Archive

  • 2021 Volume 23 Issue 9
    Published: 25 September 2021
      

  • Select all
    |
  • FANG Zhixiang
    Download PDF ( ) HTML ( )   Knowledge map   Save

    During the development of COVID-19 virus's global epidemic, the fundamental research and various applications of crowd dynamics-oriented observation theories have attracted much attention from many researchers and people all over the world within some related disciplines, such as public health, clinical medicine, geography, public management, etc. Researchers conducted many interdisciplinary explorations in theories and methods of monitoring epidemic dynamics scientifically, preventing and controlling spatial transmission precisely, predicting accurately, and responding effectively. However, no crowd dynamics-oriented observation theories have been proposed in literature so far. This paper revisits the concept and introduces a theory framework of crowd dynamics-oriented observation, which tries to include the core theories of observation from geospatial big data and to support diverse potential developments. Firstly, this article introduces the research background of crowd dynamics-oriented observation, and then summarizes its three core questions (how to observe its change, how to analyze its change, and how to control its change). From the inter-discipline view of geographic information science, surveying and mapping science, this paper explains the research significance and disciplinary value of crowd dynamics-oriented observation theories. Secondly, this paper introduces a framework of crowd dynamics-oriented observation and its spatiotemporal application, and then elaborates on the bottleneck problems of the key observation theories of crowd dynamics, such as fundamental space-time framework theory, space-time quantification and comprehensive observation theory, spatiotemporal process optimization theory, etc. Thirdly, this paper preliminarily introduces some changes of crowd dynamics-oriented observation theories, for example, refined observation driven by the application needs of digital society governance and public safety/health emergency, personal privacy protection and personalized observations by balancing the public interest and personal privacies, the development of integrated observation theories for human-oriented observation and earth-oriented observation, and the theory of crowd dynamics-oriented observation for high-level management and service. Finally, this article points out the potential directions of crowd dynamics-oriented observation theory and methods, such as, the development of big data-driven crowd perception, multi-space refined crowd dynamics observation, and human-land systematical interaction modeling, so as to realize some differentiated, integrated, and hierarchical crowd dynamics-oriented observations. All potential theories are helpful to the scientific decision-making of public management and public service. The crowd dynamics-oriented observation theory should focus on the fundamental research questions related to studying, analyzing, and servicing human beings, which has become a research frontier in geospatial information science, and could play very important roles in supporting national development strategies, such as "New urbanization", "beautiful China", "artificial intelligence", and "new infrastructure", so as to contribute to a green, efficient, smart, and sustainable regional and urban development.

  • SHI Jinlin, ZHOU Liangchen, LV Guonian, LIN Bingxian
    Download PDF ( ) HTML ( )   Knowledge map   Save

    In order to avoid crowd trampling, it is very important to accurately obtain information on the number of crowds from surveillance images. Early crowd counting studies used a feature engineering approach, in which human-designed feature extraction algorithms were used to obtain features that represented the number of people to be counted. However, the counting accuracy of such methods is not sufficient to meet the practical requirements when facing heavily occluded counting scenes with large changes in scene scale. In recent years, with the development of neural network, breakthroughs have been made in image classifications, object detections, and other fields. Neural network methods have also advanced the accuracy and robustness of dense crowd counting. In view of the difficulty of counting dense crowds, small crowd targets, and large changes in scene scale, this paper proposes a new neural network structure named: VGG-ResNeXt. The features extracted by VGG-16 are used as general-purpose visual description features. ResNet has more hidden layers, more activation functions and has more powerful feature extraction capabilities to extract more features from crowd images. Improved residual structure ResNeXt expands on the base of ResNet to further enhance feature extraction capabilities while maintaining the same computing power requirements and number of parameters. Therefore, in this paper, the first 10 layers of VGG-16 are used as the coarse-grained feature extractor, and the improved residual neural network ResNeXt is used as the fine-grained feature extractor. With the improved residual neural network feature of "multi-channel, co-activation", the single-column crowd counting neural network obtains the advantages of the multicolumn crowd counting network (i.e., extracting more features from dense crowd images with small targets and multiple scales), while avoiding the disadvantages of the multicolumn crowd counting network, such as the difficulty of training and structural redundancy. The experimental results show that our model achieves the highest accuracy in the UCF-CC-50 dataset with a very large number of people per image, the ShangHaiTech PartB dataset with a sparse crowd, and the UCF-QNRF dataset with the largest number of images currently included. Our model outperforms other models in the same period by 7.5%, 18.8%, and 2.4%, respectively, in MAE in the above three datasets, demonstrating the effectiveness of the model in improving counting accuracy in dense crowds. The results of this research can effectively help city management, relieve the pressure on public security, and protect people's lives and property.

