An automatic water detection approach based on Dempster-Shafer theory for multi-spectral images

Abstract

Abstract-Detection of surface water in natural environment via multi-spectral imagery has been widely utilized in many fields, such land cover identification. However, due to the similarity of the spectra of water bodies, built-up areas, approaches based on high-resolution satellites sometimes confuse these features. A popular direction to detect water is spectral index, often requiring the ground truth to find appropriate thresholds manually. As for traditional machine learning methods, they identify water merely via differences of spectra of various land covers, without taking specific properties of spectral reflection into account. In this paper, we propose an automatic approach to detect water bodies based on Dempster-Shafer theory, combining supervised learning with specific property of water in spectral band in a fully unsupervised context. The benefits of our approach are twofold. On the one hand, it performs well in mapping principle water bodies, including little streams and branches. On the other hand, it labels all objects usually confused with water as 'ignorance', including half-dry watery areas, built-up areas and semi-transparent clouds and shadows. 'Ignorance' indicates not only limitations of the spectral properties of water and supervised learning itself but insufficiency of information from multi-spectral bands as well, providing valuable information for further land cover classification.

Automatic flood detection in SentineI-2 images using deep convolutional neural networks

“…In contrast all the indices in gure 3, except Li et al [17], show water pixels above 0 value which can be variable due to atmospheric or brightness eects. Moreover, while Li et al [17] had a threshold that lies near 0.4, we didn't change the axis to retain their original range in order to dierentiate between water and non-water pixels. For the proposed index we also see that, for cloud free images the pixels increased in positive side as water is clearly visible.…”

“…Another approach in order to reduce mis-classication due to built-up areas creates a three-dimensional feature space by utilising NDWI, Normalised dierence vegetation index (NDVI), and Red Edge NDWI (RE-NDWI) [12] to detect water using supervised learning [17].…”

“…Since the data provided does include information across a range of wavebands, we can immediately apply a number of water indices directly to the data to attempt to identify ooded regions. Resultant images for each of NDWI, MNDWI and AWEI, as well as the index proposed by Mishra & Prasad [20] and Li et al's [17] three-dimensional feature space , can be seen in Figure 1. As noted earlier, cloud coverage and cloud shadow are two of the major problems when dealing with remote sensing images.…”

“…Given the relatively good performance of threshold based models such as [20], we apply this model as a baseline for our proposed index. However, considering the cloud coverage problem in remote (a) Original Image (b) NDWI [18] (c) MNDWI [29] (d) AWEI [5] (e) Mishra et al [20] (f) Li et al [17] Figure 1: Dierent Water Index Techniques Figure 2: Convolutional Neural Network sensing satellite images, it is good to look for its modication in indexing technique. As SWIR and Blue band are both helpful for images that present clouds, snow and penetration of thin clouds.…”

“…A ood mapping is a technique to analyse the extent of ood water, and identies its spatial distribution over the land area; these are now common place. In recent years there has been an increase in techniques proposed that aim to map water bodies [18], [29], [5], [17], [20] in order to assist with ood mapping and response. While automatic detection of oods from raw satellite image data seems like an obvious method for accurate detection, the problem is in practice non-trivial.…”