High Resolution Optical Imagery Research Articles

Creating near real-time alerts of illegal gold mining in the Peruvian Amazon using Synthetic Aperture Radar

Abstract Peru’s Southeastern Amazon deforestation trends can be attributed to alluvial gold mining. Illegal mining occurring in forestry concessions, national parks, and the territories of Indigenous People Organizations is of particular concern. We present a methodology to create near real-time alerts of deforestation caused by alluvial gold mining. A time series of Sentinel-1 Synthetic Aperture Radar (SAR) data from February to December 2022 is created in Google Earth Engine (GEE) and assessed using Morton Canty’s Omnibus Q-test change detection algorithm. Resulting detections are validated with high-resolution optical imagery from Planet NICFI’s monthly basemaps and Planet Scope daily imagery. The alerts identify the location and timing of large areas (group pixels of <1 ha) of forest loss due to gold mining activities within buffer zones of indigenous territories and protected areas. The overall accuracy of the forest loss analysis conducted with this change detection method was 99.98%, based on an independent accuracy assessment. This effort has resulted in a public web platform that displays the location of near real time alerts, so Peruvian enforcement agencies can more effectively allocate resources and staff to addressing active illegal mining operations. These results demonstrate the applicability of open-source SAR data to monitor forest loss over areas where cloud cover is more persistent and to improve tools that deliver timely, critical information to decision-makers. Future applications of our method could expand this approach to other drivers of deforestation.

Scaling Terrain-Aware Spatial Machine Learning for Flood Mapping on Large Scale Earth Imagery Data

The accurate and prompt mapping of flood-affected regions is important for effective disaster management, including damage assessment and relief efforts. While high-resolution optical imagery from satellites during disasters presents an opportunity for automated flood inundation mapping, existing segmentation models face challenges due to noises like cloud cover and tree canopies. Thanks to the digital elevation model (DEM) data readily available from sources such as United States Geological Survey (USGS), terrain guidance was utilized by recent graphical models such as hidden Markov trees (HMTs) to improve segmentation quality. Unfortunately, these methods either can only handle a small area where water levels at different locations are assumed to be consistent, or require restricted assumptions such as there is only one river channel. This paper presents an algorithm for flood extent mapping on large-scale Earth imagery, applicable to a large geographic area with multiple river channels. Since water level can vary a lot from upstream to downstream, we propose to detect river pixels in order to partition the remaining pixels into localized zones, each with a unique water level. In each zone, water at all locations flow to the same river entry point. Pixels in each zone are organized by an HMT to capture water flow directions guided by elevations. Moreover, a novel regularization scheme is designed to enforce inter-zone consistency by penalizing pixel-pairs of adjacent zones that violate terrain guidance. Efficient parallelization is made possible by coloring the zone adjacency graph to identify zones and zone-pairs that have no dependency and hence can be processed in parallel, and incremental one-pass terrain-guided scanning is conducted wherever applicable to reuse computations. Experiments demonstrate that our solution is more accurate than existing solutions, and can efficiently and accurately map out flooding pixels in a giant area of size 24,805 × 40,129. Despite the imbalanced workloads caused by a few large zonal HMTs dominating the serial computing time, our parallelization approach is effective and manages to achieve up to 14.3 × speedup on a machine with Intel Xeon Gold 6126 CPU @ 2.60GHz (24 cores, 48 threads) using 32 threads.

High-resolution regional sea-ice model based on the discrete element method with boundary conditions from a large-scale model for ice drift

Abstract Understanding sea-ice dynamics at the floe scale is crucial to comprehend polar climate systems. While continuum models are commonly used to simulate large-scale sea-ice dynamics, they have limitations in accurately representing sea-ice behaviour at small scales. DEMs, on the other hand, are well-suited for modelling the behaviour of individual ice floes but face limitations due to computational constraints. To address the limitations of both approaches while combining their strengths, we explored the feasibility of using a DEM within a continuum model, where the latter provides boundary conditions for a rectangular high-resolution DEM domain. This paper presents a feasibility study where a discrete model of a 100 × 100 km2 icefield was created using high-resolution optical satellite imagery. Sea-ice dynamics were simulated in the DEM considering environmental forces and integrating large-scale ice-drift velocities as boundary conditions. Model predictions were compared with satellite observations for ice drift and deformation parameters. This numerical approach showed potential for offering accurate, high-resolution predictions of sea ice, particularly in coastal areas and near islands, and may find applications in ice navigation and climate studies. However, further development of the DEM, along with upgrades to the coupled ocean models providing data for the ice component, may be necessary. Additionally, challenges remain to develop a two-way coupling between the DEM and a continuum model, which may be needed to improve the accuracy of large-scale simulations.

