Archive

  • 2022 Volume 24 Issue 3
    Published: 25 March 2022
      

  • Select all
    |
  • XUE Shouye, XU Haiyan, GAN Zipeng, LIANG Bingyan, CHONG Biying, WANG Li, ZHANG Bo, LI Xiaoming, LI Lisha, MAO Nan, LIU Guimin, WU Xiaodong
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Pan-Arctic area belongs to the high latitudes, and the vegetation in Pan-Arctic mainly belongs to the temperature-limited ecosystem, and thus the vegetation is sensitive to global warming. A vegetation atlas with high accuracy is the scientific basis for the study of response of vegetation to climate change and its driving mechanisms. However, there is a large room for improvement of vegetation type classification in the Pan-Arctic regions. For a better knowledge of the current status, history, and future trends in the vegetation mapping, we comprehensively reviewed the data resources, approaches, and methods of vegetation type classification in the Pan-Arctic region. Overall, the field survey of vegetation began in the 1920s, and there have been some land cover type data at the regional scale. However, there are still many challenges for Pan-Arctic vegetation type mapping, which can be attributed to several reasons such as differences in field survey data, standards for the land cover type classification, and the heterogeneities in tundra ecosystem, and differences in mapping methods and difficulties of optimizing algorithm. In the future work, more attentions should be paid to standard classification rules, and standard dataset preparation and integration, which will benefit the development of new methods for vegetation type mapping in the Pan-Arctic area.

  • ZHANG Zhikun, FAN Junfu, XU Shaobo, CHEN Zheng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Vatti algorithm is one of the widely used vector polygon clipping algorithms. In the process of intersection calculation based on the construction of scanning beams, the data structure of binary tree and recursive calculation method lead to an obvious increase in time consumption when dealing with polygons with plenty of vertices. The calculation efficiency of the algorithm is significantly affected by the boundary vertex number of the overlapping polygons. Aiming at the efficiency improvement of the time-consuming scanning beam construction process of Vatti algorithm, this paper proposes an optimization approach based on polygon boundary vertex pre-sorting method, which is named VCS (Vertex Coordinate Pre-Sorting) method, and realizes a fine-grained level parallel Vatti algorithm on GPU device based on CUDA. The VCS method replaces the original binary tree data structure of Vatti algorithm with a double linked list, and obtains an obvious efficiency improvement on polygon boundary vertex information searching process with a smaller additional storage space. In GPU environment, the bitonic sorting method is used to sort the elements of polygon boundary vertex array in parallel and filter out the valid values, which overcomes the defect of low efficiency caused by the original algorithm using binary tree data structure. Experimental results show that the improved algorithm presents higher efficiency with the same calculation accuracy as the original algorithm. When the number of polygon vertices is 920 000 and the number of threads in each thread block is 32, compared with the serial algorithm which builds scan beams using CPU, the VCS optimization method with GPU-based parallel polygon clipping algorithm obtains a relative acceleration ratio of 39.6 times, and the efficiency of vector polygon superposition analysis algorithm is improved by 4.9 times overall.

