.png){kind=icon}
Return
Hyperspectral image classification by sparse tensor based support tensor machine
.png){kind=icon}
PRE
AI
DOI：10.1016/j.infrared.2024.105446.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Spectral vector based Hyperspectral Image (HSI) classification algorithms are difficult to effectively and completely utilize the rich spatial and spectral information contained in HSI, leading to the limited classification accuracy. This paper presents a sparse tensor based Support Tensor Machine (STM) HSI classification algorithm to solve this problem. Firstly, the spatial-spectral tensor for each pixel in a given HSI is generated by combining spectral vectors of the pixel and its neighboring pixels to express its spatial and spectral information, which is used as basic processing unit for HSI classification. Then, the sparse tensors taken as classification features for pixels are obtained directly from the spatial-spectral tensors by dimension reduction in the tensor framework. Finally, the STM classifier is designed to distinguish sparse tensors, effectively preserving the structural information of the tensor. Experiments with three real HSIs are conducted to validate the effectiveness of the Tensor Sparse Representation based STM (TSR-STM) algorithm. The experimental results show that the TSR-STM algorithm achieves better performance compared to three other classification algorithms.
Keywords:
Hyperspectral image classification
sparse representation (SR)
Sparse tensor
Spatial-spectral tensor
support tensor machine (STM)
Journal
I
IF:
3.4
Papers:
5.8K
Citations:
1.2W
Organization
Cited Papers
ShareSaveShareSaveLocal adaptive joint sparse representation for hyperspectral image classification.png)
NEUROCOMPUTING
NEUROCOMPUTINGIF6.5
PREAI
ShareSave
View more