Use of Hyper-Spectral Visible and Near-Infrared Satellite Data for Timely Estimates of the Earth’s Surface Reflectance in Cloudy and Aerosol Loaded Conditions: Part 1–Application to RGB Image Restoration Over Land With GOME-2

Space-based quantitative passive optical remote sensing of the Earth’s surface typically involves the detection and elimination of cloud-contaminated pixels as an initial processing step. We explore a fundamentally different approach; we use machine learning with cloud contaminated satellite hyper-spectral data to estimate underlying terrestrial surface reflectances at red, green, and blue (RGB) wavelengths. An artificial neural network (NN) reproduces land RGB reflectances with high fidelity, even in scenes with moderate to high cloud optical thicknesses. This implies that spectral features of the Earth’s surface can be detected and distinguished in the presence of clouds, even when they are partially and visibly obscured by clouds; the NN is able to separate the spectral fingerprint of the Earth’s surface from that of the clouds, aerosols, gaseous absorption, and Rayleigh scattering, provided that there are adequately different spectral features and that the clouds are not completely opaque. Once trained, the NN enables rapid estimates of RGB reflectances with little computational cost. Aside from the training data, there is no requirement of prior information regarding the land surface spectral reflectance, nor is there need for radiative transfer calculations. We test different wavelength windows and instrument configurations for reconstruction of surface reflectances. This work provides an initial example of a general approach that has many potential applications in land and ocean remote sensing as well as other practical uses such as in search and rescue, precision agriculture, and change detection.

Monitoring of Spectral Signatures of Maize Crop using Temporal SAR and Optical Remote Sensing data

A study was carried out using the temporal Sentinel-1B microwave data (June to November at 12 days interval) and Sentinel-2A/2B optical data (June to November) to discriminate the maize crop from other competing crops rice and cotton in Siddipet district, Telangana state, India during kharif, 2019 (June to November). The study utilized the data from multiple sources such as Multi-temporal VH backscatter intensity from Sentinel-1B SAR and NDVI values from Sentinel-2A/2B in combination with field data to discriminate the maize crop. Synchronous to satellite pass, ground truth data on crop parameters viz., crop stage, crop vigour, biomass, plant height, plant density, soil moisture, LAI and chlorophyll content were collected. Multi-temporal VH backscatter intensity and Normalized Difference Vegetation Index (NDVI) data were used to characterize backscatter and greenness behaviour of the maize crop. The backscatter intensity (dB) for maize crop ranged from -21.83 (the lowest backscatter values) at planting to -12.52 (the highest backscatter values) at peak growth stage. The NDVI values during vegetative and reproductive stages (August and September) were >0.6 and during senescence to harvesting the values were less than or equal to 0.52. The increase in backscatter intensity values from initial vegetative stage to peak stage was due to increased volume scattering of the maize crop canopy and a continuous decline in backscatter intensity values of VH band at maturity stage, was due to decrease in greenness and moisture content in leaves of the maize crop helped in maize crop discrimination from other dominant kharif crops in the study area.

Semi-Supervised SAR Target Detection Based on an Improved Faster R-CNN

In the remote sensing image processing field, the synthetic aperture radar (SAR) target-detection methods based on convolutional neural networks (CNNs) have gained remarkable performance relying on large-scale labeled data. However, it is hard to obtain many labeled SAR images. Semi-supervised learning is an effective way to address the issue of limited labels on SAR images because it uses unlabeled data. In this paper, we propose an improved faster regions with CNN features (R-CNN) method, with a decoding module and a domain-adaptation module called FDDA, for semi-supervised SAR target detection. In FDDA, the decoding module is adopted to reconstruct all the labeled and unlabeled samples. In this way, a large number of unlabeled SAR images can be utilized to help structure the latent space and learn the representative features of the SAR images, devoting attention to performance promotion. Moreover, the domain-adaptation module is further introduced to utilize the unlabeled SAR images to promote the discriminability of features with the assistance of the abundantly labeled optical remote sensing (ORS) images. Specifically, the transferable features between the ORS images and SAR images are learned to reduce the domain discrepancy via the mean embedding matching, and the knowledge of ORS images is transferred to the SAR images for target detection. Ultimately, the joint optimization of the detection loss, reconstruction, and domain adaptation constraints leads to the promising performance of the FDDA. The experimental results on the measured SAR image datasets and the ORS images dataset indicate that our method achieves superior SAR target detection performance with limited labeled SAR images.

