GF-1 Images Research Articles

Harmonizing Multi-Source Remote Sensing Images for Summer Corn Growth Monitoring

Continuous monitoring of crop growth status using time-series remote sensing image is essential for crop management and yield prediction. The growing season of summer corn in the North China Plain with the period of rain and hot, which makes the acquisition of cloud-free satellite imagery very difficult. Therefore, we focused on developing image datasets with both a high temporal resolution and medium spatial resolution by harmonizing the time-series of MOD09GA Normalized Difference Vegetation Index (NDVI) images and 30-m-resolution GF-1 WFV images using the improved Kalman filter model. The harmonized images, GF-1 images, and Landsat 8 images were then combined and used to monitor the summer corn growth from 5th June to 6th October, 2014, in three counties of Hebei Province, China, in conjunction with meteorological data and MODIS Evapotranspiration Data Set. The prediction residuals ( Δ P R K ) in NDVI between the GF-1 observations and the harmonized images was in the range of −0.2 to 0.2 with Gauss distribution. Moreover, the obtained phenological curves manifested distinctive growth features for summer corn at field scales. Changes in NDVI over time were more effectively evaluated and represented corn growth trends, when considered in conjunction with meteorological data and MODIS Evapotranspiration Data Set. We observed that the NDVI of summer corn showed a process of first decreasing and then rising in the early growing stage and discuss how the temperature and moisture of the environment changed with the growth stage. The study demonstrated that the synthesized dataset constructed using this methodology was highly accurate, with high temporal resolution and medium spatial resolution and it was possible to harmonize multi-source remote sensing imagery by the improved Kalman filter for long-term field monitoring.

Tree species and age groups classification based on GF-2 image

Tree species and age groups classification based on GF-2 image

Variation of leaf area index estimation in forests based on remote sensing images of different spatial scales.

There are several important issues in quantitative remote sensing and product authenticity testing, including how well do the ground measurement points represent the remote sensing pixels, how to obtain the relative truth value of pixels, and how much spatial resolution can truly reflect fore-st leaf area index (LAI). In this study, the measured space scope of two plant canopy analyzers [LAI-2200 and tracing radiation and architecture of canopies (TRAC)] were calculated, which were combined with remote sensing images with three different spatial resolutions: GF-2 with 4.1 m spatial resolution, the Sentinel-2 with 10 m spatial resolution, and Landsat-8 OLI with 30 m spatial resolution, to get the relative true value of pixel at each scale. Under the condition of keeping the real observed area consistent with that obtained by remote sensing, the effects of different spatial resolution images for estimating forest LAI were compared and analyzed based on the unary exponential and multiple regression statistical models. Moreover, the optimal statistical models of the three images were tested on 30 m and 100 m scales and the spatial representation of dataset were evaluated, to find the most suitable scale for the description of forest LAI in the study area. The results showed that high resolution did not necessarily fully reflect LAI of forests. The statistical model based on three kinds of resolution images could well estimate forest LAI. Among the three models, the model based on the Sentinel-2 image had the highest accuracy, and the one based on the GF-2 images had the lowest. The test results at 30 and 100 m scales indicated that the forest LAI was overestimated by the GF-2 inversion model, and underestimated by the Landsat-8 inversion model. The statistical model based on Sentinel-2 could well estimated forest LAI in the study area.

Optimization of Fusion Method for GF-2 Satellite Remote Sensing Images based on the Classification Effect

With the successful launch of China’s GF series satellites, it is more important to study the image data quality, the adaptability of processing method and information extraction method. The panchromatic and multi-spectral data which is based on the GF-2 images data of Chinese sub-meter high-resolution remote sensing satellite is fused by PCA, Pansharp, Gram-Schmidt and NNDiffuse fusion. Then, the quality of the fusion images were evaluated subjectively and objectively. In order to evaluate the applicability of different classification algorithms to the classification, the object-oriented classification algorithm which is based on machine learning algorithm, such as KNN, SVM and Random Trees were used to classify the different GF-2 fusion images. The results showed that: (1) The best visual effect of GF-2 fusion image was the Pansharp fusion image; The quantitative evaluation results showed that the brightness and information retention of Gram-Schmidt fusion image was the best，while the Pansharp fusion image had the highest correlation with the original multi-spectral image; the NNDiffuse fusion image had the highest clarity, and the PCA fusion image quantitative evaluation effect was the worst; (2) According to the applicability analysis of the fusion images based on different classification algorithms with features information extraction, it could be seen that the NNDiffuse fusion method was used for the fusion of GF-2 image data, and the classification of the fusion images was more suitable by using KNN or Random Trees classification algorithm.

