Deep Convolutional Neural Network for Flood Extent Mapping Using Unmanned Aerial Vehicles Data

Abstract

Flooding is one of the leading threats of natural disasters to human life and property, especially in densely populated urban areas. Rapid and precise extraction of the flooded areas is key to supporting emergency-response planning and providing damage assessment in both spatial and temporal measurements. Unmanned Aerial Vehicles (UAV) technology has recently been recognized as an efficient photogrammetry data acquisition platform to quickly deliver high-resolution imagery because of its cost-effectiveness, ability to fly at lower altitudes, and ability to enter a hazardous area. Different image classification methods including SVM (Support Vector Machine) have been used for flood extent mapping. In recent years, there has been a significant improvement in remote sensing image classification using Convolutional Neural Networks (CNNs). CNNs have demonstrated excellent performance on various tasks including image classification, feature extraction, and segmentation. CNNs can learn features automatically from large datasets through the organization of multi-layers of neurons and have the ability to implement nonlinear decision functions. This study investigates the potential of CNN approaches to extract flooded areas from UAV imagery. A VGG-based fully convolutional network (FCN-16s) was used in this research. The model was fine-tuned and a k-fold cross-validation was applied to estimate the performance of the model on the new UAV imagery dataset. This approach allowed FCN-16s to be trained on the datasets that contained only one hundred training samples, and resulted in a highly accurate classification. Confusion matrix was calculated to estimate the accuracy of the proposed method. The image segmentation results obtained from FCN-16s were compared from the results obtained from FCN-8s, FCN-32s and SVMs. Experimental results showed that the FCNs could extract flooded areas precisely from UAV images compared to the traditional classifiers such as SVMs. The classification accuracy achieved by FCN-16s, FCN-8s, FCN-32s, and SVM for the water class was 97.52%, 97.8%, 94.20% and 89%, respectively.

Patch Similarity Convolutional Neural Network for Urban Flood Extent Mapping Using Bi-Temporal Satellite Multispectral Imagery

“…Unlike pixel-based classification for flood mapping [6][7][8]12,39,40], this study investigated image patch based flood mapping similar to the study in [3]. Major motivations include: (1) reducing the impact of heterogeneous image background over urban area, which is challenging for pixel-based classification; and (2) accelerating human annotation of training samples since pixel-wise labeling would be much more time-consuming and labor-intensive.…”

“…Flood extent mapping is a process to identify the land areas impacted by flooding. Various definitions of such flooding areas have been proposed [3,[6][7][8]. For example, only land areas covered by visible floodwaters are considered as being flooded [6,8].…”

“…Various definitions of such flooding areas have been proposed [3,[6][7][8]. For example, only land areas covered by visible floodwaters are considered as being flooded [6,8]. However, according to the National Flood Mapping Products from the Federal Emergency Management Agency (FEMA) [32], small areas covered by invisible floodwaters due to trees or surrounded by floodwaters are also treated as being flooded [3,7] (see Figure 2).…”

“…With regard to the mapping methods, most of literature focused on pixel-based dense classification approaches such as artificial neural network (ANN) [19], SVM [18], DT [14], RF [8], HMT [7], particle swarm optimization (PSO) [12], and deep CNN such as FCNs [6], U-Net [20], and Deeplab [21]. While pixel-based image segmentation approaches in the aforementioned studies could generate higher resolution flood maps, they depend on high resolution flooding masks for model training, which require intensive human annotation of training samples.…”

“…As a result, corresponding ground objects from the bi-temporal (pre-and post-flooding) imagery would have consistent surface reflectance. Similar to the studies in [6][7][8]29], we conducted extensive experiments with PSNet and other baseline methods including SVM, DT, RF, and AdaBoost (ADB), using two datasets: (1) the 2017 Hurricane Harvey flood in Houston, Texas, and (2) the 2018 Hurricane Florence flood in Lumberton, North Carolina. We used default parameters in scikit-learn [31] for experiments with baseline methods as in [7].…”