High-resolution Optical Satellite Images Research Articles

Extraction of Coastal Levees Using U-Net Model with Visible and Topographic Images Observed by High-Resolution Satellite Sensors.

Coastal levees play a role in protecting coastal areas from storm surges and high waves, and they provide important input information for inundation damage simulations. However, coastal levee data with uniformity and sufficient accuracy for inundation simulations are not always well developed. Against this background, this study proposed a method to extract coastal levees by inputting high spatial resolution optical satellite image products (RGB images, digital surface models (DSMs), and slope images that can be generated from DSM images), which have high data availability at the locations and times required for simulation, into a deep learning model. The model is based on U-Net, and post-processing for noise removal was introduced to further improve its accuracy. We also proposed a method to calculate levee height using a local maximum filter by giving DSM values to the extracted levee pixels. The validation was conducted in the coastal area of Ibaraki Prefecture in Japan as a test area. The levee mask images for training were manually created by combining these data with satellite images and Google Street View, because the levee GIS data created by the Ibaraki Prefectural Government were incomplete in some parts. First, the deep learning models were compared and evaluated, and it was shown that U-Net was more accurate than Pix2Pix and BBS-Net in identifying levees. Next, three cases of input images were evaluated: (Case 1) RGB image only, (Case 2) RGB and DSM images, and (Case 3) RGB, DSM, and slope images. Case 3 was found to be the most accurate, with an average Matthews correlation coefficient of 0.674. The effectiveness of noise removal post-processing was also demonstrated. In addition, an example of the calculation of levee heights was presented and evaluated for validity. In conclusion, this method was shown to be effective in extracting coastal levees. The evaluation of generalizability and use in actual inundation simulations are future tasks.

Ice floe segmentation and floe size distribution in airborne and high-resolution optical satellite images: towards an automated labelling deep learning approach

Abstract. Floe size distribution (FSD) has become a parameter of great interest in observations of sea ice because of its importance in affecting climate change, marine ecosystems, and human activities in the polar ocean. A most effective way to monitor FSD in the ice-covered regions is to apply image processing techniques to airborne and satellite remote sensing data, where the segmentation of individual ice floes is a challenge in obtaining FSD from remotely sensed images. In this study, we adopt a deep learning (DL) semantic segmentation network to fast and adaptive implement the task of ice floe instance segmentation. In order to alleviate the costly and time-consuming data annotation problem of model training, classical image processing technique is applied to automatically label ice floes in local-scale marginal ice zone (MIZ). Several state-of-the-art (SoA) semantic segmentation models are then trained on the labelled MIZ dataset and further applied to additional large-scale optical Sentinel-2 images to evaluate their performance in floe instance segmentation and to determine the best model. A post-processing algorithm is also proposed in our work to refine the segmentation. Our approach has been applied to both airborne and high-resolution optical (HRO) satellite images to derive acceptable FSDs at local and global scales.

Effect of Urbanization Through Land Coverage Classification

AbstractLand cover changes from rural or natural habitats to impervious surfaces that support various human activities are called urbanization. The evaluation of policy alternatives for future expansion and the promotion of sustainable urban development therefore place great emphasis on the mapping, analysis, and monitoring of land cover changes in metropolitan environments. By assessing changes in relevant environmental indicators at a range of scales, from local to regional, this study seeks to determine the rate of urbanization and analyze its impact on the environment in and around two key cities in Tamilnadu. Chennai and Madurai are the urban areas under investigation. The results are based on classifications of medium to high spatial resolution optical satellite images from 1995 to 2020. Various classification approaches with combinations of spectral, shape, and texture input information were combined to achieve high classification accuracy. Based on the categories, environmental indicators were created and used to calculate the impact of urbanization on the environment. The findings show that each research area experienced varying degrees of urban expansion and environmental impact. Urban areas grew tenfold between 1995 and 2015, according to a study comparing Chennai and Madurai. To evaluate the spatiotemporal dynamics of urbanization and its environmental impact in various metropolitan regions, this research shows the value of combining urban and environmental indicators with remote sensing data.

