|
| Title | Year | Methods | Datasets | Limitations |
|
| Integrating multilayer features of convolutional neural networks for remote sensing scene classification [127] | 2017 | Fisher kernel coding | WHU_RS dataset, UCM dataset | To reduce computational time and to improve classification accuracy. |
|
| A deep-local-global feature fusion framework for high spatial resolution imagery scene classification [126] | 2018 | Deep-local-global feature fusion | 21-class UC Merced dataset, 12-class Google dataset of SIRI-WHU | Same process could be implemented on non-optical images. |
|
| Remote sensing scene classification using multilayer stacked covariance pooling [129] | 2018 | Multilayer stacked covariance pooling (MSCP) | UC Merced land use dataset, AID30, NWPU-RESISC45 dataset | There should be an end to end CNN model which is able to classify with better accuracy using lesser features maps at each layer. |
|
| Remote sensing image scene classification using rearranged local features [128] | 2019 | Global and rearranged local features | UC Merced, WHU-RS19, Sydney, AID | To explore more techniques for feature fusion. |
|
| A feature aggregation convolutional neural network for remote sensing scene classification [130] | 2021 | Feature aggregation convolutional neural network (FACNN) | AID, UC Merced, WHU-RS19 | There should be a technique that can get semantic information of images without cropping or resizing of images. |
|