  • LI Yifei, CHEN Jing
    Download PDF ( ) HTML ( )   Knowledge map   Save

    With the wide application of Unmanned Aerial Vehicle (UAV) monitoring, inspection, mapping, and other low-altitude technologies, low-altitude and long-distance aerial path planning has become a great challenge for low-altitude aircraft applications. However, the traditional Rapidly Exploring Random Trees (RRT) and its improved algorithm face the problem of slow computational efficiency in a large-range, long-distance, low-elevation three-dimensional space. Therefore, a bidirectional heuristic RRT* algorithm with R-tree spatial index is proposed in this paper. Based on the bidirectional RRT* algorithm, the heuristic function is set for random sampling process to avoid the occurrence of local minimum when facing the gap between small urban obstacles. On this basis, R tree spatial index is established for urban obstacles, which reduces the time of collision detection in the case of massive obstacles and improves the efficiency of low altitude and long distance air path planning. In addition, in order to obtain a path more in line with the motion law of unmanned aerial vehicle and improve the practicability of the algorithm, a turning threshold is set in the sampling process to control the turning angle to prevent it from being too large, and the planning result path is smoothed by using the cubic B-spline function. Finally, in the three-dimensional city scene of Wuhan, the experiment is carried out using the 3D building data of Wuhan. The experiment results show that, compared with RRT algorithm and bidirectional RRT* algorithm, the planning time of the two-way heuristic RRT* algorithm with R-tree spatial index is reduced by more than 90% at different distances, i.e., 500 m, 2000 m, and 10 000 m. Compared to RRT algorithm, the sampling time is reduced by 51.6%, 75%, and 86.7%, respectively, at different distances; and compared to bidirectional RRT* algorithm, the sampling time is reduced by 20%, 24.7%, and 57.3%, respectively, at different distances. Compared to RRT algorithm, the turning time is reduced by 77.3%, 73.5%, and 78.3%, respectively, at different distances; and compared to bidirectional RRT * algorithm, it is reduced by 37.5%, 30.8%, and 16.8%, respectively, at different distances. The result path length of the bidirectional heuristic RRT * algorithm with R-tree spatial index is also shorter than that of the other two algorithms. Our algorithm applied to low-altitude long-distance air path planning can effectively improve the calculation efficiency and reduce the planning time, reduce sampling times, shorten the result path, reduce the swerve times, and enrich the application scenarios of unmanned aerial vehicle.

  • ZHU Qiuzhen, WU Qunyong, YAO Chengxin, SUN Haoyu
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Compared with traditional traffic detection data, floating vehicle trajectory data have the characteristics of wide coverage, low cost, and high mobility, which have been widely used in urban traffic status recognition and have gradually become one of the main data sources for urban traffic state recognition. However, most of the existing traffic state recognition based on floating vehicle trajectory data is based on high-precision or multi-source trajectory data. To address the problem of low accuracy of sparse trajectory data in urban traffic state recognition, this paper proposes a dynamic traffic state classification method combining Davies-Bouldin Index (DBI) and trajectory similarity metric to finely classify urban road traffic state, and then realizes the spatial and temporal characteristics of urban traffic state. Firstly, we pre-process the trajectory data and road network data, including data cleaning, map matching, and format conversion, record the relative spatial distances between trajectory points and matching road sections, and build a set of trajectories of road sections in different time slices. Secondly, we dynamically extend the spatial dimension of trajectories by using Davies-Bouldin Index (DBI) according to the trajectory speed-spatial similarity, and construct the best vehicle queue according to the trajectory similarity measure. After that, the different vehicle queues before and after are processed twice, and the vehicle queues are merged according to the rules and connected to form traffic flow clusters, so as to achieve the purpose of dividing local road sections and laying the foundation for subsequent recognition of local traffic states. Finally, the global trajectory points are clustered based on the fuzzy C-means clustering method to divide traffic states, and the speed bounds of different traffic states are obtained and compared with the previous traffic flow cluster speed The comparison is carried out to realize the local road traffic state identification, and then realize the fine analysis of traffic state. The real cab trajectory data at the intersection sections of Xiamen Xiahe Road, Hubin West Road, and Hubin South Road are used for testing. The results show that the traffic speed map calculated by the trajectory similarity metric method can reflect the changes of traffic speed on the road section more clearly, and the method ensures that the vehicle queuing speed distribution is basically consistent with the original trajectory speed distribution. Compared with the comparison methods, i.e., Kmeans++ and ST-DBSCAN, the root-mean-square error of the proposed method decreases by 18.77% and 21.22% on average, and it performs more stable and robust in different experimental road sections. It can effectively and reliably use sparse trajectory data to identify urban traffic states, and then realize the fine analysis of urban traffic states, which provides auxiliary decisions for the management of urban road traffic problems.