GEOMETRIC PROCESSING OF VERY HIGH-RESOLUTION SATELLITE IMAGERY: QUALITY ASSESSMENT FOR 3D MAPPING NEEDS

Abstract. In recent decades, the geospatial domain has benefitted from technological advances in sensors, methodologies, and processing tools to expand capabilities in mapping applications. Airborne techniques (LiDAR and aerial photogrammetry) generally provide most of the data used for this purpose. However, despite the relevant accuracy of these technologies and the high spatial resolution of airborne data, updates are not sufficiently regular due to significant flight costs and logistics. New possibilities to fill this information gap have emerged with the advent of Very High Resolution (VHR) optical satellite images in the early 2000s. In addition to the high temporal resolution of the cost-effective datasets and their sub-meter geometric resolutions, the synoptic coverage is an unprecedented opportunity for mapping remote areas, multi-temporal analyses, updating datasets and disaster management. For all these reasons, VHR satellite imagery is clearly a relevant study for National Mapping and Cadastral Agencies (NMCAs). This work, supported by EuroSDR, summarises a series of experimental analyses carried out over diverse landscapes to explore the potential of VHR imagery for large-scale mapping.

Freely accessible inventory and spatial distribution of large-scale landslides in Xianyang City, Shaanxi Province, China

In this study, we used high-resolution optical satellite images on the Google Earth platform to map large-scale landslides in Xianyang City, Shaanxi Province, China. After mapping, a comprehensive and detailed large-scale landslide inventory that contains 2924 large-scale landslides was obtained. We analyzed the spatial distribution of landslides with seven influencing factors, including elevation, slope angle, aspect, curvature, lithology, distance to a river, and distance to the fault. Landslide Number, Landslide Area, Landslide Number Density (LND), and Landslide Area Percentage (LAP) were selected as indexes for the spatial distribution analysis. The results show that the number and area of landslides in the elevation range of 1000–1200 ​m is the highest. The highest number of landslides was observed in the slope angle of 25°–30°. North-facing slopes are prone to sliding. The area and number of landslides are the largest when the slope curvature ranges from −1.28 to 0. The LND and LAP reach their maxima when the slope curvature is less than −2.56. Areas covered by the Tertiary stratum with weakened fine-grained sandstone and siltstone show the highest LND and LAP values. Regarding distance to a river, the LAP peaks in the range of 300–600 ​m, whereas the LND peaks in an area larger than 2100 ​m. The values of LND and LNP rise as the distance from the faults increases, except for the locations 30 ​km away from active faults. This phenomenon is because active faults in this area pass through the plain areas, while landslides mostly occur in mountainous areas. The cataloging of landslide development in Xianyang City provides a significant scientific foundation for future research on landslides. In addition, the spatial distribution results are useful for landslide hazard prevention decisions and provide valuable references in this area.

Data Fusion for Satellite-Derived Earth Surface: The 2021 Topographic Map of Etna Volcano

We present a new automatic procedure for updating digital topographic data from multi-source satellite imagery, which consists in the production of Digital Surface Models (DSMs) from high resolution optical satellite images, followed by a context-aware fusion that exploits the complementary characteristics of the multi-source DSMs. The fused DSM minimizes blunders and artifacts due to occlusions (e.g., the presence of clouds, snow or ash plumes) in the source images, resulting in improved accuracy and quality versus those that are not merged. The procedure has been tested to produce the 2021 digital topography of Mt Etna, whose summit area is constantly changing and shows the new peak of 3347 m on the north rim of the South East Crater. We also employ the 2021 DSM to measure the volcanic deposits emplaced in the last five years, finding about 120 million cubic meters, with a yearly average volume of about 24 million cubic meters in agreement with the large eruptive rates registered at Mt Etna since the nineteen seventies. The flexibility and modularity of the presented procedure make it easily exportable to other environmental contexts, allowing for a fast and frequent reconstruction of topographic surfaces even in extreme environments.

Ice Aprons in the Mont Blanc Massif (Western European Alps): Topographic Characteristics and Relations with Glaciers and Other Types of Perennial Surface Ice Features

Ice aprons are poorly studied and not well-defined thin ice bodies adhering to high altitude steep rock faces, but are present in most Alpine-type high mountain environments worldwide. This study aims to precisely define ice aprons based on a detailed analysis of their topographical characteristics in the Mont Blanc massif (western European Alps). For this, we accurately identified and precisely mapped 423 ice aprons using a combination of high-resolution optical satellite images from 2019. To better understand their relationship with other types of glaciers, especially the steep slope glaciers and other surface ice bodies, we built a detailed inventory at the scale of the massif that incorporates nine different types of perennial surface ice bodies. In addition, an analysis using different topographic factors helped us to better understand the preferred locations of the ice aprons. We show that they predominantly occur on west-oriented steep and topographically rugged rock slopes above the local Equilibrium Line Altitude (~3200 m a.s.l.), with concave profile curvatures around them that facilitate snow accumulation. They are also found in areas underlain by permafrost. The extensive inventory also helped us to identify different types of ice aprons based on their relationships with glaciers/ice bodies. The analysis shows that ice aprons existing at the headwall of large glaciers above a bergschrund are the most dominant ice apron type in the study area, with ~82% of the total.