$\mathcal{R}^2$ -CNN: Fast Tiny Object Detection in Large-Scale Remote Sensing Images

Recently, the convolutional neural network has brought impressive improvements for object detection. However, detecting tiny objects in large-scale remote sensing images still remains challenging. First, the extreme large input size makes the existing object detection solutions too slow for practical use. Second, the massive and complex backgrounds cause serious false alarms. Moreover, the ultratiny objects increase the difficulty of accurate detection. To tackle these problems, we propose a unified and self-reinforced network called remote sensing region-based convolutional neural network ($\mathcal{R}^2$-CNN), composing of backbone Tiny-Net, intermediate global attention block, and final classifier and detector. Tiny-Net is a lightweight residual structure, which enables fast and powerful features extraction from inputs. Global attention block is built upon Tiny-Net to inhibit false positives. Classifier is then used to predict the existence of targets in each patch, and detector is followed to locate them accurately if available. The classifier and detector are mutually reinforced with end-to-end training, which further speed up the process and avoid false alarms. Effectiveness of $\mathcal{R}^2$-CNN is validated on hundreds of GF-1 images and GF-2 images that are 18 000 $\times$ 18 192 pixels, 2.0-m resolution, and 27 620 $\times$ 29 200 pixels, 0.8-m resolution, respectively. Specifically, we can process a GF-1 image in 29.4 s on Titian X just with single thread. According to our knowledge, no previous solution can detect the tiny object on such huge remote sensing images gracefully. We believe that it is a significant step toward practical real-time remote sensing systems.

Estimating Pylon Height Using Differences in Shadows Between GF-2 Images

Information on height of pylon is important when planning the routes of unmanned aerial vehicles. This paper proposes a new automatic method for estimating pylon height using shadow differences between GF-2 satellite images taken at different times. Initially, the spectral distribution features and triangular shape of pylon shadows are used for shadow detection to enhance the difference from other land objects using images taken at a greater sun elevation angle. Then, the difference in the shadows of the same pylon in images taken at different times and shadow-imaging principles are used to estimate the vertical of the pylon top to avoid interference from complex land objects. Finally, the height of the pylon is approximated using the distance from the vertical projection point to the shadow point of the pylon top and sun elevation angle. GF-2 images were selected and included five pylons in the target area. The average error of the estimated height was 1.56 m, and the relative error was only 2.74%.

Associations between Road Density, Urban Forest Landscapes, and Structural-Taxonomic Attributes in Northeastern China: Decoupling and Implications

A better understanding on the associations between road density (RD), urban forest structural-taxonomic attributes, and landscape metrics is vital for forest ecological service evaluations and suitable management in sprawling urban areas with increasing road networks. We chose Harbin, a fast growing provincial capital city in northeast China, as a case study to address this issue. We utilized ArcGIS software (Esri, version 10.0; Redlands, CA, USA) and FRAGSTATS (V4.2.589) to digitize GF-1 images (Gaofen No.1 remote sensing images) to acquire road net characteristic information and landscape metrics of urban forests in Harbin. Together with forest structural-taxonomic attributes from a stratified random sampling survey, statistical methods such as an analysis of variance, a regression analysis, and a redundancy analysis were used to determine the road-dependent differences and to decouple the associations between them. The results indicated that road area percentages, road length/imperious surface area (ISA) ratios, road area/ISA ratios, and road cross-points sharply increased from low to heavy RD areas. This road intensification was strongly associated with increased urban forest area, patch density, and diverse patch shapes; smaller tree sizes, lower tree densities, and diverse tree species compositions were generally observed. Redundancy-based variation partitioning showed that part of the variations in structural-taxonomic attributes of forests could be explained by road intensity characteristics. In low RD (0–1.5 km/km2) regions, the road characteristics significantly affected forest characteristics (Shannon Wiener diversity index, species richness, and evenness index); however, such associations weakened with increasing forest landscape-related associations in medium to heavy RD (1.5–6 km/km2) regions. Our findings highlighted that road development is strongly associated with forest characteristics in Harbin city, and RD-dependent forest landscape regulating management could favor the maximization of forest ecological services that are related to structural and species identities.

