Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery
This study compares some different types of spectral domain transformations for convolutional neural network (CNN)-based land cover classification. A novel approach was proposed, which transforms one-dimensional (1-D) spectral vectors into two-dimensional (2-D) features: Polygon graph images (CNN-Po...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-03-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/12/7/1097 |
| _version_ | 1827719626321035264 |
|---|---|
| author | Junghee Lee Daehyeon Han Minso Shin Jungho Im Junghye Lee Lindi J. Quackenbush |
| author_facet | Junghee Lee Daehyeon Han Minso Shin Jungho Im Junghye Lee Lindi J. Quackenbush |
| author_sort | Junghee Lee |
| collection | DOAJ |
| description | This study compares some different types of spectral domain transformations for convolutional neural network (CNN)-based land cover classification. A novel approach was proposed, which transforms one-dimensional (1-D) spectral vectors into two-dimensional (2-D) features: Polygon graph images (CNN-Polygon) and 2-D matrices (CNN-Matrix). The motivations of this study are that (1) the shape of the converted 2-D images is more intuitive for human eyes to interpret when compared to 1-D spectral input; and (2) CNNs are highly specialized and may be able to similarly utilize this information for land cover classification. Four seasonal Landsat 8 images over three study areas—Lake Tapps, Washington, Concord, New Hampshire, USA, and Gwangju, Korea—were used to evaluate the proposed approach for nine land cover classes compared to several other methods: Random forest (RF), support vector machine (SVM), 1-D CNN, and patch-based CNN. Oversampling and undersampling approaches were conducted to examine the effect of the sample size on the model performance. The CNN-Polygon had better performance than the other methods, with overall accuracies of about 93%–95 % for both Concord and Lake Tapps and 80%–84% for Gwangju. The CNN-Polygon particularly performed well when the training sample size was small, less than 200 per class, while the CNN-Matrix resulted in similar or higher performance as sample sizes became larger. The contributing input variables to the models were carefully analyzed through sensitivity analysis based on occlusion maps and accuracy decreases. Our result showed that a more visually intuitive representation of input features for CNN-based classification models yielded higher performance, especially when the training sample size was small. This implies that the proposed graph-based CNNs would be useful for land cover classification where reference data are limited. |
| first_indexed | 2024-03-10T20:47:44Z |
| format | Article |
| id | doaj.art-482589cd9fa847f9987290c67641336a |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T20:47:44Z |
| publishDate | 2020-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-482589cd9fa847f9987290c67641336a2023-11-19T20:10:18ZengMDPI AGRemote Sensing2072-42922020-03-01127109710.3390/rs12071097Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite ImageryJunghee Lee0Daehyeon Han1Minso Shin2Jungho Im3Junghye Lee4Lindi J. Quackenbush5School of Urban and Environmental Engineering, Ulsan National Institute of Science and Technology (UNIST), Ulsan 44919, KoreaSchool of Urban and Environmental Engineering, Ulsan National Institute of Science and Technology (UNIST), Ulsan 44919, KoreaSchool of Urban and Environmental Engineering, Ulsan National Institute of Science and Technology (UNIST), Ulsan 44919, KoreaSchool of Urban and Environmental Engineering, Ulsan National Institute of Science and Technology (UNIST), Ulsan 44919, KoreaSchool of Management Engineering, UNIST, Ulsan 44949, KoreaDepartment of Environmental Resources Engineering, State University of New York, College of Environmental Science and Forestry, Syracuse, NY 13210, USAThis study compares some different types of spectral domain transformations for convolutional neural network (CNN)-based land cover classification. A novel approach was proposed, which transforms one-dimensional (1-D) spectral vectors into two-dimensional (2-D) features: Polygon graph images (CNN-Polygon) and 2-D matrices (CNN-Matrix). The motivations of this study are that (1) the shape of the converted 2-D images is more intuitive for human eyes to interpret when compared to 1-D spectral input; and (2) CNNs are highly specialized and may be able to similarly utilize this information for land cover classification. Four seasonal Landsat 8 images over three study areas—Lake Tapps, Washington, Concord, New Hampshire, USA, and Gwangju, Korea—were used to evaluate the proposed approach for nine land cover classes compared to several other methods: Random forest (RF), support vector machine (SVM), 1-D CNN, and patch-based CNN. Oversampling and undersampling approaches were conducted to examine the effect of the sample size on the model performance. The CNN-Polygon had better performance than the other methods, with overall accuracies of about 93%–95 % for both Concord and Lake Tapps and 80%–84% for Gwangju. The CNN-Polygon particularly performed well when the training sample size was small, less than 200 per class, while the CNN-Matrix resulted in similar or higher performance as sample sizes became larger. The contributing input variables to the models were carefully analyzed through sensitivity analysis based on occlusion maps and accuracy decreases. Our result showed that a more visually intuitive representation of input features for CNN-based classification models yielded higher performance, especially when the training sample size was small. This implies that the proposed graph-based CNNs would be useful for land cover classification where reference data are limited.https://www.mdpi.com/2072-4292/12/7/1097spectral curve transformationconvolutional neural networksensitivity analysisland cover classification |
| spellingShingle | Junghee Lee Daehyeon Han Minso Shin Jungho Im Junghye Lee Lindi J. Quackenbush Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery Remote Sensing spectral curve transformation convolutional neural network sensitivity analysis land cover classification |
| title | Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery |
| title_full | Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery |
| title_fullStr | Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery |
| title_full_unstemmed | Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery |
| title_short | Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery |
| title_sort | different spectral domain transformation for land cover classification using convolutional neural networks with multi temporal satellite imagery |
| topic | spectral curve transformation convolutional neural network sensitivity analysis land cover classification |
| url | https://www.mdpi.com/2072-4292/12/7/1097 |
| work_keys_str_mv | AT jungheelee differentspectraldomaintransformationforlandcoverclassificationusingconvolutionalneuralnetworkswithmultitemporalsatelliteimagery AT daehyeonhan differentspectraldomaintransformationforlandcoverclassificationusingconvolutionalneuralnetworkswithmultitemporalsatelliteimagery AT minsoshin differentspectraldomaintransformationforlandcoverclassificationusingconvolutionalneuralnetworkswithmultitemporalsatelliteimagery AT junghoim differentspectraldomaintransformationforlandcoverclassificationusingconvolutionalneuralnetworkswithmultitemporalsatelliteimagery AT junghyelee differentspectraldomaintransformationforlandcoverclassificationusingconvolutionalneuralnetworkswithmultitemporalsatelliteimagery AT lindijquackenbush differentspectraldomaintransformationforlandcoverclassificationusingconvolutionalneuralnetworkswithmultitemporalsatelliteimagery |