The Use of Artificial Intelligence-Based Optical Remote Sensing and Positioning Technology in Microelectronic Processing Technology

The purpose is to make defect detection in microelectronic processing technology fast, accurate, reliable, and efficient. A new optical remote sensing-optical beam induced resistance change (ORS-OBIRCH) target recognition and location defect detection method is proposed based on an artificial intelligence algorithm, optical remote sensing (ORS), and optical beam induced resistance change (OBIRCH) location technology using deep convolutional neural network. This method integrates the characteristics of high resolution and rich details of the image obtained by ORS technology and combines the advantages of photosensitive temperature characteristics in OBIRCH positioning technology. It can be adopted to identify, capture, and locate the defects of microdevices in the process of microelectronic processing. Simulation results show that this method can quickly reduce the detection range and locate defects accurately and efficiently. The experimental results reveal that the ORS-OBIRCH target recognition defect location detection method can complete the dynamic synchronization of the IC detection system and obtain high-quality images by changing the laser beam irradiation cycle. Moreover, it can analyze and process the detection results to quickly, accurately, and efficiently locate the defect location. Unlike the traditional detection methods, the success rate of detection has been greatly improved, which is about 95.8%, an increase of nearly 40%; the detection time has been reduced by more than half, from 5.5 days to 1.9 days, and the improvement rate has reached more than 65%. In a word, this method has good practical application value in the field of microelectronic processing.

A saliency model-oriented convolution neural network for cloud detection in remote sensing images

Remote sensing is an indispensable technical way for monitoring earth resources and environmental changes. However, optical remote sensing images often contain a large number of cloud, especially in tropical rain forest areas, make it difficult to obtain completely cloud-free remote sensing images. Therefore, accurate cloud detection is of great research value for optical remote sensing applications. In this paper, we propose a saliency model-oriented convolution neural network for cloud detection in remote sensing images. Firstly, we adopt Kernel Principal Component Analysis (KCPA) to unsupervised pre-training the network. Secondly, small labeled samples are used to fine-tune the network structure. And, remote sensing images are performed with super-pixel approach before cloud detection to eliminate the irrelevant backgrounds and non-clouds object. Thirdly, the image blocks are input into the trained convolutional neural network (CNN) for cloud detection. Meanwhile, the segmented image will be recovered. Fourth, we fuse the detected result with the saliency map of raw image to further improve the accuracy of detection result. Experiments show that the proposed method can accurately detect cloud. Compared to other state-of-the-art cloud detection method, the new method has better robustness.

Unraveling the hydrology and sediment balance of an ungauged lake in the Sudano-Sahelian region of West Africa using remote sensing

The presence of ephemeral ponds and perennial lakes in the Sudano-Sahelian region of West Africa is strongly variable in space and time. Yet, they have important ecological functions and societies are reliant on their surface waters for their lives and livelihoods. It is essential to monitor and understand the dynamics of these lakes to assess past, present, and future water resource changes. In this paper, we present an innovative approach to unravel the sediment and water balance of Lac Wégnia, a small ungauged lake in Mali near the capital of Bamako. The approach uses optical remote sensing data to identify the shoreline positions over a period of 22 years (2000–2021) and then attributes water surface heights (WSHs) to each observation using the lake bathymetry. The method represents a significant advancement over previous attempts to remotely monitor lakes in the West African drylands, since it considers not only changes in water depth to explain recent declining trends in lake areas, but also changes in the storage capacity. We recognize silting at the tributaries to the lake, but overall, erosion processes are dominant and threaten the persistence of the lake because of a continuous decrease of the floor level at the outflow. This explains the decreasing trend in WSH even for the wet-season, in spite of positive rainfall patterns.

A Comparative Study of Active Rock Glaciers Mapped from Geomorphic- and Kinematic-Based Approaches in Daxue Shan, Southeast Tibetan Plateau

Active rock glaciers (ARGs) are important permafrost landforms in alpine regions. Identifying ARGs has mainly relied on visual interpretation of their geomorphic characteristics with optical remote sensing images, while mapping ARGs from their kinematic features has also become popular in recent years. However, a thorough comparison of geomorphic- and kinematic-based inventories of ARGs has not been carried out. In this study, we employed a multi-temporal interferometric synthetic aperture radar (InSAR) technique to derive the mean annual surface displacement velocity over the Daxue Shan, Southeast Tibet Plateau. We then compiled a rock glacier inventory by synergistically interpreting the InSAR-derived surface displacements and geomorphic features based on Google Earth images. Our InSAR-assist kinematic-based inventory (KBI) was further compared with a pre-existing geomorphic-based inventory (GBI) of rock glaciers in Daxue Shan. The results show that our InSAR-assist inventory consists of 344 ARGs, 36% (i.e., 125) more than that derived from the geomorphic-based method (i.e., 251). Only 32 ARGs in the GBI are not included in the KBI. Among the 219 ARGs detected by both approaches, the ones with area differences of more than 20% account for about 32% (i.e., 70 ARGs). The mean downslope velocities of ARGs calculated from InSAR are between 2.8 and 107.4 mm∙a−1. Our comparative analyses show that ARGs mapping from the InSAR-based kinematic approach is more efficient and accurate than the geomorphic-based approach. Nonetheless, the completeness of the InSAR-assist KBI is affected by the SAR data acquisition time, signal decorrelation, geometric distortion of SAR images, and the sensitivity of the InSAR measurement to ground deformation. We suggest that the kinematic-based approach should be utilized in future ARGs-based studies such as regional permafrost distribution assessment and water storage estimates.