Two public inventories of landslides induced by the 10 June 2022 Maerkang Earthquake swarm, China and ancient landslides in the affected area

The June 10, 2022 Ms5.8, Ms6.0 and Ms5.2 earthquake swarm of Maerkang, Sichuan, China has triggered a lot of landslides. The primary purpose of this paper is to prepare a detailed coseismic landslides inventory and ancient landslide inventory in the earthquake swarm-affected area. Based on pre- and post-quake high-resolution optical satellite images (planet), we delineated 650 individual coseismic landslides in the VI-intensity area and above, occupying an area of about 1.2 ​km2. The largest landslide covers about 70217 ​m2 in horizontal projection, while the smallest one is just 81 ​m2. In addition, based on the series of high-precise images on the Google Earth Platform, 759 ancient landslides were also delineated with an area of 117 ​km2. The maximum area reaches about 2 ​km2 and the minimum is 2580 ​m2. The two inventories provide a basis for landslide spatial distribution and hazard assessment, disaster prevention and mitigation of earthquake-triggered landslides in similar regions. We opened these two landslide inventories to facilitate researchers to carry out more in-depth work.

A COMPARATIVE STUDY FOR BUILDING SEGMENTATION IN REMOTE SENSING IMAGES USING DEEP NETWORKS: CSCRS ISTANBUL BUILDING DATASET AND RESULTS

Abstract. Building semantic segmentation is an exceedingly important issue in the field of remote sensing. A new building dataset as created consisting of very high-resolution optical satellite images provided by the Center for Satellite Communications and Remote Sensing (CSCRS). The imagery is obtained by Pleiades satellite and have a resolution of 0.5 meters. Segmentation results have been obtained using post-FCN architectures. Architectures examined in this work fall under one of few categories. The first category is Encoder-Decoder Network: an encoder that reduces the spatial resolution of the data and a decoder that recreates the lower resolution result of the encoder and upsamples it. The second category is Feature Pyramid Network, in this type of network scene information is aggregated across pyramid structures which produce more comprehensive results. The third category is Dilated Network, due to its atrous structure, which can calculate any layer at any desired resolution, with the presence of holes in the filter. The final category is Attention-Based Network, in these networks, certain aspects of the data are emphasized while other aspects are ignored. After this work, it can be seen that according to several metrics Dilated and Attention-Based Networks perform better than their counterparts. As a result of the training of 100 epochs with the data set in architectures belonging to Dilated and Attention-Based Networks, IoU values above 0.90 were obtained.

IMPROVING PAIRWISE DSM WITH 3SGM: A SEMANTIC SEGMENTATION FOR SGM USING AN AUTOMATICALLY REFINED NEURAL NETWORK

Abstract. The amount of very high resolution optical satellite images at our disposal is continuously increasing. Besides, associated satellite programs often come with high revisit rates and geometric properties that allow for either opportunistic or by-design 3D stereo reconstruction. Digital Surface Models (DSM) computed from these satellite images offer new possibilities. In the past, the high revisit rate has largely benefited glacier monitoring studies. Now, DSM with increased resolution provided on urban areas can be used for smart city applications as well. However, most of these require 3D modeling of buildings with level of details ranging from 0 to 2. This is where the need for better reconstructed buildings inside DSM arises. Indeed, building edges and corners tend to be smoothed and softened by the stereo matching step of a DSM computation pipeline. This undesired behavior can mostly be linked to the difficult task of optimizing the Disparity Space Image, thus finding good balance between smoothing untextured areas while conserving sharp discontinuities where needed. In this paper, we show how the optimization can benefit from an input building semantic segmentation. We also provide a method to create it from a very high satellite image in epipolar geometry using a convolutional neural network. To help our network generalize well on unseen areas we propose an interactive learning method based on clicked annotations. Eventually, we show that annotations can be automatically created, hence removing the need for an operator and making our solution suitable for operational conditions.

