Automated Cloud and Cloud-Shadow Masking for Landsat 8 Using Multitemporal Images in a Variety of Environments

Small-scale placer mining in Colombia takes place in rural areas and involves excavations resulting in large footprints of bare soil and water ponds. Such excavated areas comprise a mosaic of challenging terrains for cloud and cloud-shadow detection of Sentinel-2 (S2A and S2B) data used to identify, map, and monitor these highly dynamic activities. This paper uses an efficient two-step machine-learning approach using freely available tools to detect clouds and shadows in the context of mapping small-scale mining areas, one which places an emphasis on the reduction of misclassification of mining sites as clouds or shadows. The first step is comprised of a supervised support-vector-machine classification identifying clouds, cloud shadows, and clear pixels. The second step is a geometry-based improvement of cloud-shadow detection where solar-cloud-shadow-sensor geometry is used to exclude commission errors in cloud shadows. The geometry-based approach makes use of sun angles and sensor view angles available in Sentinel-2 metadata to identify potential directions of cloud shadow for each cloud projection. The approach does not require supplementary data on cloud-top or bottom heights nor cloud-top ruggedness. It assumes that the location of dense clouds is mainly impacted by meteorological conditions and that cloud-top and cloud-base heights vary in a predefined manner. The methodology has been tested over an intensively excavated and well-studied pilot site and shows 50% more detection of clouds and shadows than Sen2Cor. Furthermore, it has reached a Specificity of 1 in the correct detection of mining sites and water ponds, proving itself to be a reliable approach for further related studies on the mapping of small-scale mining in the area. Although the methodology was tailored to the context of small-scale mining in the region of Antioquia, it is a scalable approach and can be adapted to other areas and conditions.

Section: Location Of Shadow With Respect To Cloud Projectionmentioning

confidence: 99%

Cloud and Cloud-Shadow Detection for Applications in Mapping Small-Scale Mining in Colombia Using Sentinel-2 Imagery

Ibrahim

Jiang

Lema³

et al. 2021

“…Candra et al [10] proposed an automated cloud and cloud shadow detection method by using multi temporal Lansat8 images, which is named MCM. This approach makes use of the reflectance differences between two images at the same location to identify cloud and cloud shadows, which is especially effective in tropical areas.…”

Section: An Overview Of Cloud Detection Approachesmentioning

confidence: 99%

Cloud Detection Using an Ensemble of Pixel-Based Machine Learning Models Incorporating Unsupervised Classification

Lary

2021

Remote sensing imagery, such as that provided by the United States Geological Survey (USGS) Landsat satellites, has been widely used to study environmental protection, hazard analysis, and urban planning for decades. Clouds are a constant challenge for such imagery and, if not handled correctly, can cause a variety of issues for a wide range of remote sensing analyses. Typically, cloud mask algorithms use the entire image; in this study we present an ensemble of different pixel-based approaches to cloud pixel modeling. Based on four training subsets with a selection of different input features, 12 machine learning models were created. We evaluated these models using the cropped LC8-Biome cloud validation dataset. As a comparison, Fmask was also applied to the cropped scene Biome dataset. One goal of this research is to explore a machine learning modeling approach that uses as small a training data sample as possible but still provides an accurate model. Overall, the model trained on the sample subset (1.3% of the total training samples) that includes unsupervised Self-Organizing Map classification results as an input feature has the best performance. The approach achieves 98.57% overall accuracy, 1.18% cloud omission error, and 0.93% cloud commission error on the 88 cropped test images. By comparison to Fmask 4.0, this model improves the accuracy by 10.12% and reduces the cloud omission error by 6.39%. Furthermore, using an additional eight independent validation images that were not sampled in model training, the model trained on the second largest subset with an additional five features has the highest overall accuracy at 86.35%, with 12.48% cloud omission error and 7.96% cloud commission error. This model’s overall correctness increased by 3.26%, and the cloud omission error decreased by 1.28% compared to Fmask 4.0. The machine learning cloud classification models discussed in this paper could achieve very good performance utilizing only a small portion of the total training pixels available. We showed that a pixel-based cloud classification model, and that as each scene obviously has unique spectral characteristics, and having a small portion of example pixels from each of the sub-regions in a scene can improve the model accuracy significantly.

“…The traditional threshold method [1][2][3] excessively relies on artificial calibration based on pixel values, that is the characteristics of clouds-high reflection and low temperatureand distinguishes clouds from ground objects by analyzing their Nearinfrared Spectrum (NIS). When it comes to complex situations, it is difficult to distinguish clouds using these manually calibrated characteristics.…”

Section: Introductionmentioning

confidence: 99%

“…Experience has proven that the overall accuracy of cloud detection by the Fmask method is improved, but the detection rate of thin clouds is still not ideal. Candra et al [3] added a Haze-Optimized Transformation (HOT) test and thermal band in the previous multitemporal cloud masking algorithm to improve the algorithm in the detection of haze, thin cirrus clouds, and thick clouds. They also improved the previous multitemporal cloud masking in the detection of cloud shadow by adding a blue band.…”

Section: Introductionmentioning

confidence: 99%

CDUNet: Cloud Detection UNet for Remote Sensing Imagery

Zhang

Xia

2021

Cloud detection is a key step in the preprocessing of optical satellite remote sensing images. In the existing literature, cloud detection methods are roughly divided into threshold methods and deep-learning methods. Most of the traditional threshold methods are based on the spectral characteristics of clouds, so it is easy to lose the spatial location information in the high-reflection area, resulting in misclassification. Besides, due to the lack of generalization, the traditional deep-learning network also easily loses the details and spatial information if it is directly applied to cloud detection. In order to solve these problems, we propose a deep-learning model, Cloud Detection UNet (CDUNet), for cloud detection. The characteristics of the network are that it can refine the division boundary of the cloud layer and capture its spatial position information. In the proposed model, we introduced a High-frequency Feature Extractor (HFE) and a Multiscale Convolution (MSC) to refine the cloud boundary and predict fragmented clouds. Moreover, in order to improve the accuracy of thin cloud detection, the Spatial Prior Self-Attention (SPSA) mechanism was introduced to establish the cloud spatial position information. Additionally, a dual-attention mechanism is proposed to reduce the proportion of redundant information in the model and improve the overall performance of the model. The experimental results showed that our model can cope with complex cloud cover scenes and has excellent performance on cloud datasets and SPARCS datasets. Its segmentation accuracy is better than the existing methods, which is of great significance for cloud-detection-related work.