work and considering the aforementioned drawbacks of the existing deep models, the maximum correntropy criterion (MCC)-based robust 3-D-CapsNet architecture is proposed in this article. To the best of our knowledge, it is the first time that the MCC is used in CapsNet for addressing the noise and outlier problem in HSIs. 2) A novel MCC-based dual channel robust CapsNet framework is proposed to fuse multisource remote sensing data, e.g., hyperspectral data and LiDAR data, in which the spatial–spectral information of HSI and the elevation information of LiDAR can be efficiently fused to extract more discriminative features for the classification.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