Farmland Extraction from High Spatial Resolution Remote Sensing Images Based on Stratified Scale Pre-Estimation

Extracting farmland from high spatial resolution remote sensing images is a basic task for agricultural information management. According to Tobler’s first law of geography, closer objects have a stronger relation. Meanwhile, due to the scale effect, there are differences on both spatial and attribute scales among different kinds of objects. Thus, it is not appropriate to segment images with unique or fixed parameters for different kinds of objects. In view of this, this paper presents a stratified object-based farmland extraction method, which includes two key processes: one is image region division on a rough scale and the other is scale parameter pre-estimation within local regions. Firstly, the image in RGB color space is converted into HSV color space, and then the texture features of the hue layer are calculated using the grey level co-occurrence matrix method. Thus, the whole image can be divided into different regions based on the texture features, such as the mean and homogeneity. Secondly, within local regions, the optimal spatial scale segmentation parameter was pre-estimated by average local variance and its first-order and second-order rate of change. The optimal attribute scale segmentation parameter can be estimated based on the histogram of local variance. Through stratified regionalization and local segmentation parameters estimation, fine farmland segmentation can be achieved. GF-2 and Quickbird images were used in this paper, and mean-shift and multi-resolution segmentation algorithms were applied as examples to verify the validity of the proposed method. The experimental results have shown that the stratified processing method can release under-segmentation and over-segmentation phenomena to a certain extent, which ultimately benefits the accurate farmland information extraction.

Automatic Ship Detection in Optical Remote Sensing Images Based on Anomaly Detection and SPP-PCANet

Automatic ship detection technology in optical remote sensing images has a wide range of applications in civilian and military fields. Among most important challenges encountered in ship detection, we focus on the following three selected ones: (a) ships with low contrast; (b) sea surface in complex situations; and (c) false alarm interference such as clouds and reefs. To overcome these challenges, this paper proposes coarse-to-fine ship detection strategies based on anomaly detection and spatial pyramid pooling pcanet (SPP-PCANet). The anomaly detection algorithm, based on the multivariate Gaussian distribution, regards a ship as an abnormal marine area, effectively extracting candidate regions of ships. Subsequently, we combine PCANet and spatial pyramid pooling to reduce the amount of false positives and improve the detection rate. Furthermore, the non-maximum suppression strategy is adopted to eliminate the overlapped frames on the same ship. To validate the effectiveness of the proposed method, GF-1 images and GF-2 images were utilized in the experiment, including the three scenarios mentioned above. Extensive experiments demonstrate that our method obtains superior performance in the case of complex sea background, and has a certain degree of robustness to external factors such as uneven illumination and low contrast on the GF-1 and GF-2 satellite image data.

Building extraction method based on the spectral index for high-resolution remote sensing images over urban areas

With the advent of high-resolution remote sensing images, automatic building extraction methods play a more important role in rapidly acquiring information about large-scale buildings. Although advanced building extraction methods have been introduced to improve building extraction results, these methods involve complex processing and high-computation times. We put forward an effective method to extract building information, based on a proposed spectral building index. The basic idea of the spectral building index is to generate an optimized index based on the computation and analysis of spectral bands, which are beneficial for image enhancement for buildings in images. Aiming at the band number of the multispectral satellite images in high-resolution remote sensing images, we propose two spectral indices for building extraction, including the normalized spectral building index (NSBI) and the difference spectral building index (DSBI). Considering the current spectral band number of high-resolution satellite images, NSBI is suited for satellite images with eight spectral bands, whereas DSBI is suited for satellite images with four spectral bands. The proposed method is validated on various high-resolution images including WorldView-2, GF-1, GF-2, and QuickBird images with 13 experiment datasets, as well as a detailed comparison to the state-of-the-art methods, such as the morphological building index, nonhomogeneous feature difference, and building condition index. The experimental results reveal that the proposed method can achieve promising results for different building conditions, such as regular and irregular building shapes and concrete and metal roofing materials. The average overall accuracy was over 85% with low-time consumption (<1 s).