  • DENG Yawen, HOU Peng, JIANG Weiguo, PENG Kaifeng, LI Zhuo, DENG Yue
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The boundary of river source region is a kind of important and fundamental national geographical information. Apart from the available river source region boundary information for major rivers such as the Yangtze river and the Yellow River, many middle and small-sized river basins in China still lack explicit and detailed river source region boundary information at present. Thus, it is necessary to scientifically and reasonably delineate the river source region boundary to support the implementation of water-related ecological protection and compensation policy. In this paper, based on established principles for identifying river source region, we proposed a method for river source region boundary demarcation based on multi-characteristic indexes and hierarchical clustering analysis. The Qinhe River Basin was taken as a case study area. Firstly, we utilized mean change point analysis to determine the optimal quantile threshold (0.15%) of flow accumulation to extract sub-catchments units of Qinhe River Basin. Then, the boundary of its river source region was determined by hierarchical clustering analysis based on the multi-characteristic indexes of sub-catchments. The method was also applied to the Yangtze River and Yellow River basins for verification analysis. The results are as follows: (1) The area of the source area of Qinhe River basin based on multi-characteristic indexes and hierarchical cluster analysis was between that obtained by the slope inflection point and hydrological station method; (2) In the Yangtze river basin and the Yellow River basin, the IoU (Intersection over Union) results of this method reached 85.40% and 79.99%, respectively, which indicates that this method is applicable to the extraction of source regions. To summarize, the automatic extraction method for river source region boundary information we proposed could simply and efficiently identify large, middle, and small river source regions’ boundary, which provides scientific support for the identification of ecological security barriers and the implementation of water resources protection policies in China's river source regions.

  • GUO Chenchen, LIANG Juanzhu
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Owing to the rapid advent of urbanization and the increasing demand for medical services by residents, the pressure on medical services in densely populated areas is surging. The analysis of the accessibility of medical service facilities is of primordial importance. In this study, the medical data was garnered from the Fuzhou Municipal Health Commission, and the crawler technology was used to yield the number of residential households to estimate the population. By use of the Baidu map to obtain the real time road condition information of the peak and non-peak time periods, the access time under the optimal route from the community residential districts to the hospital based on the real-time road condition was calculated, and the time zones of medical services were drawn. The accessibility of general hospitals in the main urban area of Fuzhou was analyzed using the two-step mobile (Ga-2SFCA) search method boosted by the Gaussian distance attenuation function, considering factors such as the travel mode, searching time threshold, and travel peak hours. The results yielded show that: (1) By integrating Baidu Map API into Ga-2SFCA model, multivariate and fine-grained analysis of accessibility was implemented, leading to the accurate measurement of urban medical service supply and demand; (2) The time cost of public transportation at different periods was less affected by traffic congestion, and reaching tertiary hospitals was faster. Under the premise of advocating green transportation, this mode of public transportation was recommended for medical treatment; (3) Under different conditions, the accessibility of medical facilities depended on the space of residential differentiation characteristics significantly, on the whole presenting a "single center" and "diminishing layer coil" distribution. High accessibility of residential areas was mainly distributed in urban core areas, and the lower level of accessibility settlement distribution was in the peripheral urban areas. However, other factors can also influence accessibility, such as the time threshold. The accessibility level of medical services markedly differed with the transportation mode, and the accessibility of medical services was significantly higher along the subway. The choice of off-peak travel time can effectively improve the level of medical service; (4) Due to the layout of urban expressways, the spatial distribution of medical accessibility in driving mode was consistent with that of roads, presenting a "loop level" pattern. However, the spatial distribution of accessibility under the public transport mode was affected by the urban bus microcirculation system, displaying the trait of "axial expansion." The method used in this paper provides a new scientific method for refined measurement and analysis of the accessibility of medical service facilities.

  • LIN Yu, ZHAO Quanhua, LI Yu
    Download PDF ( ) HTML ( )   Knowledge map   Save

