Abstract
Remote sensing images are usually contaminated by cloud and corresponding shadow regions, making cloud and shadow detection one of the essential prerequisites for processing and translation of remote sensing images. Edge-precise cloud and shadow segmentation remains challenging due to the inherent high-level semantic acquisition of current neural segmentation fashions. We, therefore, introduce the Refined UNet series to partially achieve edge-precise cloud and shadow detection, including two-stage Refined UNet, v2 with a potentially efficient gray-scale guided Gaussian filter-based CRF, and v3 with an efficient multi-channel guided Gaussian filter-based CRF. However, it is visually demonstrated that the locally linear kernel used in v2 and v3 is not sufficiently sensitive to potential edges in comparison with Refined UNet. Accordingly, we turn back to the investigation of an end-to-end UNet-CRF architecture with a Gaussian-form bilateral kernel and its relatively efficient approximation. In this paper, we present Refined UNet v4, an end-to-end edge-precise segmentation network for cloud and shadow detection, which is capable of retrieving regions of interest with relatively tight edges and potential shadow regions with ambiguous edges. Specifically, we inherit the UNet-CRF architecture exploited in the Refined UNet series, which concatenates a UNet backbone of coarsely locating cloud and shadow regions and an embedded CRF layer of refining edges. In particular, the bilateral grid-based approximation to the Gaussian-form bilateral kernel is applied to the bilateral message-passing step, in order to ensure the delineation of sufficiently tight edges and the retrieval of shadow regions with ambiguous edges. Our TensorFlow implementation of the bilateral approximation is relatively computationally efficient in comparison with Refined UNet, attributed to the straightforward GPU acceleration. Extensive experiments on Landsat 8 OLI dataset illustrate that our v4 can achieve edge-precise cloud and shadow segmentation and improve the retrieval of shadow regions, and also confirm its computational efficiency.
Highlights
Remote sensing images are usually contaminated by cloud and corresponding shadow regions when acquired, which notoriously perturbs the recognition of land cover and leads to invalid resolved results [1]; more and more remote sensing applications require cloud- and shadow-free images [2,3,4,5,6]
Since each pixel within a remote sensing image should be identified as the category of cloud, shadow, or background, intelligent cloud and shadow detection is in practice formulated as a semantic segmentation task, driven by large-scale coarsely labeled data and sophisticated neural segmentation models
We introduce a UNet-conditional random field (CRF) architecture to address the issue of edge-precise cloud and shadow detection, the instances of which are referred to as the Refined UNet series [1,2,7]
Summary
Remote sensing images are usually contaminated by cloud and corresponding shadow regions when acquired, which notoriously perturbs the recognition of land cover and leads to invalid resolved results [1]; more and more remote sensing applications require cloud- and shadow-free images [2,3,4,5,6]. Edge-sensitive approaches refine the segmentation performance around edges in a pipeline way, using nonlinear filters to visually improve segmentation proposals As a consequence, it is worth developing an end-to-end solution to the edge-precise cloud and shadow segmentation. Refined UNet v4, an end-to-end edge-precise segmentation network for cloud and shadow detection, which is capable of retrieving regions of interest with relatively tight edges and potential shadow regions with ambiguous edges. Refined UNet v4: we propose an end-to-end network for cloud and shadow segmentation of remote sensing images, which can perform cloud and shadow detection in an edge-precise way, improve the retrieval of shadow regions with potential edges, and enable a relatively speed-up in comparison with Refined UNet [1].
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