Monitoring Potential Geological Hazards with Different InSAR Algorithms: The Case of Western Sichuan

In recent years, the number of geological disasters in Sichuan Province has significantly increased due to the influence of earthquakes and extreme climate, as well as the disturbance to the geological environment by human activities. Thus, geological disaster monitoring is particularly important, which can provide some scientific basis for disaster prevention and reduction. In this paper, the interferometric synthetic aperture radar (InSAR) technology was introduced to monitor potential geological hazards, taking parts of Dujiangyan City, Wenchuan County, and Mao County in Sichuan Province, China as examples. Firstly, the data such as Sentinel-1A Terrain Observation with Progressive Scans (TOPS) Synthetic Aperture Radar (SAR) images and Precision Orbit Determination (POD) precise orbit ephemerides from 2018 to 2020, high-resolution optical satellite images and Digital Elevation Model (DEM) were collected. Secondly, the Differential InSAR (D-InSAR), Persistent Scatterer InSAR (PS-InSAR), Small Baseline Subset InSAR (SBAS-InSAR), Offset-Tracking, and Distributed Scatterer InSAR (DS-InSAR) algorithms were used to invert the surface deformation of the study area. Thirdly, according to the deformation results obtained by experiments, we used GF-1 and GF-2 optical images as a reference and combine the experimental results of InSAR algorithms to delineate the areas affected by geological disasters. A total of 49 geological disaster areas were obtained, mainly including landslides, collapses, and debris flow. Through field verification, the overall accuracy rate of InSAR deformation monitoring achieved 69.23%, and the accuracy rate of new potential hazards monitoring reached 63.64%. Among all InSAR methods, the DS-InSAR method outperformed and monitored the geological disaster areas well. Finally, the study area was divided into three elevation intervals and the applicability of different InSAR algorithms in different elevation intervals was discussed.

MF-Dfnet: a deep learning method for pixel-wise classification of very high-resolution remote sensing images

ABSTRACT Semantic segmentation of high-resolution remote sensing images is very important. However, the targets in the high-resolution optical satellite images are always various in size, which lead to multiscale problems resulting in difficulty of locating and identifying the target. High-resolution remote sensing is more complex than natural phenomena; this leads to false alarms due to a greater intraclass inconsistency. Thus, the pixel-wise classification of high-resolution remote sensing images becomes challenging. Aiming at the above problems, we propose a multiscale feature and discriminative feature network (MF-DFNet). We introduce the hierarchical-split block (HSB) and the residual receptive field block module (RRFBM) to extract multiscale information to address multiscale problems. We also introduce a foreground-scene relation module to enhance the discrimination of features and deal with the false alarm phenomenon. In addition, the channel attention block (CAB) is introduced to select more discriminative features. We use two publicly available remote sensing image datasets (Vaihingen and Massachusetts building) for the experiments in this paper. Compared to current advanced models, our results show that MF-DFNet achieves state-of-the-art performance and can effectively improve the integrity and correctness of semantic segmentation in high-resolution remote sensing images.

Predicting road blockage due to building damage following earthquakes

Transportation infrastructure supports the social and economic activities of communities. One of the impacts of roads’ disruption is the obstruction of emergency services (e.g., ambulance, firefighting, evacuation). Furthermore, the recovery process of a community following an extreme event (e.g., a natural hazard) depends on the functionality of the transportation infrastructure. Therefore, conducting a risk and resilience analysis of transportation infrastructure is critical to help communities minimize the initial impact of hazards and promote a rapid recovery. Current approaches model the probability of road blockage due to building damage using high-resolution optical satellite images and aerial photographs collected after past events. However, the data used by these methods are limited, and few data have been collected before 2010. Besides, data may not be available for specific regions that have not experienced recent earthquakes. Thus, a probabilistic predictive method is needed for risk and resilience analysis of roads. This paper proposes a probabilistic model using the data from the 2010 Haiti Earthquake and calibrated by Bayesian approach to predict the debris distance from undamaged buildings (e.g., the distance debris can reach from the footprint of the undamaged building). The model is then used to construct fragility curves that give the conditional probability of road blockage at a given road section for a given seismic intensity. The proposed model considers the relevant factors affecting the road blockage probability, including building types, damage level, and road characteristics. The probability of road blockage at a given road section is estimated for the four general road section types, considering buildings on only one side of the road or both sides, and with or without a raised traffic median. The probability of road blockage for an entire road is then calculated by system and parallel reliability analysis. The proposed models apply to any general urban area without the dependence on historical data from past earthquakes.

