Deep Learning Based Fossil-Fuel Power Plant Monitoring in High Resolution Remote Sensing Images: A Comparative Study
Abstract
:1. Introduction
2. Related Works
2.1. Deep Learning Based Object Detection
2.2. Deep Learning Based Object Detection in RSIs
2.3. Fossil-Fuel Power Plant Monitoring in RSIs
3. Deep Learning Based Fossil-Fuel Power Plant Monitoring
3.1. Deep Learning Models for Comparative Study
3.1.1. Faster R-CNN
3.1.2. FPN
3.1.3. R-FCN
3.1.4. DCN
3.1.5. SSD
3.1.6. DSSD
3.1.7. YOLO
3.1.8. RetinaNet
3.2. Implementation Details
3.2.1. Backbone Network
3.2.2. Training Details
4. Experimental Results
4.1. BUAA-FFPP60 Dataset
4.2. Index for Evaluation
4.3. Training Loss
4.4. Accuracy Comparison
4.5. Model Size and Memory Cost
4.6. Running Time
5. Discussion
6. Conclusions
Author Contributions
Funding
Conflicts of Interest
References
- He, X.; Xue, Y.; Guang, J.; Shi, Y.; Xu, H.; Cai, J.; Mei, L.; Li, C. The analysis of the haze event in the north china plain in 2013. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Quebec City, QC, Canada, 13–18 July 2014; pp. 5002–5005. [Google Scholar] [CrossRef]
- Feng, X.; Li, Q.; Zhu, Y. Bounding the role of domestic heating in haze pollution of bei**g based on satellite and surface observations. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Quebec City, QC, Canada, 13–18 July 2014; pp. 5727–5728. [Google Scholar] [CrossRef]
- Kai, Q.; Hu, M.; Lixin, W.; Rao, L.; Lang, H.; Wang, L.; Yang, B. Satellite remote sensing of aerosol optical depth, SO2 and NO2 over China’s Bei**g-Tian**-Hebei region during 2002–2013. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Bei**g, China, 10–15 July 2016; pp. 5727–5728. [Google Scholar] [CrossRef]
- Yao, Y.; Jiang, Z.; Zhang, H.; Cai, B.; Meng, G.; Zuo, D. Chimney and condensing tower detection based on faster R-CNN in high resolution remote sensing images. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Fort Worth, TX, USA, 23–28 July 2017; pp. 3329–3332. [Google Scholar] [CrossRef]
- Zhang, F.; Du, B.; Zhang, L.; Xu, M. Weakly supervised learning based on coupled convolutional neural networks for aircraft detection. IEEE Trans. Geosci. Remote Sens. 2016, 5553–5563. [Google Scholar] [CrossRef]
- Cai, B.; Jiang, Z.; Zhang, H.; Yao, Y.; Nie, S. Online Exemplar-Based Fully Convolutional Network for Aircraft Detection in Remote Sensing Images. IEEE Geosci. Remote Sens. Lett. 2018, 15, 1095–1099. [Google Scholar] [CrossRef]
- Zou, Z.; Shi, Z. Ship detection in spaceborne optical image with svd networks. IEEE Trans. Geosci. Remote Sens. 2016, 5832–5845. [Google Scholar] [CrossRef]
- Yao, Y.; Jiang, Z.; Zhang, H.; Zhao, D.; Cai, B. Ship detection in optical remote sensing images based on deep convolutional neural networks. J. Appl. Remote Sens. 2017, 11, 11–12. [Google Scholar] [CrossRef]
- Yang, X.; Sun, H.; Fu, K.; Yang, J.; Sun, X.; Yan, M.; Guo, Z. Automatic Ship Detection in Remote Sensing Images from Google Earth of Complex Scenes Based on Multiscale Rotation Dense Feature Pyramid Networks. Remote Sens. 2018, 10, 132. [Google Scholar] [CrossRef]
- Zhang, L.; Shi, Z.; Wu, J. A Hierarchical Oil Tank Detector With Deep Surrounding Features for High-Resolution Optical Satellite Imagery. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2015, 8, 4895–4909. [Google Scholar] [CrossRef]
- Yao, Y.; Jiang, Z.; Zhang, H. Oil tank detection based on salient region and geometric features. Proc. SPIE 2016, 92731G. [Google Scholar] [CrossRef]
- Wang, W.; Zhao, D.; Jiang, Z. Oil Tank Detection via Target-Driven Learning Saliency Model. In Proceedings of the 2017 4th IAPR Asian Conference on Pattern Recognition (ACPR), Nan**g, China, 26–29 November 2017; pp. 156–161. [Google Scholar] [CrossRef]
- Zhu, X.X.; Tuia, D.; Mou, L.; ** Sites Associated with Slavery from Space. Remote Sens. 2019, 11, 266. [Google Scholar] [CrossRef]
