Waterbody detection from satellite SAR images using deep learning
Document Type
Oral Presentation
Campus where you would like to present
Ellensburg
Event Website
https://digitalcommons.cwu.edu/source
Start Date
18-5-2020
Abstract
Nowadays, Synthetic Aperture Radar (SAR) images have been widely used in the industry and the scientific community for different remote sensing applications. The main advantage of SAR technology is that it can acquire images from night time since it does not require sunlight. Additionally, it can penetrate the cloud which can capture images where the traditional optical sensor is limited. One of the remarkable applications of SAR image is water detection since the water body reflects off all the energy from the radar so it appears in a SAR image as dark pixels. The traditional way to mark out water from SAR image is using the threshold method where each pixel is classified as water when its value is below a certain threshold. This method works fine in a plain rural area but the complex features of urban areas make it more challenging, for example, highways and buildings shadows can be easily misclassified as water. To solve this problem, we propose a deep learning solution to detect water from SAR image. The implemented convolutional neural network will no only identify water by the intensity of each pixel, it also learns the spatial information of neighborhood pixels. To train the network we used so2sat dataset which is processed from Sentinel-1 satellite SAR images. After training, we tested the neural network in many real SAR images and it gave us promising results that are more clear and better than the thresholding method. Moreover, to speed up the proposed solution, we were able to update it with the convolutional sliding windows. College of the Sciences Presentation Award Winner.
Recommended Citation
Lin, Chao Huang, "Waterbody detection from satellite SAR images using deep learning" (2020). Symposium Of University Research and Creative Expression (SOURCE). 47.
https://digitalcommons.cwu.edu/source/2020/COTS/47
Department/Program
Computer Sciences
Additional Mentoring Department
https://cwu.studentopportunitycenter.com/2020/04/waterbody-detection-from-satellite-sar-images-using-deep-learning/
Waterbody detection from satellite SAR images using deep learning
Ellensburg
Nowadays, Synthetic Aperture Radar (SAR) images have been widely used in the industry and the scientific community for different remote sensing applications. The main advantage of SAR technology is that it can acquire images from night time since it does not require sunlight. Additionally, it can penetrate the cloud which can capture images where the traditional optical sensor is limited. One of the remarkable applications of SAR image is water detection since the water body reflects off all the energy from the radar so it appears in a SAR image as dark pixels. The traditional way to mark out water from SAR image is using the threshold method where each pixel is classified as water when its value is below a certain threshold. This method works fine in a plain rural area but the complex features of urban areas make it more challenging, for example, highways and buildings shadows can be easily misclassified as water. To solve this problem, we propose a deep learning solution to detect water from SAR image. The implemented convolutional neural network will no only identify water by the intensity of each pixel, it also learns the spatial information of neighborhood pixels. To train the network we used so2sat dataset which is processed from Sentinel-1 satellite SAR images. After training, we tested the neural network in many real SAR images and it gave us promising results that are more clear and better than the thresholding method. Moreover, to speed up the proposed solution, we were able to update it with the convolutional sliding windows. College of the Sciences Presentation Award Winner.
https://digitalcommons.cwu.edu/source/2020/COTS/47
Faculty Mentor(s)
Razvan Andonie