    In the practical task of remote sensing image classification, the biggest problem with the use of deep neural network method is the lack of sufficient labeled samples. How to use fewer labeled samples to achieve higher accuracy of remote sensing image classification is a problem that needs to be solved at present. ImageNet is the largest image recognition dataset in the world, the model trained on it has rich underlying features. Fine-tuning the ImageNet pre-training model is the most common transfer learning method, which can make use of the rich underlying features to improve the classification accuracy. However, there is a big difference between ImageNet image features and remote sensing image features, and the improvement of classification effect is limited. In order to solve the above problems, a remote sensing image classification method based on deep transitive transfer learning combined with deep neural network is proposed in this paper. This method constructs an intermediate domain using the open-source remote sensing scene recognition datasets as the data source and uses ImageNet pre-training weight as the source domain and remote sensing images to be classified as the target domain for transfer learning to improve remote sensing image classification accuracy. First, based on ImageNet pre-training VGG16 network, the fully connected layer is replaced by the global average pooling layer in order to speed up the weight update of convolutional layer, and the GAP-VGG16 is constructed. The intermediate domain dataset is used for training the ImageNet pre-training GAP-VGG16 to obtain the weight. Then, based on the SegNet, the T-SegNet is designed by adding the convolutional layer into the SegNet to further extract the obtained weight. Finally, the obtained weight is transferred to T-SegNet, and the remote sensing image classification is achieved by training the target domain dataset. In this paper, the Aerial Image Dataset and UC Merced Land-Use Dataset are selected as the data sources of the intermediate domain dataset, and the ZY-3 Panjin area image is selected as the target domain image, 50% and 25% of the training samples are selected for the experiment. The experimental results show that using 50% and 25% of the training samples, the Kappa coefficient of the classification results using the proposed method in this paper is increased by 0.0459 and 0.0545, respectively compared to SegNet, and is increased by 0.0377 and 0.0346, respectively compared to ImageNet pre-training SegNet. For classes with a smaller number of samples, the classification accuracy of the method in this paper is improved more significantly.

  • SHAO Pan, FAN Hongmei, GAO Ziang
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Remote sensing change detection plays an important role in earth observation. This paper presents a novel Adaptive and Semi-Supervised Fuzzy C-Means (ASFCM) clustering algorithm for remote sensing change detection. The proposed ASFCM method integrates semi-supervised technique, spatial attraction model, and fuzzy factor into fuzzy clustering process, and thus can make full use of the characteristics of difference image. ASFCM consists of three main steps. Firstly, according to the property “the larger the gray values of the pixels in difference image, the more likely the pixels belong to the change region”, the difference image is partitioned into two parts-the nearly certain part and the uncertain part-by analyzing the histogram of difference image using thresholds, which are adaptively and automatically obtained. The pixels in the nearly certain part possess a high probability of belonging to the unchanged or changed class. Secondly, the gray values of pixels, the spatial-contextual information, and the pseudo labels of the nearly certain pixels are jointly exploited in the well-designed ASFCM algorithm to compute the membership functions of difference image. Finally, a change detection map is generated by applying the maximum membership principle to the computed membership functions. On the one hand, ASFCM uses the labeling information of the nearly certain pixels to guide the clustering process through semi-supervised scheme. As a result, the change information is enhanced and more accurate membership is achieved. On the other hand, ASFCM employs the spatial attraction model to improve the traditional fuzzy factor. The improved fuzzy factor is then used to integrate the spatial information adaptively, for reducing the effect of noise and outliers. Owing to the above two aspects, the proposed ASFCM method can obtain more accurate change detection result. Three real remote sensing datasets were used to evaluate the performance of the proposed ASFCM method. Eight state-of-the-art change detection methods were used as the comparative methods. The first dataset consists of two synthetic aperture radar images acquired by Envisat in April 2007 and July 2007. For the first dataset, the proposed ASFCM produced the lowest overall error rate (1.99%) and highest Kappa coefficient (91.88%). The second dataset is made up of two Landsat Thematic Mapper images acquired in July 2000 and July 2006. For the second dataset, ASFCM generated the lowest overall error rate (1.69%) and highest Kappa coefficient (93.79%). The third dataset contains two Landsat-7 ETM+ images acquired in August 2001 and August 2002. For the third dataset, ASFCM also gave the lowest overall error rate (2.71%) and highest Kappa coefficient (86.96%). Experimental results demonstrated the effectiveness of the proposed ASFCM method for change detection.