Building Extraction from High-Resolution Remotely Sensed Imagery Based on Multi-subgraph Matching

Building extraction is still a difficult issue in the field of remote sensing. In order to extract the buildings with similar structures efficiently, an algorithm based on multi-subgraph matching is proposed using only the panchromatic high-resolution remotely sensed imagery (RSI). Firstly, scale-invariant feature transform feature is detected within both RSI and building template, and the corresponding graphs are constructed. Then, binary matching rules are defined to reconstruct the graphs to reduce the complexity. At last, according to the homogeneity of the building top, disconnected subgraphs are isolated from the reconstructed graphs. To improve the algorithm accuracy, the matched subgraphs are optimized on the basis of the differences in the structure and size. For verifying the validity of the proposed method, nine representatives are chosen from GF-2 images covering Guangzhou, China. Experimental results show that the precision and recall of the proposed method are 97.73% and 87.16%, respectively, and its overall performance F1 is higher than the three other similar methods.

Coastline carrying capacity monitoring and assessment based on GF-1 satellite remote sensing images

The coastline of Zhejiang is the most intensively developed region in China for industry. Based on GF-1 images, an object-oriented classification for coastline was created to monitor coastline exploitation status. A carrying capacity model was then developed to assess the coastline carrying capacity in Zhejiang by means of coastline carrying capacity index (CCI) which is based on coastline artificial parameters and evaluation criterion. The result indicated that coastlines in Cixi, Zhenhai, Fenghua, and Longwan were overloaded with CCI less than 0. The coastlines in Pinghu, Yinzhou, Ninghai, and Dinghai were critically overloaded with CCI ranging from 0 to 0.20. The other 17 sections with CCI bigger than 0.20 were considered load sections. It can be argued that the overloaded sections and critical overload sections in urban regions should be in primary consideration for protection. This study provided a possible model for coastline management with respect to exploitation and ecological function conservation.

A method based on harmonic model for generating synthetic GF-1 images

A method based on harmonic model for generating synthetic GF-1 images

Direct tangible damage assessment for regional snowmelt flood disasters with HJ-1 and HR satellite images: a case study of the Altay region, northern Xinjiang, China

Regional snowmelt flood disasters (RSFDs) can cause significant direct tangible damage which generally refers to the physical destruction due to direct contact with the flood water, such as damage to buildings, croplands, livestock, and infrastructure. Information about people, habitations, and infrastructure affected by the flood is essential for disaster responders and the humanitarian community to plan and coordinate emergency response activities. However, this direct tangible damage information obtained in the ground is limited, incomplete, contradictory, and sometimes impossible to obtain in a short time. Earth observation satellites help overcome operational uncertainties after the RSFDs. Here, we present an improved rapid direct tangible damage assessment model using HJ-1 and GF-1/2 satellite images. We selected the Altay region in northern Xinjiang, China, as the study area, and investigated a RSFD occurring in spring 2017. A series of HJ-1 and GF-1 images were used to track the flood extent over the duration of the disaster, and the maximum affected flood area was assigned as the area in which direct tangible damage occurred. Pre-disaster GF-2 images were then used to estimate direct tangible damage to habitations (2375 households and 6388 rooms), infrastructure (102 km of roads), and affected population (7125) in the flood area, which covered an area of 185,240 m2. Our method is an effective approach for the design of rescue plans and disaster subsidy programs.

