IMAFD: An Interpretable Multi-stage Approach to Flood Detection from time series Multispectral Data

TL;DR

This paper tackles large-scale flood detection from time-series multispectral data by addressing two major gaps: computationally expensive dense change detection and lack of interpretable AI. It introduces IMAFD, a four-stage, interpretable pipeline that starts with anomalous-image detection, followed by targeted binary change detection, semantic change detection using the IDSS+ prototype-based method, and a decision-making step that fuses all information to declare Flooding or No Flooding. IDSS+ extends prototype-based explanations by using real-pixel prototypes and providing confidence maps and UMAP-based visualizations, enabling humans to understand the decision process. The framework is validated on WorldFloods, RaVAEn, and MediaEval, showing competitive performance relative to baselines while offering improved interpretability and insight into flooded areas, making it practical for large-scale remote sensing tasks.

Abstract

In this paper, we address two critical challenges in the domain of flood detection: the computational expense of large-scale time series change detection and the lack of interpretable decision-making processes on explainable AI (XAI). To overcome these challenges, we proposed an interpretable multi-stage approach to flood detection, IMAFD has been proposed. It provides an automatic, efficient and interpretable solution suitable for large-scale remote sensing tasks and offers insight into the decision-making process. The proposed IMAFD approach combines the analysis of the dynamic time series image sequences to identify images with possible flooding with the static, within-image semantic segmentation. It combines anomaly detection (at both image and pixel level) with semantic segmentation. The flood detection problem is addressed through four stages: (1) at a sequence level: identifying the suspected images (2) at a multi-image level: detecting change within suspected images (3) at an image level: semantic segmentation of images into Land, Water or Cloud class (4) decision making. Our contributions are two folder. First, we efficiently reduced the number of frames to be processed for dense change detection by providing a multi-stage holistic approach to flood detection. Second, the proposed semantic change detection method (stage 3) provides human users with an interpretable decision-making process, while most of the explainable AI (XAI) methods provide post hoc explanations. The evaluation of the proposed IMAFD framework was performed on three datasets, WorldFloods, RavAEn and MediaEval. For all the above datasets, the proposed framework demonstrates a competitive performance compared to other methods offering also interpretability and insight.

Figures (8)

• Figure 1: IMAFD architecture. The proposed IMAFD architecture consists of the following four stages, (1) anomalous image detection: identifying suspected images from time series multispectral optical images (2) binary change detection: detecting changes within suspected images (3) semantic change detection: semantic segmentation of changes into Land, Water or Cloud class (4) decision making.
• Figure 2: RGB visualization of Sentinel-2 time series multispectral images (All of the above images were taken in 2018).
• Figure 3: IDSS+ architecture.
• Figure 4: Comparison of segmentation results. The meaning of the colours is: green - Land, yellow - Cloud, blue - Water. For the confidence map, lighter colors indicate lower confidence, while darker colors indicate high confidence.
• Figure 5: Comparison of segmentation results. The meaning of the colours is: green - Land, yellow - Cloud, blue - Water.