Sea surface Fresnel reflections difference driven de-glint algorithm for airborne optical images.

This paper presents a glint correction algorithm for high spatial resolution optical remote sensing imagery captured by the ER-2 Airborne Visual Infrared Imaging Spectrometer (AVIRIS). The algorithm employs linear and differential techniques to mitigate sun glint and sky glint effects, encompassing statistical glint reflections resulting from variations in imaging angles within strips and inter-strip variations due to Fresnel reflectance disparities. It aims to diminish Fresnel reflectance diversity on water surfaces and mitigate the distortions induced by glint reflectance during spectral and ocean color inversion. A comparative analysis of spectral and ocean color information in AVIRIS images before and after correction reveals enhanced accuracy following the glint correction. By systematically addressing multiple glint reflections and their ramifications, this method offers a valuable framework for correcting water surface glint in diverse high spatial resolution optical imagery.

Detection of forest disturbance across California using deep-learning on PlanetScope imagery

California forests have recently experienced record breaking wildfires and tree mortality from droughts, However, there is inadequate monitoring, and limited data to inform policies and management strategies across the state. Although forest surveys and satellite observations of forest cover changes exist at medium to coarse resolutions (30–500 m) annually, they remain less effective in mapping small disturbances of forest patches (<5 m) occurring multiple times a year. We introduce a novel method of tracking California forest cover using a supervised U-Net deep learning architecture and PlanetScope’s Visual dataset which provides 3-band RGB (Red, Green, and Blue) mosaicked imagery. We created labels of forest and non-forest to train the U-Net model to map tree cover based on a semi-unsupervised classification method. We then detected changes of tree cover and disturbance with the U-Net model, achieving an overall accuracy of 98.97% over training data set, and 95.5% over an independent validation dataset, obtaining a precision of 82%, and a recall of 74%. With the predicted tree cover mask, we created wall to wall monthly tree cover maps over California at 4.77 m resolution for 2020, 2021, and 2022. These maps were then aggregated in a post-processing step to develop annual maps of disturbance, while accounting for the time of disturbance and other confounding factors such as topography, phenological and snow cover variability. We compared our high-resolution disturbance maps with wildfire GIS survey data from CALFIRE, and satellite-based forest cover changes and achieved an F-1 score of 54% and 88% respectively. The results suggest that high-resolution maps capture variability of forest disturbance and fire that wildfire surveys and medium resolution satellite products cannot. From 2020 to 2021, California maintained 30,923.5 sq km of forest while 5,994.9 sq km were disturbed. The highest observed forest loss rate was located at the Sierra Nevada mountains at 21.4% of the forested area being disturbed between 2020 and 2021. Our findings highlight the strong potential of deep learning and high-resolution RGB optical imagery for mapping complex forest ecosystems and their changes across California, as well as the application of these techniques on a national to global scale.

Monitoring dynamics of Kyagar Glacier surge and repeated draining of Ice-dammed lake using multi-source remote sensing

Glacier surges, a primary factor contributing to various glacial hazards, has long captivated the attention of the global glaciological community. This study delves into the dynamics of Kyagar Glacier surging and the associated drainage features of its Ice-dammed lake, employing high temporal resolution optical imagery. Our findings indicate that the surge on Kyagar Glacier began in late spring and early summer of 2014 and concluded during the summer of 2016. This surge resulted in the transfer of 0.321 ± 0.012 km3 of glacier mass from the reservoir zone to the receiving zone, leading to the formation of an ice-dammed lake at the glacier's terminus. The lake experienced five outbursts between 2015 and 2019, with the largest discharge occurring in 2017. And the maximum water depth during this period was 112 ± 11 m, resulting in a water storage volume of (158.37 ± 28.32) × 106 m3. On the other hand, our analysis of the relationship between glacier surface velocity and albedo, coupled with an examination of subglacial dynamics, revealed that increased precipitation during the active phase of the Kyagar Glacier results in accumulation of mass in the upper glacier. This accumulation induces changes in basal shear stress, triggering the glacier's transition into an unstable state. Consequently, glacier deformation rates escalate, surface crevasses proliferate, potentially providing conduits for surface meltwater to infiltrate the glacier bed. This, in turn, leaded to elevated basal water pressure, initiating glacier sliding. Furthermore, we postulated that the repetitive drainage of Kyagar Ice-dammed lake was primarily influenced by the opening and closing of subglacial drainage pathways and variations in inflow volumes. Future endeavors necessitate rigorous field observations to enhance glacier surge simulations, deepening our comprehension of glacier surge mechanisms and mitigating the impact of associated glacial hazards.