Mapping Mangrove Forests Based on Multi-Tidal High-Resolution Satellite Imagery

Mangrove forests, which are essential for stabilizing coastal ecosystems, have been suffering from a dramatic decline over the past several decades. Mapping mangrove forests using satellite imagery is an efficient way to provide key data for mangrove forest conservation. Since mangrove forests are periodically submerged by tides, current methods of mapping mangrove forests, which are normally based on single-date, remote-sensing imagery, often underestimate the spatial distribution of mangrove forests, especially when the images used were recorded during high-tide periods. In this paper, we propose a new method of mapping mangrove forests based on multi-tide, high-resolution satellite imagery. In the proposed method, a submerged mangrove recognition index (SMRI), which is based on the differential spectral signature of mangroves under high and low tides from multi-tide, high-resolution satellite imagery, is designed to identify submerged mangrove forests. The proposed method applies the SMRI values, together with textural features extracted from high-resolution imagery and geographical features of mangrove forests, to an object-based support vector machine (SVM) to map mangrove forests. The proposed method was evaluated via a case study with GF-1 images (high-resolution satellites launched by China) in Yulin City, Guangxi Zhuang Autonomous Region of China. The results show that our proposed method achieves satisfactory performance, with a kappa coefficient of 0.86 and an overall accuracy of 94%, which is better than results obtained from object-based SVMs that use only single-date, remote sensing imagery.

Comparison of CBERS-04, GF-1, and GF-2 Satellite Panchromatic Images for Mapping Quasi-Circular Vegetation Patches in the Yellow River Delta, China.

Vegetation in arid and semi-arid regions frequently exists in patches, which can be effectively mapped by remote sensing. However, not all satellite images are suitable to detect the decametric-scale vegetation patches because of low spatial resolution. This study compared the capability of the first Gaofen Satellite (GF-1), the second Gaofen Satellite (GF-2), and China-Brazil Earth Resource Satellite 4 (CBERS-04) panchromatic images for mapping quasi-circular vegetation patches (QVPs) with K-Means (KM) and object-based example-based feature extraction with support vector machine classification (OEFE) in the Yellow River Delta, China. Both approaches provide relatively high classification accuracy with GF-2. For all five images, the root mean square errors (RMSEs) for area, perimeter, and perimeter/area ratio were smaller using the KM than the OEFE, indicating that the results from the KM are more similar to ground truth. Although the mapped results of the QVPs from finer-spatial resolution images appeared more accurate, accuracy improvement in terms of QVP area, perimeter, and perimeter/area ratio was limited, and most of the QVPs detected only by finer-spatial resolution imagery had a more than 40% difference with the actual QVPs in these three parameters. Compared with the KM approach, the OEFE approach performed better for vegetation patch shape description. Coupling the CBERS-04 with the OEFE approach could suitably map the QVPs (overall accuracy 75.3%). This is important for ecological protection managers concerned about cost-effectiveness between image spatial resolution and mapping the QVPs.

Agricultural Oasis Expansion and Its Impact on Oasis Landscape Patterns in the Southern Margin of Tarim Basin, Northwest China

Oasis landscape change and its pattern dynamics are considered one of the vital research areas on global land use and landscape change in arid regions. An agricultural oasis is the main site of food security and ecosystem services in arid areas. Recently, the dramatic exploitation of agricultural oases has affected oasis stability, inducing some ecological and environmental issues such as water shortage and land degradation. In this study, the Qira oasis on the southern margin of Tarim Basin, Northwest China, was selected as a study area to examine the spatiotemporal changes in an agricultural oasis and the influence on oasis landscape pattern. Based on the integration of Thematic Mapper, Enhanced Thematic Mapper Plus, and GF-1 images, the agricultural Qira oasis has rapidly increased, with annual change rates of −0.3%, 1.6%, 3.7%, and 1.5% during 1970–1990, 1990–2000, 2000–2013, and 2013–2016, respectively. With the agricultural oasis expansion, the agricultural land has increased from 91.10 km2 in 1970 to 105.04 km2 in 2016. The percentage of farmland area has increased by 15.3% in 2016 compared with that in 1970. The natural vegetation is decreasing owing to the reclamation of desert–oasis ecotone. The oasis landscape change and pattern are mainly affected by agricultural expansion under water-saving technological utilization, land use policy, and regional economic development demand. The expansion of agricultural oasis is alarming due to human overexploitation. Thus, the government should adjust the layout of agricultural development and pay considerable attention to the oasis environment sustainability. This study can provide a valuable reference on the impact of climate change and human activities on a landscape.