Diffuse Deformation and Surface Faulting Distribution from Submetric Image Correlation along the 2019 Ridgecrest, California, Ruptures

ABSTRACT The 2019 Mw 6.4 and 7.1 Ridgecrest, California, earthquake sequence (July 2019) ruptured consecutively a system of high-angle strike-slip cross faults (northeast- and northwest-trending) within 34 hr. The complex rupture mechanism was illuminated by seismological and geodetic data, bringing forward the issue of the interdependency of the two fault systems both at depth and at the surface, and of its effect on the final surface displacement pattern. Here, we use high-resolution (WorldView and Pleiades) optical satellite image correlation to measure the near-fault horizontal and vertical surface displacement fields at 0.5 m ground resolution for the two earthquakes. We point out significant differences with previous geodetic- and geologic-based measurements, and document the essential role of distributed faulting and diffuse deformation in producing the observed surface displacement patterns. We derive strain fields from the horizontal displacement maps, and highlight the predominant role of rotation and shear strain in the surface rupture process. We discuss the segmentation of the rupture based on the fault geometry and along-strike slip variations. We also image several northeast-trending faults with similar orientation to the deeply embedded shear fabric identified in aftershock studies, and show that these cross faults are present all along the rupture, including at a scale <100 m. Finally, we compare our results to kinematic slip inversions, and show that the surface diffuse deformation is primarily associated with areas of shallow slip deficit; however, this diffuse deformation cannot be explained using elastic modeling. We conclude that inelastic processes play an important role in contributing to the total surface deformation associated with the 2019 Ridgecrest sequence.

A deep learning method for building height estimation using high-resolution multi-view imagery over urban areas: A case study of 42 Chinese cities

Knowledge of building height is critical for understanding the urban development process. High-resolution optical satellite images can provide fine spatial details within urban areas, while they have not been applied to building height estimation over multiple cities and the feasibility of mapping building height at a fine scale (< 5 m) remains understudied. Multi-view satellite images can describe vertical information of buildings, due to the inconsistent response of buildings (e.g., spectral and structural variations) to different viewing angles, but they have not been employed to deep learning-based building height estimation. In this context, we introduce high-resolution ZY-3 multi-view images to estimate building height at a spatial resolution of 2.5 m. We propose a multi-spectral, multi-view, and multi-task deep network (called M3Net) for building height estimation, where ZY-3 multi-spectral and multi-view images are fused in a multi-task learning framework. A random forest (RF) method using multi-source features is also carried out for comparison. We select 42 Chinese cities with diverse building types to test the proposed method. Results show that the M3Net obtains a lower root mean square error (RMSE) than the RF, and the inclusion of ZY-3 multi-view images can significantly lower the uncertainty of building height prediction. Comparison with two existing state-of-the-art studies further confirms the superiority of our method, especially the efficacy of the M3Net in alleviating the saturation effect of high-rise building height estimation. Compared to the vanilla single/multi-task models, the M3Net also achieves a lower RMSE. Moreover, the spatial-temporal transferability test indicates the robustness of the M3Net to imaging conditions and building styles. The test of our method on a relatively large area (covering about 14,120 km2) further validates the scalability of our method from the perspectives of both efficacy and quality. The source code will be made available at https://github.com/lauraset/BuildingHeightModel.

DISTRIBUTION AND EVOLUTION OF ICE APRONS IN A CHANGING CLIMATE IN THE MONT-BLANC MASSIF (WESTERN EUROPEAN ALPS)

Abstract. Ice Apron (IA) is a poorly studied ice feature, commonly existing in all the world’s major mountain regions. This study aims to map the locations of the IAs in the Mont Blanc massif (MBM), making use of the very high-resolution optical satellite images from 2001, 2012 and 2019. 423 IAs were identified and accurately delineated in the MBM on the images from 2019, and their topographic characteristics were studied. We generated our own Digital Elevation Model (DEM) at 4 m resolution since the freely available products predominantly suffer from significant inconsistencies, especially in steep mountain areas. Results show that most IAs exist at elevations above the regional Equilibrium Line Altitude (ELA), on steep slopes, on concave surfaces, on northern and southern aspects and on the most rugged terrains. They are also commonly associated with steep slope glaciers as 85% of them occur on these glaciers’ headwalls. A comparison between 2001 and 2019 shows that IAs have lost around 29% of their area over a period of 18 years. This is significant and the rate of area loss is very alarming in comparison with the larger glacier bodies. We also studied the effect of topographic parameters on the area loss. We found that topographic factors like slope, aspect, curvature, elevation and Terrain Ruggedness Index (TRI) strongly influence the rate of area loss of IAs.