  • LI Yu, ZHANG Youshui
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Remote sensing based studies of urban thermal environment usually analyze the relationship among vegetation, impervious surface, and Land Surface Temperature (LST). Although the cooling effects of vegetation and warming effects of impervious surface have been widely recognized, quantitative studies on cooling and warming efficiencies are lacking. In this study, Land Surface Cooling Rate (LSCR) and Land Surface Warming Rate (LSWR) were used to quantify the cooling efficiency of vegetation and the warming efficiency of impervious surface, respectively. Taking the central urban area of Nanjing, Jiangsu Province in 2017 as the research area, Landsat 8 OLI remote sensing data of four dates were selected as the data source. Firstly, Linear Spectral Mixture Analysis (LSMA) was used to obtain Fractional Vegetation Coverage (FVC) and Impervious Surface Percentage (ISP). High-resolution Google earth images were used for precision verification. Then, with LST inversion results, the LSCR and LSWR of each season were calculated, and the influence of different LSTs on the LSCR and LSWR was analyzed. Finally, using a thresholding method, FVC and ISP were divided into four intervals of 0%~25%, 25%~50%, 50%~75% and 75%~100%. The LSCR and LSWR of each interval were calculated. On this basis, the changes of LSCR and LSWR of different intervals were analyzed. The results show that: (1) LST is positively correlated with the overall LSCR and LSWR. The cooling effect of vegetation and the warming effect of impervious layer are the strongest in summer, with LSCR being 5.6% and LSWR being 5.1%. (2) In summer, LSCR in every interval is positively correlated with FVC. When FVC is 75%~100%, LSCR reaches the maximum value of 7.5%. In addition, LSWR in every interval is negatively correlated with ISP in summer. When ISP is 75%~100%, LSWR reaches the minimum value of 2.4%. (3) In the future planning, the cooling effect of vegetation can best inhibit the warming effect of impervious surface when FVC is 0%~25% while ISP is 75%~100%, which is the best areal combination of vegetation and impermeable surface. The LSCR and LSWR analysis methods adopted in this study can select the best FVC and ISP intervals from the perspective of preventing the rise of surface temperature. Based on this, different cities can be compared with each other in the future. Considering the impacts of latitude, topography, climate, tree species, etc. on LSCR and LSWR, we can further explore the influencing factors and changing rules of LSCR and LSWR.

  • WANG Shuoqi, ZHAO Liang
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The overlooking landscape of a mountain city is not only the natural landscape resources given by the nature, but also an important part of urban characteristics. Now, with rapid urbanization, the mountain body is gradually integrated with urban construction, forming the spatial characteristics of a mountain city. However, with the development of urban construction, the relationship between the city and the mountain gradually changes. The mountain landscape could be constantly changed during extensive urban construction, and the overlooking landscape of the mountain city will be disappeared in tall buildings and dense blocks. When people look far away, the green mountains will been gone, replaced with dense and airtight facade of high-rise buildings. Therefore, the overlooking landscape of a mountain city needs to be protected through appropriate methods. Taking Jinan city center as an example, this paper explores the rules of the evolution of architectures constructed based on the controlling zones and develops a spatial evolution model of mountain city by combining Cellular Automata theory and Multi-Agents theory which is called Overlooking Space Cellular Automata model (OSCA). We also use some data processing platforms such as GIS, to analyze and modify the results and discuss the developed models of urban spatial structure in a more intuitive way. Finally, we put forward strategies for future spatial development in mountain cities in order to protect or control the overlooking landscape of mountain cities. The OSCA model theoretically combines cellular automaton theory and multi-agent theory and technically integrates NetLogo simulation platform and ArcGIS 3D display platform, which is very innovative. Our research results show that the OSCA model can realize the dynamic evolution of a mountain city’s overlooking landscape and predict the future growth mode and development and construction mode of mountain city's overlooking space evacuation. It provides important theoretical and application values to guide urban planning and control the development of urban space. Although this study cannot bring all the complex elements into the model, it still provides reasonable evacuation suggestions for overlooking space according to the simulation results. We also put forward some solutions for the follow-up study on urban dynamic model construction and provide reasonable suggestions for the development of urban space and urban structure in future.