Shoreline delineation and change analysis in response to sea level rise and coastal bathymetry along the coast of Visakhapatnam, India using high-resolution optical imagery

Shoreline delineation and change analysis in response to sea level rise and coastal bathymetry along the coast of Visakhapatnam, India using high-resolution optical imagery

Inventory of Landslides in the Northern Half of the Taihang Mountain Range, China

The Taihang Mountains are a critical mountain range and geographical boundary in eastern China. Landslide disasters are particularly common in this region and usually cause serious casualties and property damage. However, previous landslide inventories in the region are limited and lack comprehensive landslide cataloguing. To address this gap, the northern half of the Taihang Mountain Range was selected for this study. A landslide database for the area was constructed using multi-temporal high-resolution optical imagery from the Google Earth and human–computer interactive visual interpretation technology. The results indicate that at least 8349 landslides have occurred in the Taihang Mountain Range, with a total landslide area of about 151.61 km2. The size of the landslides varies, averaging about 18,159.23 m2, with the largest landslide covering 2.83 km2 and the smallest landslide only 5.95 m2. The significance of this study lies in its ability to enhance our understanding of the distribution of landslides in the northern half of the Taihang Mountains. Furthermore, it offers valuable data references and supports for landslide assessment, early warning systems, disaster management, and ecological protection efforts.

Leveraging Deep Learning for High-Resolution Optical Satellite Imagery From Low-Cost Small Satellite Platforms

Leveraging Deep Learning for High-Resolution Optical Satellite Imagery From Low-Cost Small Satellite Platforms

Integrating Multiple Geospatial Datasets for Precise and Efficient Surface Water Extent Mapping in Malaysia

Accurate mapping of surface water extent is essential for several applications, including water resources management, climate modelling, biodiversity conservation, agricultural irrigation, hydropower, and food security. Inland water management is critical in Malaysia due to dams, river basins, and heavy rainfall. In a tropical country, mapping surface water is challenging due to cloud cover and dense vegetation. This study proposes an approach that uses rule sets in ArcGIS Model Builder to automate the identification of water bodies. At the regional scale, the extent of surface waters can be mapped by integrating multiple geospatial datasets, including surface terrain, global surface water data, and synthetic aperture radar (SAR) imagery. Accuracy is calculated by comparing the results to high-resolution optical imagery via the SecureWatch® application. The accuracy of 200 stratified random points is 95%, with a kappa value of 0.90 for water body and land area classes. This study maps a 12,563.95 km2 study area covering several states in Malaysia. These results show that the integration of multiple geospatial datasets using rule sets and data modelling can map the extent of surface waters in Malaysia. The proposed approach could be applied in other tropical countries with similar topography.

Integrating object-based and pixel-based segmentation for building footprint extraction from satellite images

Accurately delineating building footprints from optical satellite imagery presents a formidable challenge, particularly in urban settings characterized by intricate and diverse structures. Consequently, enhancing the utility of these images for geospatial data updates demands meticulous refinement. Machine learning algorithms have made notable contributions in this context, yet the pursuit of precision remains an ongoing challenge. This paper aims to enhance the accuracy of building footprint extraction through the integration of object-based and pixel-based segmentation techniques. Additionally, it evaluates the performance of machine learning methodologies, specifically LightGBM, XGBoost, and Neural Network (NN) approaches. The model's evaluation employed low spectral resolution optical images, widely accessible and cost-effective for acquisition. The study's outcomes demonstrate a substantial enhancement in extraction accuracy compared to extant literature. The proposed methodology attains an overall accuracy of 99.39%, an F1 measurement of 0.9935, and a Cohen Kappa index of 0.9870. Thus, the proposed approach signifies a noteworthy advancement over existing techniques for building footprint extraction from high-resolution optical imagery.