A Method for Monitoring Iron and Steel Factory Economic Activity Based on Satellites

The Chinese government has promulgated a de-capacity policy for economic growth and environmental sustainability, especially for the iron and steel industry. With these policies, this study aimed to monitor the economic activities and evaluate the production conditions of an iron and steel factory based on satellites via Landsat-8 Thermal Infrared Sensor (TIRS) data and high-resolution images from January 2013 to October 2017, and propel next economic adjustment and environmental protection. Our methods included the construction of a heat island intensity index for an iron and steel factory (ISHII), a heat island radio index for an iron and steel factory (ISHRI) and a dense classifying approach to monitor the spatiotemporal changes of the internal heat field of an iron and steel factory. Additionally, we used GF-2 and Google Earth images to identify the main production area, detect facility changes to a factory that alters its heat field and verify the accuracy of thermal analysis in a specific time span. Finally, these methods were used together to evaluate economic activity. Based on five iron and steel factories in the Beijing-Tianjin-Hebei region, when the ISHII curve is higher than the seasonal changes in a time series, production is normal; otherwise, there is a shut-down or cut-back. In the spatial pattern analyses, the ISHRI is large in normal production and decreases when cut-back or shut-down occurs. The density classifying images and high-resolution images give powerful evidence to the above-mentioned results. Finally, three types of economic activities of normal production, shut-down or cut-back were monitored for these samples. The study provides a new perspective and method for monitoring the economic activity of an iron and steel factory and provides supports for sustainable development in China.

THE STUDY OF RESIDENTIAL AREAS EXTRACTION BASED ON GF-3 TEXTURE IMAGE SEGMENTATION

Abstract. The study chooses the standard stripe and dual polarization SAR images of GF-3 as the basic data. Residential areas extraction processes and methods based upon GF-3 images texture segmentation are compared and analyzed. GF-3 images processes include radiometric calibration, complex data conversion, multi-look processing, images filtering, and then conducting suitability analysis for different images filtering methods, the filtering result show that the filtering method of Kuan is efficient for extracting residential areas, then, we calculated and analyzed the texture feature vectors using the GLCM (the Gary Level Co-occurrence Matrix), texture feature vectors include the moving window size, step size and angle, the result show that：window size is 11*11, step is 1, and angle is 0°, which is effective and optimal for the residential areas extracting. And with the FNEA (Fractal Net Evolution Approach), we segmented the GLCM texture images, and extracted the residential areas by threshold setting. The result of residential areas extraction verified and assessed by confusion matrix. Overall accuracy is 0.897, kappa is 0.881, and then we extracted the residential areas by SVM classification based on GF-3 images, the overall accuracy is less 0.09 than the accuracy of extraction method based on GF-3 Texture Image Segmentation. We reached the conclusion that，residential areas extraction based on GF-3 SAR texture image multi-scale segmentation is simple and highly accurate. although, it is difficult to obtain multi-spectrum remote sensing image in southern China, in cloudy and rainy weather throughout the year, this paper has certain reference significance.

IMAGE QUALITY ASSESSMENT OF HIGH-RESOLUTION SATELLITE IMAGES WITH MTF-BASED FUZZY COMPREHENSIVE EVALUATION METHOD

Abstract. A Modulation Transfer Function (MTF)-based fuzzy comprehensive evaluation method was proposed in this paper for the purpose of evaluating high-resolution satellite image quality. To establish the factor set, two MTF features and seven radiant features were extracted from the knife-edge region of image patch, which included Nyquist, MTF0.5, entropy, peak signal to noise ratio (PSNR), average difference, edge intensity, average gradient, contrast and ground spatial distance (GSD). After analyzing the statistical distribution of above features, a fuzzy evaluation threshold table and fuzzy evaluation membership functions was established. The experiments for comprehensive quality assessment of different natural and artificial objects was done with GF2 image patches. The results showed that the calibration field image has the highest quality scores. The water image has closest image quality to the calibration field, quality of building image is a little poor than water image, but much higher than farmland image. In order to test the influence of different features on quality evaluation, the experiment with different weights were tested on GF2 and SPOT7 images. The results showed that different weights correspond different evaluating effectiveness. In the case of setting up the weights of edge features and GSD, the image quality of GF2 is better than SPOT7. However, when setting MTF and PSNR as main factor, the image quality of SPOT7 is better than GF2.