  • ZHANG Kun, WANG Tao, ZHANG Yan, ZHENG Yinghui, ZHAO Xiang, LI Fangfang
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Feature matching is a key step in the area array swing-scan aerial image processing. Traditional feature matching method has some limitations in the area array swing-scan aerial image matching, e.g., the number of matching points is small, and the distribution is uneven. This paper proposes a method based on adaptive feature matching method in brightness space. First, according to the image Position Oriental System (POS) information, we solve the transformation relationship between the images to be matched for image correction, build an adaptive brightness space on the corrected image, and use ORB operator and BEBLID algorithm to obtain feature points and binary feature descriptors in the brightness space. Then, based on the Hamming distance, we obtain the initial matching points, use the RANSAC algorithm to eliminate gross error. Finally, the matching points are transformed to the original image to obtain the final matching result. This paper selects six groups of swept aerial images with different viewing angles and brightness changes for experiments and compares the algorithm proposed in this paper with SIFT, SURF, ORB, ORB+BEBLID, and ASIFT matching methods. The results show that the algorithm in this paper builds an adaptive brightness space by establishing the transformation relationship between images, so that the number of feature points extracted by the algorithm is increased by 1.5 times, and the number of matching points obtained is more than 3 times that of other algorithms. The distribution of matching points is more uniform, and the matching efficiency is higher than other algorithms. The algorithm verifies the effectiveness of the algorithm proposed in this paper to match the area array swing-scan aerial images with brightness changes and viewing angle differences.

  • LI Bo, FAN Junfu, HAN Liusheng, SUN Guangwei, ZHANG Dafu, ZHANG Panpan
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Aiming at the problem of insufficient quantity and spatial refinement in the extraction of industrial heat source from annual scale thermal anomaly data, a neural network industrial heat source extraction method based on temperature feature template is proposed by using VIIRS active fire data. This study took Beijing-Tianjin-Hebei and its surrounding areas as the study area, Firstly, according to the spatial aggregation characteristics of industrial heat sources, the heat source objects were divided by the OPTICS algorithm. Secondly, according to the thermal radiation characteristics of the heat sources, the temperature characteristic template of industrial heat sources and non-industrial heat sources were constructed. Finally, the BP neural network was used to extract industrial heat source objects using the temperature feature template and heat source statistical characteristics as parameters. The results show that: (1) the extraction precision of industrial heat source of the neural network algorithm of temperature feature template proposed in this paper reached 96.31%. Compared with time filtering and logistic regression methods, the extraction precision of industrial heat sources was improved by 8.45% and 7.53%, respectively; (2) From 2015 to 2020, the number of industrial heat sources in the six provinces and cities in Beijing-Tianjin-Hebei and its surrounding areas decreased by 27.46%. The number of industrial heat source objects and heat anomalies in Hebei Province decreased by 8.06% and 7.44% annually, respectively, which was the largest decrease compared with other provinces and cities. The concentration of industrial heat sources in Shandong and Tianjin increased by 25.72% and 86.64%, respectively, indicating that the industrial transformation and upgrade policies in the two places have achieved remarkable results; (3) Tangshan, Handan, Lvliang, and Changzhi accounted for 31.37% of the total industrial heat sources in the study area, which are the main cities in Beijing-Tianjin-Hebei and its surrounding areas. The degree of industrial heat source accumulation and energy consumption in seven cities such as Linfen and Taiyuan was higher than those in other cities; The degree of industrial heat source accumulation and energy consumption in 11 cities such as Beijing and Zhoukou was lower than those in other cities; (4) From January to May 2020, the number of industrial heat anomalies in Beijing-Tianjin-Hebei and its surrounding areas remained unchanged or increased compared with the same period in 2019 and 2021. The COVID-19 had no significant impact on the industrial heat source in the study area. The number of industrial heat anomalies in Wuhan in January and February 2020 decreased by more than 66.67% compared with that in the same period in 2019 and 2021, the number of industrial heat anomalies from March to May 2020 was lower than that in the same period of 2019. The COVID-19 has had a significant impact on industrial heat sources in Wuhan from January to May 2020. This study reflects the current situation and trend of industrial heat source development in Beijing-Tianjin-Hebei and its surrounding areas, which provides a valuable reference for the formulation and adjustment of relevant policies such as reducing energy consumption and improving secondary industry concentration.