Dual Parallel Branch Fusion Network for Road Segmentation in High-Resolution Optical Remote Sensing Imagery

Road segmentation from high-resolution (HR) remote sensing images plays a core role in a wide range of applications. Due to the complex background of HR images, most of the current methods struggle to extract a road network correctly and completely. Furthermore, they suffer from either the loss of context information or high redundancy of details information. To alleviate these problems, we employ a dual branch dilated pyramid network (DPBFN), which enables dual-branch feature passing between two parallel paths when it is merged to a typical road extraction structure. A DPBFN consists of three parts: a residual multi-scaled dilated convolutional network branch, a transformer branch, and a fusion module. Constructing pyramid features through parallel multi-scale dilated convolution operations with multi-head attention block can enhance road features while suppressing redundant information. Both branches after fusing can solve shadow or vision occlusions and maintain the continuity of the road network, especially on a complex background. Experiments were carried out on three datasets of HR images to showcase the stable performance of the proposed method, and the results are compared with those of other methods. The OA in the three data sets of Massachusetts, Deep Globe, and GF-2 can reach more than 98.26%, 95.25%, and 95.66%, respectively, which has a significant improvement compared with the traditional CNN network. The results and explanation analysis via Grad-CAMs showcase the effective performance in accurately extracting road segments from a complex scene.

Unveiling the Potential of Drone-Borne Optical Imagery in Forest Ecology: A Study on the Recognition and Mapping of Two Evergreen Coniferous Species

The use of drone-borne imagery for tree recognition holds high potential in forestry and ecological studies. Accurate species identification and crown delineation are essential for tasks such as species mapping and ecological assessments. In this study, we compared the results of tree crown recognition across three neural networks using high-resolution optical imagery captured by an affordable drone with an RGB camera. The tasks included the detection of two evergreen coniferous tree species using the YOLOv8 neural network, the semantic segmentation of tree crowns using the U-Net neural network, and the instance segmentation of individual tree crowns using the Mask R-CNN neural network. The evaluation highlighted the strengths and limitations of each method. YOLOv8 demonstrated effective multiple-object detection (F1-score—0.990, overall accuracy (OA)—0.981), enabling detailed analysis of species distribution. U-Net achieved less accurate pixel-level segmentation for both species (F1-score—0.981, OA—0.963). Mask R-CNN provided precise instance-level segmentation, but with lower accuracy (F1-score—0.902, OA—0.822). The choice of a tree crown recognition method should align with the specific research goals. Although YOLOv8 and U-Net are suitable for mapping and species distribution assessments, Mask R-CNN offers more detailed information regarding individual tree crowns. Researchers should carefully consider their objectives and the required level of accuracy when selecting a recognition method. Solving practical problems related to tree recognition requires a multi-step process involving collaboration among experts with diverse skills and experiences, adopting a biology- and landscape-oriented approach when applying remote sensing methods to enhance recognition results. We recommend capturing images in cloudy weather to increase species recognition accuracy. Additionally, it is advisable to consider phenological features when selecting optimal seasons, such as early spring or late autumn, for distinguishing evergreen conifers in boreal or temperate zones.

Summer sea ice floe perimeter density in the Arctic: high-resolution optical satellite imagery and model evaluation

Abstract. Size distribution of sea ice floes is an important component for sea ice thermodynamic and dynamic processes, particularly in the marginal ice zone. Recently processes related to the floe size distribution (FSD) have been incorporated into sea ice models, but the sparsity of existing observations limits the evaluation of FSD models, thus hindering model improvements. In this study, perimeter density has been applied to characterise the floe size distribution for evaluating three FSD models – the Waves-in-Ice module and Power law Floe Size Distribution (WIPoFSD) model and two branches of a fully prognostic floe size-thickness distribution model: CPOM-FSD and FSDv2-WAVE. These models are evaluated against a new FSD dataset derived from high-resolution satellite imagery in the Arctic. The evaluation shows an overall overestimation of floe perimeter density by the models against the observations. Comparison of the floe perimeter density distribution with the observations shows that the models exhibit a much larger proportion for small floes (radius <10–30 m) but a much smaller proportion for large floes (radius >30–50 m). Observations and the WIPoFSD model both show a negative correlation between sea ice concentration and the floe perimeter density, but the two prognostic models (CPOM-FSD and FSDv2-WAVE) show the opposite pattern. These differences between models and the observations may be attributed to limitations in the observations (e.g. the image resolution is not sufficient to detect small floes) or limitations in the model parameterisations, including the use of a global power-law exponent in the WIPoFSD model as well as too weak a floe welding and enhanced wave fracture in the prognostic models.