  • ZHANG Ziwei, HUANG Qiuhao, LU Yu, LI Manchun, CHEN Zhenjie, LI Feixue
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Good Health and Human Well-being is one of The Sustainable Development Goals proposed by the United Nations, and increasing the life expectancy is a significant step towards this goal. Due to differences in the natural environment and social development of Chinese cities, understanding the factors that affect life expectancy in different regions is the key to formulate urban public health policy. Based on the data of 286 cities in China in 2015, this paper used exploratory regression, ordinary least squares, and geographically weighted regression to screen out the most relevant influencing factors to life expectancy and explore their spatial differences. Then, the two-step cluster analysis was used to make targeted policy recommendations for each type of cities. The results show that: (1) Economic development, educational conditions, and medical facilities had a significant positive impact on life expectancy, while average altitude and environmental pollution had a negative impact; (2) Compared with other regions, economic development in the southeast region had a greater impact on local life expectancy; medical facilities in the northeast and southwest regions had a higher degree of promotion of life expectancy for its residents; education conditions in the northern region had a higher impact on the life expectancy of local residents; average altitude had the greatest impact on the life expectancy of residents in the West region; The life expectancy of residents in the northwest region was more susceptible to the negative impact of environmental pollution than in other regions; (3) Cities were divided into three categories based on spatial differences, and the key factors affecting the life expectancy are economic development and environmental pollution, educational conditions, and medical facilities in order. City managers in each category of cities should pay attention to different factors to increase their life expectancy.

  • WANG Xingfeng, LIU Junsheng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    With the increase of indoor space application and the development of indoor positioning technology, the integration and application of indoor location information has become one of the hot spots in indoor GIS research. Emergency rescue and navigation for indoor space, such as large venues, has become a research hotspot of indoor GIS application. The construction of indoor network is the key technology to realize the indoor emergency navigation service. In this paper, aiming at the problem of indoor navigation routing, we proposed and constructed Indoor Cognitive Hierarchical Coding Method (ICHCM) based on indoor space perception and hierarchical cognition. The main contents are as follows: (1) Based on the law of indoor space perception and the way of hierarchical cognition, the indoor road network was simplified into four levels: street-building level, building-floor level, floor-block level, and block-room level. Thus a tree network of multi-level expression was formed; (2) In order to meet the needs of semantic analysis and path finding, the "virtual room" unit was introduced to divide the indoor closed unit and associated unit into room unit based on the analysis of the indoor unit function. The partition strategy of horizontal and vertical connection space was also provided; (3) Based on the cognitive hierarchical model of interior architecture and the results of indoor unit division, the indoor units were coded successively from high level to low level. This indoor unit encoding method is of great significance to semantic relations, spatial queries, topological relations, and path finding in indoor space. In order to verify the feasibility and effectiveness of the proposed road network construction and coding method, a commercial center was selected as study area, four levels of indoor road network were constructed. The number of nodes and arcs of every level of the network was reduced by layered and partitioned processing while the ICHCM network was effectively simplified and the efficiency of calculation was improved. The time used in path-finding was less than those of traditional network models. The same floor routing time was -55 millisecond while the cross floor routing time was -100 millisecond. The results showed that ICHCM model fits the way of the cognition of science for people. ICHCM can describe the characteristics of the network of different levels, enable the integrated path-finding of indoor space, and meet the demand of the precision and efficiency of path-finding. Results from this study provide important basis for indoor navigation.