  • CAO Yin, YE Yuntao, ZHAO Hongli, JIANG Yunzhong, DONG Jiaping, Yan Dengming
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Gaofen-6 wide field of view (GF-6 WFV) imagery, with wide coverage, high temporal, spatial, and spectral resolution, has been applied in the fields of remote sensing of agriculture and forestry. However, the application potential of GF-6 imagery in the field of remote sensing of water quality lacks a systematic assessment. In this study, four empirical models of single-band model, band-ratio model, partial least squares model, and support vector machine model were developed to retrieve Chlorophyll-a (Chl-a) in Panjiakou and Daheiting reservoirs. The retrieval was based on measured Chl-a concentration and in situ remote sensing reflectance of 37 samples acquired in September 24 and 25, 2019, as well as a quasi-synchronous GF-6 WFV imagery. The application potential of GF-6 imagery in the field of remote sensing of water quality was evaluated according to the performance of four empirical models for Chl-a retrieval in Panjiakou and Daheiting reservoirs. The determination coefficients and comprehensive errors of four empirical models based on GF-6 WFV reflectance simulated by in situ reflectance were above 0.9 and less than 15% for Chl-a retrieval in Panjiakou and Daheiting reservoirs, respectively. The partial least squares model had the highest accuracy among the four empirical models, with a determination coefficient of 0.96 and a comprehensive error of 13.22%. Finally, the partial least squares model was applied to retrieve the spatial distribution of Chl-a concentration in Panjiakou and Daheiting reservoirs based on the GF-6 WFV imagery acquired on September 26, 2019. The Chl-a retrieval result indicated that Chl-a concentration was less than 10 µg/L in Panjiakou reservoir but more than 10 µg/L in Daheiting reservoir. The trophic states of Panjiakou reservoir and Daheiting reservoir were respectively mesotrophic and eutrophic according to trophic level index calculated by Chl-a concentration. GF-6 WFV imagery, with eight bands in visible and near-infrared, has application potential in remote sensing of Chl-a concentration in inland water. In particular, the newly added yellow band and red-edge band 1 in GF-6 WFV imagery contribute to improving the performance of Chl-a retrieval. The band reflectance of the GF-6 WFV imagery, derived from atmospheric correction, has obvious systematic deviations and correction errors, especially for band 4 (Near Infrared, NIR) and band 6 (Red-edge 2). Atmospheric correction error reduces the performance of the GF-6 WFV imagery in Chl-a retrieval in Panjiakou and Daheiting reservoirs. In order to improve the capability of GF-6 WFV imagery in remote sensing of water quality in inland water, the atmospheric correction accuracy of GF-6 WFV imagery needs to be further improved.

  • LI Qingshuo, KE Changqing, ZHANG Jie, FAN Yubin, SHEN Xiaoyi
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The Greenland Ice Sheet is one of the important factors affecting global climate change. Minor changes in Greenland Ice Sheet can cause significant change in sea level. Thus, it is extremely essential to estimate the Greenland Ice Sheet mass balance quantitatively, which lays foundation for understanding global sea level rise and climate change. In this study, the latest ICESat-2 satellite laser altimetry data (November 2018 to September 2019) and the ICESat data (February 2003 to October 2009) are used to estimate Greenland Ice Sheet mass balance from February 2003 to September 2019. The elevation change of the Greenland Ice Sheet is obtained by crossover analysis. During the process of calculating mass balance, we have corrections for deformation of the Firn Air Content, Glacial Isostatic Adjustment, and Elastic Rebound, and the ice-column density is estimated by ice physical process parameters, such as snow accumulation, glacier surface melting, and glacier surface dynamic change. To further analyze the spatial variation of Greenland Ice Sheet mass balance, we compared the results of mass balance across the glacier hydrological basins. Results show: (1) The main body of the Greenland Ice Sheet from 2003 to 2019 is melting with a mean annual elevation change rate of about -11.27 ± 0.83 cm/yr; (2) For ice sheet below 2000 m, the overall volume change rate is -206.0 km 3/yr, which indicates a relatively large ablation and the maximum ablation rate is -6.0 m/yr. In contrast, the ice sheet above 2000 m shows an accumulation trend. Its volume change rate is 14.2 km3/yr, with the maximum accumulation rate of 1.1 m/yr; (3) The total mass balance of Greenland Ice Sheet grounded-ice from 2003 to 2019 is -195.2 ± 13.1 Gt/yr after correction. And the mass balance change has obvious regional variation. The southeastern and northwestern drainage basins show a large ablation trend, and the northeastern drainage basin is the only accumulation zone; (4) The annual mean temperature in Greenland Ice Sheet is rising at a rate of 0.8 K/yr during the period of 2003-2019, while its mean precipitation shows a downward trend especially in the east and northwest of Greenland Ice Sheet, with a maximum decreasing rate of -0.1 mm/yr. Hence, we can conclude that the accelerated melting pattern of Greenland Ice Sheet is due to the combined effect of the increasing annual mean temperature and decreasing annual mean precipitation in this area.

