Hierarchical Terrain Attention and Multi-Scale Rainfall Guidance For Flood Image Prediction

Feifei Wang,Yong Wang,Bing Li,Qidong Huang,Shaoqing Chen

2023-12-12

Abstract:With the deterioration of climate, the phenomenon of rain-induced flooding has become frequent. To mitigate its impact, recent works adopt convolutional neural network or its variants to predict the floods. However, these methods directly force the model to reconstruct the raw pixels of flood images through a global constraint, overlooking the underlying information contained in terrain features and rainfall patterns. To address this, we present a novel framework for precise flood map prediction, which incorporates hierarchical terrain spatial attention to help the model focus on spatially-salient areas of terrain features and constructs multi-scale rainfall embedding to extensively integrate rainfall pattern information into generation. To better adapt the model in various rainfall conditions, we leverage a rainfall regression loss for both the generator and the discriminator as additional supervision. Extensive evaluations on real catchment datasets demonstrate the superior performance of our method, which greatly surpasses the previous arts under different rainfall conditions.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper primarily addresses the issue of flood prediction caused by rainfall and proposes a new framework to improve prediction accuracy and generalization capability. Specifically, existing methods often directly use Convolutional Neural Networks (CNN) and their variants to reconstruct flood images, but these methods do not fully utilize the information in terrain features and rainfall patterns. To solve this problem, the paper introduces the following innovations:

1. **Hierarchical Terrain Spatial Attention (HTA)**: Through this mechanism, the model can focus on more significant spatial regions at different scales in terrain features, thereby better utilizing terrain information.

2. **Multi-Scale Rainfall Embedding (MRE)**: This method can fully integrate the information of rainfall patterns into the generation process, helping the model understand the impact of different rainfall patterns.

3. **Adding Rainfall Regression Loss**: To make the model better adapt to various rainfall conditions, the paper also introduces rainfall regression loss as an additional supervision signal, which helps the model extract high-level information of rainfall patterns from flood images.

Through the above methods, the proposed framework can significantly improve the accuracy of flood image prediction. Extensive experiments on multiple datasets show that this method significantly outperforms existing technical solutions. It performs particularly well in terms of generalization capability under different watershed and rainfall conditions.