International Journal of New Developments in Engineering and Society, 2024, 8(3); doi: 10.25236/IJNDES.2024.080306.
Haoyu Guan
St. Stephen’s Episcopal School, Texas, The United States of America
Wildfire is a fire disaster in the forests or on the grassland of the wasteland area, potentially destroying urban and rural properties, polluting the air, and causing global warming. The subsequent fire is more likely to appear in vulnerable areas, so the images after the last fire become the images before the next one. Thus, detecting wildfire and its affected area is significant for evaluating its ecological consequence and spatial pattern. The application background of this study is to delineate the degree of impact accurately and to provide valuable information for disaster response. This study develops a self-response system based on high-resolution multispectral imagery (MSI) from the satellite Sentinel-2, adopting an advanced deep-learning approach to detect the change of burned areas. This research proposes SimCLR using contrastive learning techniques to extract features with unlabeled models. Based on SimCLR, we create our framework, FireCLR, and apply a simple CNN to capture and compare features without artificial supervision to learn and mark the wildfire on the satellite maps. By substituting the simple CNN with more advanced ResNet and VGG networks, we further improve the accuracy and ability of wildfire delineation and analysis. Our experiment proves that the model performs well in judging the wildfire conditions and vulnerable areas with given satellite images. With further studies and extensions, this model will contribute to the worldwide detection and prediction of various natural disasters.
Wildfire, SimCLR, CNN, FireCLR, VGG, Disasters
Haoyu Guan. Unsupervised Wildfire Change Detection Based on Contrastive Learning. International Journal of New Developments in Engineering and Society (2024) Vol.8, Issue 3: 29-38. https://doi.org/10.25236/IJNDES.2024.080306.
[1] Simmerman, A. (2024, March 1). Wildfires burn more than one million acres in Texas panhandle. track wildfires in real-time. USA Today. https://www.usatoday.com/story/news/state/2024/02/27/texas-wildfire-map-panhandle-smokehouse-creek-windy-deuce-grape-vine-creek-juliet-pass/72756264007/
[2] Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278–2324. https://doi.org/10.1109/5.726791
[3] Simonyan, K. (2014, September 4). Very Deep Convolutional Networks for Large-Scale Image Recognition. arXiv.org. https://doi.org/10.48550/arXiv.1409.1556
[4] He, K. (2015, December 10). Deep Residual Learning for Image Recognition. arXiv.org. https://doi. org/ 10.48550/arXiv.1512.03385
[5] Hu, X., Ban, Y., & Nascetti, A. (2021, April 14). Uni-Temporal Multispectral Imagery for Burned Area Mapping with Deep Learning. Remote Sensing, 13(8), 1509. https://doi.org/10.3390/rs13081509
[6] McCarthy, N. F., Tohidi, A., Aziz, Y., Dennie, M., Valero, M. M., & Hu, N. (2021, March 3). A Deep Learning Approach to Downscale Geostationary Satellite Imagery for Decision Support in High Impact Wildfires. Forests, 12(3), 294. https://doi.org/10.3390/f12030294
[7] Monaco, S., Pasini, A., Apiletti, D., Colomba, L., Farasin, A., Garza, P., & Baralis, E. (2021, March). Double-Step deep learning framework to improve wildfire severity classification. In EDBT/ICDT Workshops.
[8] Farasin, A., Colomba, L., & Garza, P. (2020, June 24). Double-Step U-Net: A Deep Learning-Based Approach for the Estimation of Wildfire Damage Severity through Sentinel-2 Satellite Data. Applied Sciences, 10(12), 4332. https://doi.org/10.3390/app10124332
[9] Alwan, Y., & Celenk, M. (2020, February). Using combined linear regression and principal component analysis for unsupervised change detection of forest fire. In 2020 6th International Engineering Conference “Sustainable Technology and Development"(IEC) (pp. 152-156). IEEE.
[10] Chung, M., Han, Y., & Kim, Y. (2020, November 22). A Framework for Unsupervised Wildfire Damage Assessment Using VHR Satellite Images with PlanetScope Data. Remote Sensing, 12(22), 3835. https://doi.org/10.3390/rs12223835
[11] Farasin, A., Nini, G., Garza, P., & Rossi, C. (2019). Unsupervised Burned Area Estimation through Satellite Tiles: A Multimodal Approach by Means of Image Segmentation over Remote Sensing Imagery. In MACLEAN@ PKDD/ECML.
[12] Zhang, B., Wang, H., Alabri, A., Bot, K., McCall, C., Hamilton, D., & Růžička, V. (2022). Unsupervised Wildfire Change Detection based on Contrastive Learning. arXiv preprint arXiv:2211.14654.