A hierarchical object detection method in large-scale optical remote sensing satellite imagery using saliency detection and CNN

ABSTRACT Detecting geospatial objects, especially small, time-sensitive targets such as airplanes and ships in cluttered scenes, is a substantial challenge in large-scale, high-resolution optical satellite images. Directly detecting targets in countless image blocks results in higher false alarms and is also inefficient. In this paper, we introduce a hierarchical architecture to quickly locate related areas and detect these targets effectively. In the coarse layer, we use an improved saliency detection model that utilizes geospatial priors and multi-level saliency features to probe suspected regions in broad and complicated remote sensing images. Then, in the fine layer of each region, an efficacious end-to-end neural network that predicts the categories and locations of the objects is adopted. To improve the detection performance, an enhanced network, adaptive multi-scale anchors, and an improved loss function are designed to overcome the great diversity and complexity of backgrounds and targets. The experimental results obtained for both a public dataset and our collected images validated the effectiveness of our proposed method. In particular, for large-scale images (more than 500 km2), the adopted method far surpasses most unsupervised saliency models in terms of the performance in region saliency detection and can quickly detect targets within 1 minute, with 95.0% recall and 93.2% precision rates on average.

Robust Automatic Generation of True Orthoimages From Very High-Resolution Panchromatic Satellite Imagery Based on Image Incidence Angle for Occlusion Detection

Orthoimages have proven to provide useful reference data for a variety of geospatial applications. Their improved versions, true orthoimages, are even more valuable as they have been adjusted also for high objects. Even though true orthoimages have been-over the last decades-generated mainly from aerial images, we believe there is a need to produce true orthoimages also from satellite images. The aim of this article is to present the already developed automatic procedure for generating true orthoimages from very high-resolution optical satellite images. The automatic workflow consists of five modules, starting with the extraction of ground control points, through the geometric processing of image blocks, occlusion detection, orthorectification, and finally the generation of true orthoimages. The orthorectification module uses a high-resolution digital surface model derived from laser scanning data. Occlusion detection is performed using a ray-tracing method based on the incidence angle, which is suitable for images acquired by satellites and processed with rational function models. The usability of the workflow was demonstrated with the test of six Pléiades images that were acquired on different dates and with different viewing angles. The research presents the accuracy of the geometric correction and the quality of the true orthoimage generated with different numbers of images. The results show that satellite-based true orthoimages are suitable for the extraction of spatial information that can be used directly by the end-users.

Building Identification in Urban Areas using VHR Satellite Imagery

Satellite imagery is one of the emerging technologies which is extensively utilized in various applications such as detection/extraction of man-made structures, monitoring of sensitive areas, creating graphic maps etc. The main approach here is the automated detection of buildings from very high resolution (VHR) optical satellite images. Initially the shadow and the building region are investigated and building extraction is mainly focused. Once all the landscape is collected a trimming process is done so as to eliminate the landscapes that may occur due to nonbuilding objects. Finally the label method is used to extract the building regions. The label method may be altered for efficient building extraction. The images used for the analysis are the ones which are extracted from the sensors having resolution less than 1 meter (VHR). This method provides an efficient way to produce good results. The additional overhead of mid processing is eliminated without compromising the quality of the output to ease out processing steps required and time consumed in the same. /

H-YOLO: A Single-Shot Ship Detection Approach Based on Region of Interest Preselected Network

Ship detection from high-resolution optical satellite images is still an important task that deserves optimal solutions. This paper introduces a novel high-resolution image network-based approach based on the preselection of a region of interest (RoI). This pre-selected network first identifies and extracts a region of interest from input images. In order to efficiently match ship candidates, the principle of our approach is to distinguish suspected areas from the images based on hue, saturation, value (HSV) differences between ships and the background. The whole approach is the basis of an experiment with a large ship dataset, consisting of Google Earth images and HRSC2016 datasets. The experiment shows that the H-YOLO network, which uses the same weight training from a set of remote sensing images, has a 19.01% higher recognition rate and a 16.19% higher accuracy than applying the you only look once (YOLO) network alone. After image preprocessing, the value of the intersection over union (IoU) is also greatly improved.