- He, K.; Zhang, X.; Ren, S.; Sun, J. Deep Residual Learning for Image Recognition. In Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 26 June–1 July 2016; pp. 770–778. [Google Scholar] [CrossRef]
- Kingma, D.P.; Ba, J. Adam: A Method for Stochastic Optimization. In Proceedings of the ICLR 2014, Banff, AB, Canada, 14–16 April 2014. [Google Scholar]
Model | Backbone | Category | Application in Remote Sensing |
---|---|---|---|
Faster R-CNN [18] | ResNet-101 | two-stage | [4,8,38] |
FPN [22] | ResNet-101 | two-stage | [9] |
R-FCN [23] | ResNet-101 | two-stage | [6,32] |
DCN [24] | ResNet-101 | two-stage | [33] |
SSD [26] | ResNet-101 | one-stage | [34] |
DSSD [28] | ResNet-101 | one-stage | [35] |
YOLOv3 [30] | Darknet-53 | one-stage | [36] |
RetinaNet [31] | ResNet-101 | one-stage | [37] |
Models | Warm-up Step | Train Step | Initial lr | lr Schedule | Batch Size |
---|---|---|---|---|---|
Faster R-CNN [18] | 0 | 72000 | STEP | 1 | |
FPN [22] | 0 | 60000 | STEP | 1 | |
R-FCN [23] | 0 | 60000 | STEP | 1 | |
DCN [24] | 1000 | 60000 | STEP | 1 | |
SSD [26] | 200 | 5600 | EXPONENTIAL | 24 | |
DSSD [28] | 0 | 30000 | STEP | 8 | |
YOLOv3 [30] | 500 | 30000 | ADAM | 6 | |
RetinaNet [31] | 0 | 30000 | ADAM | 2 |
Class | Chimney | Condensing Tower | RSI | ||
---|---|---|---|---|---|
Working | Not Working | Working | Not Working | ||
Number in training set (augmented + original) | 198 | 435 | 408 | 426 | 861 |
Number in testing set (original) | 21 | 36 | 65 | 28 | 31 |
Total number (augmented + original) | 219 | 471 | 473 | 454 | 892 |
Model | Chimney | Condensing Tower | ||
---|---|---|---|---|
Precision | Recall | Precision | Recall | |
Faster R-CNN [18] | 0.7342 | 0.8681 | 0.9293 | 0.9683 |
FPN [22] | 0.7361 | 0.8526 | 0.9525 | 0.9711 |
R-FCN [23] | 0.7022 | 0.8092 | 0.9223 | 0.9576 |
DCN [24] | 0.6469 | 0.7813 | 0.8673 | 0.9287 |
SSD [26] | 0.6557 | 0.7861 | 0.8661 | 0.9478 |
DSSD [28] | 0.6432 | 0.7627 | 0.8593 | 0.9341 |
YOLOv3 [30] | 0.6751 | 0.7892 | 0.8723 | 0.9583 |
RetinaNet [31] | 0.7254 | 0.8461 | 0.9428 | 0.9731 |
Yao et al. [4] | 0.6710 | 0.8390 | 0.9080 | 0.9570 |
Model | AP of Chimney | AP of Condensing Tower | mAP | ||
---|---|---|---|---|---|
Working | Not Working | Working | Not Working | ||
Faster R-CNN [18] | 0.6979 | 0.6602 | 0.9283 | 0.9648 | 0.8128 |
FPN [22] | 0.6878 | 0.7239 | 0.9245 | 0.9730 | 0.8273 |
R-FCN [23] | 0.5952 | 0.6786 | 0.9144 | 0.9502 | 0.7846 |
DCN [24] | 0.5426 | 0.5696 | 0.8672 | 0.9152 | 0.7234 |
SSD [26] | 0.5323 | 0.6031 | 0.8964 | 0.9510 | 0.7457 |
DSSD [28] | 0.5336 | 0.6042 | 0.8877 | 0.9149 | 0.7376 |
YOLOv3 [30] | 0.5532 | 0.5851 | 0.8765 | 0.9396 | 0.7386 |
RetinaNet [31] | 0.6564 | 0.6725 | 0.9308 | 0.9603 | 0.8075 |
Model | Space Occupancy | Memory Cost |
---|---|---|
Faster R-CNN [21] | 425.6 MB | 4.61 GB |
FPN [22] | 464.2 MB | 4.36 GB |
R-FCN [23] | 443.1 MB | 3.60 GB |
DCN [24] | 494.6 MB | 0.95 GB |
SSD [26] | 218.7 MB | 0.86 GB |
DSSD [28] | 246.6 MB | 1.76 GB |
YOLOv3 [30] | 235.6 MB | 2.61 GB |
RetinaNet [31] | 635.1 MB | 3.79 GB |
© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Zhang, H.; Deng, Q. Deep Learning Based Fossil-Fuel Power Plant Monitoring in High Resolution Remote Sensing Images: A Comparative Study. Remote Sens. 2019, 11, 1117. https://doi.org/10.3390/rs11091117
Zhang H, Deng Q. Deep Learning Based Fossil-Fuel Power Plant Monitoring in High Resolution Remote Sensing Images: A Comparative Study. Remote Sensing. 2019; 11(9):1117. https://doi.org/10.3390/rs11091117
Chicago/Turabian StyleZhang, Haopeng, and Qin Deng. 2019. "Deep Learning Based Fossil-Fuel Power Plant Monitoring in High Resolution Remote Sensing Images: A Comparative Study" Remote Sensing 11, no. 9: 1117. https://doi.org/10.3390/rs11091117