PolSAR Image Classification with Lightweight 3D Convolutional Networks
Convolutional neural networks (CNNs) have become the state-of-the-art in optical image processing. Recently, CNNs have been used in polarimetric synthetic aperture radar (PolSAR) image classification and obtained promising results. Unlike optical images, the unique phase information of PolSAR data e...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-01-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/12/3/396 |
| _version_ | 1798018560419692544 |
|---|---|
| author | Hongwei Dong Lamei Zhang Bin Zou |
| author_facet | Hongwei Dong Lamei Zhang Bin Zou |
| author_sort | Hongwei Dong |
| collection | DOAJ |
| description | Convolutional neural networks (CNNs) have become the state-of-the-art in optical image processing. Recently, CNNs have been used in polarimetric synthetic aperture radar (PolSAR) image classification and obtained promising results. Unlike optical images, the unique phase information of PolSAR data expresses the structure information of objects. This special data representation makes 3D convolution which explicitly modeling the relationship between polarimetric channels perform better in the task of PolSAR image classification. However, the development of deep 3D-CNNs will cause a huge number of model parameters and expensive computational costs, which not only leads to the decrease of the interpretation speed during testing, but also greatly increases the risk of over-fitting. To alleviate this problem, a lightweight 3D-CNN framework that compresses 3D-CNNs from two aspects is proposed in this paper. Lightweight convolution operations, i.e., pseudo-3D and 3D-depthwise separable convolutions, are considered as low-latency replacements for vanilla 3D convolution. Further, fully connected layers are replaced by global average pooling to reduce the number of model parameters so as to save the memory. Under the specific classification task, the proposed methods can reduce up to 69.83% of the model parameters in convolution layers of the 3D-CNN as well as almost all the model parameters in fully connected layers, which ensures the fast PolSAR interpretation. Experiments on three PolSAR benchmark datasets, i.e., AIRSAR Flevoland, ESAR Oberpfaffenhofen, EMISAR Foulum, show that the proposed lightweight architectures can not only maintain but also slightly improve the accuracy under various criteria. |
| first_indexed | 2024-04-11T16:26:11Z |
| format | Article |
| id | doaj.art-dcbbcdbdc60b464ab7820c95faa25ac9 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-04-11T16:26:11Z |
| publishDate | 2020-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-dcbbcdbdc60b464ab7820c95faa25ac92022-12-22T04:14:11ZengMDPI AGRemote Sensing2072-42922020-01-0112339610.3390/rs12030396rs12030396PolSAR Image Classification with Lightweight 3D Convolutional NetworksHongwei Dong0Lamei Zhang1Bin Zou2Department of Information Engineering, Harbin Institute of Technology, Harbin 150001, ChinaDepartment of Information Engineering, Harbin Institute of Technology, Harbin 150001, ChinaDepartment of Information Engineering, Harbin Institute of Technology, Harbin 150001, ChinaConvolutional neural networks (CNNs) have become the state-of-the-art in optical image processing. Recently, CNNs have been used in polarimetric synthetic aperture radar (PolSAR) image classification and obtained promising results. Unlike optical images, the unique phase information of PolSAR data expresses the structure information of objects. This special data representation makes 3D convolution which explicitly modeling the relationship between polarimetric channels perform better in the task of PolSAR image classification. However, the development of deep 3D-CNNs will cause a huge number of model parameters and expensive computational costs, which not only leads to the decrease of the interpretation speed during testing, but also greatly increases the risk of over-fitting. To alleviate this problem, a lightweight 3D-CNN framework that compresses 3D-CNNs from two aspects is proposed in this paper. Lightweight convolution operations, i.e., pseudo-3D and 3D-depthwise separable convolutions, are considered as low-latency replacements for vanilla 3D convolution. Further, fully connected layers are replaced by global average pooling to reduce the number of model parameters so as to save the memory. Under the specific classification task, the proposed methods can reduce up to 69.83% of the model parameters in convolution layers of the 3D-CNN as well as almost all the model parameters in fully connected layers, which ensures the fast PolSAR interpretation. Experiments on three PolSAR benchmark datasets, i.e., AIRSAR Flevoland, ESAR Oberpfaffenhofen, EMISAR Foulum, show that the proposed lightweight architectures can not only maintain but also slightly improve the accuracy under various criteria.https://www.mdpi.com/2072-4292/12/3/396deep learningpolarimetric synthetic aperture radar (polsar) classification3d convolutionpseudo-3d convolutiondepthwise separable convolution |
| spellingShingle | Hongwei Dong Lamei Zhang Bin Zou PolSAR Image Classification with Lightweight 3D Convolutional Networks Remote Sensing deep learning polarimetric synthetic aperture radar (polsar) classification 3d convolution pseudo-3d convolution depthwise separable convolution |
| title | PolSAR Image Classification with Lightweight 3D Convolutional Networks |
| title_full | PolSAR Image Classification with Lightweight 3D Convolutional Networks |
| title_fullStr | PolSAR Image Classification with Lightweight 3D Convolutional Networks |
| title_full_unstemmed | PolSAR Image Classification with Lightweight 3D Convolutional Networks |
| title_short | PolSAR Image Classification with Lightweight 3D Convolutional Networks |
| title_sort | polsar image classification with lightweight 3d convolutional networks |
| topic | deep learning polarimetric synthetic aperture radar (polsar) classification 3d convolution pseudo-3d convolution depthwise separable convolution |
| url | https://www.mdpi.com/2072-4292/12/3/396 |
| work_keys_str_mv | AT hongweidong polsarimageclassificationwithlightweight3dconvolutionalnetworks AT lameizhang polsarimageclassificationwithlightweight3dconvolutionalnetworks AT binzou polsarimageclassificationwithlightweight3dconvolutionalnetworks |