Prediction of Unpaved Road Conditions Using High-Resolution Optical Satellite Imagery and Machine Learning

Rural roads play a crucial role in fostering economic and social development in Africa. Local Road Authorities (LRAs) struggle to collect road condition data using conventional means due to logistical and resource issues. Poor road conditions and restricted mobility have severe economic consequences for the transport of goods and services. Lack of maintenance can increase costs three-fold. In this work, a novel framework is proposed in which earth observations using high-resolution optical satellite imagery are applied to measure the condition of unpaved roads, providing a vital input to maintenance planning and prioritisation. A trial was conducted using this method on 83 roads in Tanzania totalling 131.7 km. The experimental results demonstrate that, by analysing variations in pixel intensity of the road surface, the condition can be estimated with an accuracy of 71.9% when compared to ground truth information. Machine Learning techniques are applied to the same network to test the performance of the system in predicting road conditions. A blended classifier approach achieves an accuracy of 88%. The proposed framework enables LRAs to define the information they receive based on their specific priorities, offering a rapid, objective, consistent and potentially cost-effective system that overcomes the current challenges faced by LRAs.

Windthrow damage detection in Nordic forests by 3D reconstruction of very high-resolution stereo optical satellite imagery

ABSTRACT We tested whether windthrow damage to Nordic conifer forest stands could be reliably detected as canopy height decrease between a pre-storm LiDAR (Light Detection and Ranging) digital surface model (DSM) and a photogrammetric DSM derived from a post-storm WorldView-3 stereo pair. The post-storm ground reference data consisted of field and unmanned aerial vehicle (UAV) observations of windthrow combined with no-damage areas collected by visual interpretation of the available very high resolution (VHR) satellite imagery. We trained and tested a thresholding model using canopy height change as the sole predictor. We undertook a two-step accuracy assessment by (1) running k-fold cross-validation on the ground reference dataset and examining the effect of the potential imperfections in the ground reference data, and (2) conducting rigorous accuracy assessment of the classified map of the study area using an extended set of VHR imagery. The thresholding model produced accurate windthrow maps in dense, productive forest stands with a sensitivity of 96%, specificity of 71%, and Matthews correlation coefficient (MCC) over 0.7. However, in sparse and high elevation stands, the classification accuracy was poor. Despite certain collection challenges during the winter months in the Nordic region, we consider VHR stereo satellite imagery to be a viable source of forest canopy height information and sufficiently accurate to map windthrow disturbance in forest stands of high to moderate density.

Land Use and Land Cover Mapping with VHR and Multi-Temporal Sentinel-2 Imagery

Land Use/Land Cover (LULC) mapping is the first step in monitoring urban sprawl and its environmental, economic and societal impacts. While satellite imagery and vegetation indices are commonly used for LULC mapping, the limited resolution of these images can hamper object recognition for Geographic Object-Based Image Analysis (GEOBIA). In this study, we utilize very high-resolution (VHR) optical imagery with a resolution of 50 cm to improve object recognition for GEOBIA LULC classification. We focused on the city of Nice, France, and identified ten LULC classes using a Random Forest classifier in Google Earth Engine. We investigate the impact of adding Gray-Level Co-Occurrence Matrix (GLCM) texture information and spectral indices with their temporal components, such as maximum value, standard deviation, phase and amplitude from the multi-spectral and multi-temporal Sentinel-2 imagery. This work focuses on identifying which input features result in the highest increase in accuracy. The results show that adding a single VHR image improves the classification accuracy from 62.62% to 67.05%, especially when the spectral indices and temporal analysis are not included. The impact of the GLCM is similar but smaller than the VHR image. Overall, the inclusion of temporal analysis improves the classification accuracy to 74.30%. The blue band of the VHR image had the largest impact on the classification, followed by the amplitude of the green-red vegetation index and the phase of the normalized multi-band drought index.

