Cloud masking is of central importance to the Earth Observation community. This paper deals with the problem of detecting clouds in visible and multispectral imagery from high-resolution satellite cameras. Recently, Machine Learning has offered promising solutions to the problem of cloud masking, allowing for more flexibility than traditional thresholding techniques, which are restricted to instruments with the requisite spectral bands. However, few studies use multi-scale features (as in, a combination of pixel-level and spatial) whilst also offering compelling experimental evidence for real-world performance. Therefore, we introduce CloudFCN, based on a Fully Convolutional Network architecture, known as U-net, which has become a standard Deep Learning approach to image segmentation. It fuses the shallowest and deepest layers of the network, thus routing low-level visible content to its deepest layers. We offer an extensive range of experiments on this, including data from two high-resolution sensors—Carbonite-2 and Landsat 8—and several complementary tests. Owing to a variety of performance-enhancing design choices and training techniques, it exhibits state-of-the-art performance where comparable to other methods, high speed, and robustness to many different terrains and sensor types.

A paradigm shift is underway in Earth observation, as deep learning (DL) replaces other methods for many predictive tasks. Nevertheless, most DL classification models for Earth observation are limited by their specificity with respect to both the sensors used (inputs) and classes predicted (outputs), leading to models that only perform well for specific satellites and on specific datasets. Cloud masking is typical of this, but is one of the most important tasks to generalize across sensors, given that it is required for all optical instruments. This work sets out a framework to relax DL’s constraints on specific inputs and outputs, using cloud and shadow masking as a case-study. Centrally, a model which is sensor independent, and which can simultaneously learn from different labeling schemes is developed. The model, Spectral ENcoder for SEnsor Independence version 2 (SEnSeI-v2) extends the original version, by permitting multimodal data [in this case Sentinel-1 synthetic aperture radar (SAR) imagery and a digital elevation model (DEM)] to be ingested, along with several other architectural improvements. SEnSeI-v2, attached to SegFormer, is shown to have state-of-the-art performance, whilst being usable on a range of multispectral band combinations, alongside SAR and DEM inputs, without retraining. The labeling schemes of eight datasets are not made compatible through a reductive approach (e.g., converting to cloud versus noncloud), rather, an ambiguous cross-entropy loss is introduced that allows the model to learn from the different labeling schemes without sacrificing the class distinctions of each, leading to a model which predicts all of the constituent classes of the different datasets.

We introduce a novel neural network architecture—spectral encoder for sensor independence (SEnSeI)—by which several multispectral instruments, each with different combinations of spectral bands, can be used to train a generalized deep learning model. We focus on the problem of cloud masking, using several preexisting datasets, and a new, freely available dataset for Sentinel-2. Our model is shown to achieve state-of-the-art performance on the satellites on which it was trained (Sentinel-2 and Landsat 8) and is able to extrapolate to sensors that it has not seen during training, such as Landsat 7, PerúSat-1, and Sentinel-3 Sea and Land Surface Temperature Radiometer (SLSTR). Model performance is shown to improve when multiple satellites are used in training, approaching, or surpassing the performance of specialized, single-sensor models. This work is motivated by the fact that the remote sensing community has access to data taken with a huge variety of sensors. This has inevitably led to labeling efforts being undertaken separately for different sensors, which limits the performance of deep learning models, given their need for huge training sets to perform optimally. Sensor independence can enable deep learning models to utilize multiple datasets for training simultaneously, boosting performance, and making them much more widely applicable. This may lead to deep learning approaches being used more frequently for onboard applications and in ground segment data processing, which generally requires models to be ready at launch or soon afterward.

We present here a dataset of nearly 5000 small craters across roughly 1700 km² of the Martian surface, in the MC-11 East quadrangle. The dataset covers twelve 2000-by-2000 pixel Context Camera images, each of which is comprehensively labelled by six annotators, whose results are combined using agglomerative clustering. Crater size-frequency distributions are centrally important to the estimation of planetary surface ages, in lieu of in-situ sampling. Older surfaces are exposed to meteoritic impactors for longer and, thus, are more densely cratered. However, whilst populations of larger craters are well understood, the processes governing the production and erosion of small (sub-km) craters are more poorly constrained. We argue that, by surveying larger numbers of small craters, the planetary science community can reduce some of the current uncertainties regarding their production and erosion rates. To this end, many have sought to use state-of-the-art object detection techniques utilising Deep Learning, which—although powerful—require very large amounts of labelled training data to perform optimally. This survey gives researchers a large dataset to analyse small crater statistics over MC-11 East, and allows them to better train and validate their crater detection algorithms. The collection of these data also demonstrates a multi-annotator method for the labelling of many small objects, which produces an estimated confidence score for each annotation and annotator.

This study proposed a new method of retrieving 10-m spectral surface albedo products. Three crucial components are incorporated into this high-resolution surface albedo generation system. Firstly, a deep learning system, CloudFCN based on the U-net paradigm has been developed. This produces the best available cloud detection of any algorithm published to date. Secondly, an advanced atmospheric correction model, the Sensor Invariant Atmospheric Correction (SIAC) is employed. The SIAC method considers the surface BRDF effects as these are usually ignored, because the atmosphere correction is a large signal and the largest uncertainty in converting top-of-atmosphere reflectance to top-of-canopy surface reflectance. Thirdly, an endmember-based new technology will be used to retrieve high-resolution albedo from high-resolution reflectance by combining downscaled MODIS BRDF. These methods are further described alongside results shown of the different stages and the final high resolution albedo.