  • LONG Yican, LEI Rong, DONG Yang, LI Dongzi, ZHAO Chenchen
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Automatic detection of aircrafts plays an important role in airport monitoring and intelligence analysis. Currently, there are various and mature methods for aircraft detection, but the aircraft type detection is still facing many problems, such as the use of unpublicized self-made datasets, the difficulty to reproduce experiment, and the inability to verify generalization ability. Therefore, detecting aircraft type quickly and accurately is still a hotspot in the field of remote sensing image analysis. The traditional detection methods are complicated in processes, poor in robustness and generalization, and cannot detect the specific type of aircrafts. In recent years, deep learning methods have been widely applied in the field of computer vision. Compared with the two-stage algorithm, YOLO, as a one-stage algorithm, rejects the steps of multiple regression, includes only a convolution network, and regards the detection problem as the regression problem of image classification and candidate box parameters. However, multi-layer convolution and pooling may weaken or completely lose aircraft features, making it challenging to extract practical features. Meanwhile, remote sensing images are susceptible to light conditions, seasonal changes, cloud occlusion, noise, and other factors, which makes the detection task harder. In order to solve these problems, this paper firstly used the MTARSI dataset to screen samples and then collected aircraft images from open-source methods such as Google Earth using random rotation, changed brightness, added noise, and other methods to form a new aircraft type detection dataset. Secondly, multi-scale adjustment and training were carried out based on YOLOv5. Finally, an across-dataset was used to identify the aircraft in the optical remote sensing images, which could verify the model’s generalization ability. The experimental results show that the method can accurately and effectively detect the number, location, and type of aircraft in the optical remote sensing images and has strong robustness and generalization ability. The accuracy of type detection reached 82.12% in the across-dataset, which can provide technical support for intelligent aircraft semantic analysis and on-board application research.

  • ZHANG Min, WU Wenting, WANG Xiaoqin, SUN Yu
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Tidal flats are important transitional zones between terrestrial and marine ecosystems and have complicated ecological processes and essential ecosystem services. Tidal flats are highly dynamic under the influences of land-sea interactions and anthropogenic activities. Limited by the accessibility, it is difficult to map the tidal flat using traditional survey. To solve the difficulty in obtaining tidal flat elevation data, a tidal flat elevation inversion model suitable for large-scale with high accuracy is needed. In this study, we proposed an algorithm incorporating tidal submergence and time-series Remote Sensing (RS) data to map the topography of tidal flats. We used Chongming Dongtan as an example and further extended the results to the whole Yangtze Estuary. Firstly, the K-means++ clustering was employed to extract the inundation extent of tidal. Then, the frequency of tidal inundation of each pixel was calculated from the time series RS data. Finally, the tidal flat topography was retrieved based on the regional tidal frequency. All available Sentinel-2 and Landsat-8 images from 2016 to 2020 were used to build the time-series dynamic of tidal flats to map the topography. Verified by the in-situ data, the results showed that the total accuracy and F1-score of the inundation extent extraction of the tidal flats were 97.73% and 0.98, respectively. The average absolute error of elevation inversion was 0.15 m. The accuracy of tidal flat elevation was positively correlated with the number of available images. The total area of tidal flats was 346.93 km2 with an elevation range of 1.00~3.84 m. The tidal flats in the Yangtze Estuary were mainly distributed in Chongming Dongtan, Jiuduansha, Hengsha Dongtan, Nanhui Biantan, and Tuanjiesha. Among them, Nanhui beach had the largest area (107.44 km2), while Chongming east beach had the largest elevation difference (2.84 m). The distribution status of tidal flat was mainly affected by sediment hydrodynamics, vegetation, and human engineering activities. Compared with the existing dataset, our results showed a more robust capacity in the inundation extent extraction of tidal flats. With the increasing number of effective observations and tidal level information from time-series RS images in coastal areas, the extraction accuracy of tidal flat information could be further improved. The proposed algorithm has a great potential in rapid mapping of tidal flat topography and is of great significance for the dynamic monitoring and management of tidal flat resources.