Mapping Pluvial Flood-Induced Damages with Multi-Sensor Optical Remote Sensing: A Transferable Approach

Pluvial floods caused by extreme overland flow inland account for half of all flood damage claims each year along with fluvial floods. In order to increase confidence in pluvial flood susceptibility mapping, overland flow models need to be intensively evaluated using observations from past events. However, most remote-sensing-based flood detection techniques only focus on the identification of degradations and/or water pixels in the close vicinity of overflowing streams after heavy rainfall. Many occurrences of pluvial-flood-induced damages such as soil erosion, gullies, landslides and mudflows located further away from the stream are thus often unrevealed. To fill this gap, a transferable remote sensing fusion method called FuSVIPR, for Fusion of Sentinel-2 & Very high resolution Imagery for Pluvial Runoff, is developed to produce damage-detection maps. Based on very high spatial resolution optical imagery (from Pléiades satellites or airborne sensors) combined with 10 m change images from Sentinel-2 satellites, the Random Forest and U-net machine/deep learning techniques are separately trained and compared to locate pluvial flood footprints on the ground at 0.5 m spatial resolution following heavy weather events. In this work, three flash flood events in the Aude and Alpes-Maritimes departments in the South of France are investigated, covering over more than 160 km2 of rural and periurban areas between 2018 and 2020. Pluvial-flood-detection accuracies hover around 75% (with a minimum area detection ratio for annotated ground truths of 25%), and false-positive rates mostly below 2% are achieved on all three distinct events using a cross-site validation framework. FuSVIPR is then further evaluated on the latest devastating flash floods of April 2022 in the Durban area (South Africa), without additional training. Very good agreement with the impact maps produced in the context of the International Charter “Space and Major Disasters” are reached with similar performance figures. These results emphasize the high generalization capability of this method to locate pluvial floods at any time of the year and over diverse regions worldwide using a very high spatial resolution visible product and two Sentinel-2 images. The resulting impact maps have high potential for helping thorough evaluation and improvement of surface water inundation models and boosting extreme precipitation downscaling at a very high spatial resolution.

GPU-Accelerated PCG Method for the Block Adjustment of Large-Scale High-Resolution Optical Satellite Imagery Without GCPs

The precise geo-positioning of high-resolution satellite imagery (HRSI) without ground control points (GCPs) is an important and fundamental step in global mapping, three-dimensional modeling, and so on. In this paper, to improve the efficiency of large-scale bundle adjustment (BA), we propose a combined Preconditioned Conjugate Gradient (PCG) and Graphic Processing Unit (GPU) parallel computing approach for the BA of large-scale HRSI without GCPs. The proposed approach consists of three main components: 1) construction of a BA model without GCPs ; 2) reduction of memory consumption using the Compressed Sparse Row sparse matrix format; and 3) improvement of the computational efficiency by the use of the combined PCG and GPU parallel computing method. The experimental results showed that the proposed method: 1) consumes less memory consumption compared to the conventional full matrix format method; 2) demonstrates higher computational efficiency than the single-core, Ceres-solver and multi-core central processing unit computing methods, with 9.48, 6.82, and 3.05 times faster than the above three methods, respectively; 3) obtains comparable BA accuracy with the above three methods, with image residuals of about 0.9 pixels; and 4) is superior to the parallel bundle adjustment method in the reprojection error.

An Assessment of YOLO Architectures for Oil Tank Detection from SPOT Imagery

Since it can be used to manage and estimate oil reserves, the inventory of oil tanks is essential for both the economy and the military applications. Considering oil tanks contain valuable materials required for transportation and industrial production, they are a significant type of target. Oil tank detection techniques have several uses, including monitoring disasters, preventing oil leaks, designing cities, and assessing damage. Huge amount of satellite imagery has recently been available and it is used in both the military and civil applications. The new spaceborne sensors' higher resolution enables the detection of targeted objects. Therefore, remote sensing instruments provide ideal tools for oil tank detection task. Conventional approaches for oil tank detection from high resolution remote sensing imagery generally relies on geometric shape, structure, contract differences and color information of the boundary or hand-crafted features. However, these methods come along with vulnerabilities and hence it can be challenging to obtain accurate detection in the presence of a number of disturbance elements, particularly a wide range of colours, size variations, and the shadows that view angle and illumination create. Therefore, deep learning-based methods can provide a big advantage for solution of this task. In this regard, this study employs four YOLO models namely YOLOv5, YOLOX, YOLOv6 and YOLOv7 for oil tank detection from high-resolution optical imagery. Our results show that YOLOv7 and YOLOv5 architectures provide more accurate detections with mean average precision values of 68.11% and 69.69%, respectively. The experiments and visual inspections reveal efficiency, generalization and transferability of these models.