  • WANG Liying, WANG Xinning
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The traditional 3D plane extraction algorithm for airborne LIDAR data have defects. For example, designing on discrete LIDAR points leads to difficulties in the design of point-based plane extraction methods. It is easy to generate false detection in the smooth transition region of plane by using only the consistency of geometric features. To overcome the above restrictions, a new 3D plane detection algorithm for airborne LIDAR data was developed based on multi-value voxel connected region construction method. The proposed algorithm is designed based on voxel structure and makes comprehensive use of the geometry and the reflection intensity formation from airborne LIDAR data. It converts the traditional plane feature point clustering into connected region construction based on voxel and reflection intensity statistics under spatial constraints. It gives the multi-valued voxel structure construction scheme of airborne LIDAR point cloud data and the planar extraction scheme on this basis, which contributes to the development of airborne LIDAR point cloud data processing and application based on the theory of multi-valued voxel model. The specific implementation process of the algorithm is showed as follows: ① The airborne LIDAR point cloud data is regularized to a multi-valued voxel structure, where voxel value is the average laser reflection intensity, curvature, and normal vector of the LIDAR point(s) within the voxel. ② In the DSM data of voxel structure, voxels with smaller curvature are selected as seeds, and then the seeds and their 3D connected regions, which are connected with the seeds and have similar reflection intensity and normal, are labelled as the plane. ③ In the non-DSM data of voxel structure, the voxels located in the contour buffer of the connected region with laser reflection intensity satisfying statistical characteristics are labeled as planes. In this paper, airborne LIDAR data provided by ISPRS were used to test the accuracy of the proposed algorithm. The quantitative evaluation results showed that the quality and Kappa coefficient of the proposed method were 92.5% and 89.4%, respectively, which were 9.68% and 11.62% higher than that of the traditional region-growing algorithm using only geometric features.

  • WANG Yujing, KONG Yunfeng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Nearby school enrollment is one of the fundamental principles of the compulsory education in China. In this paper, the spatial optimization methods have been applied to local education planning in order to comply with the educational policy on nearby school enrollment. Four planning scenarios are defined: the nearest school enrollment, the optimal nearby school enrollment constrained by the school quotas, the optimal nearby school enrollment constrained by the school quotas and contiguous service areas, and the optimal nearby school enrollment by adjusting school locations. The Mixed Integer Linear Programming (MILP) models for the four scenarios were formulated. The optimization models were tested on a real instance in a county of Henan Province, China. There are 516 residential areas and 31 junior middle schools in the study region. All the instance models were successfully solved by the IBM ILOG CPLEX Optimizer. The case study shows that: (1) It is not feasible to assign all the students to their nearest schools because some schools will be severely overloaded. (2) The total travel distance of students can be significantly reduced by assigning some students to the schools outside their township boundary. (3) Compared with the nearest school enrollment, the total travel distance of students will increase by 40.75% when the school quotas are considered. (4) The design of contiguous service areas of schools is convenient for managing school enrollment, which has no obvious influence on the total travel distance of students. (5) The school service will be significantly improved by adjusting locations of three schools and expanding quotas of two schools in the study area, which will reduce the total travel distance of students by 31.32%. The case study indicates that the nearby school enrollment of compulsory education could be spatially designed by solving the mathematical models such as the generalized assignment problem, the facility service area problem, and the capacitated facility location problem. Both the generalized assignment problem and the facility service area problem aim to minimize the travel distance of students to existing schools, while the capacitated facility location problem is capable of searching better school locations and thus reducing the travel distance of students.

  • HUANG Longyang, WANG Jing, LI Zehui, ZHAO Xiaodong, LIU Jingjing, FANG Ying
    Download PDF ( ) HTML ( )   Knowledge map   Save

    By virtue of providing multiple landscape functions, multi-functional landscape is considered as an important way to relieve the pressure of ecological environment. Supported by the big data of natural resources, the multi-functional landscape research, based on the grass-roots administrative management unit, can more quickly and accurately reflect the spatial characteristics and regional differences of regional physical geography pattern and social-economic development pattern. It can also effectively combine the landscape function management with administrative management, providing technically support for the planning and zoning of territorial space in the aspects of functional assessment and spatial identification. Taking Yantai City as an example, we extensively collected the big data of natural resources, including land use data, natural resource survey and evaluation data, climate data, and multi-source remote sensing data. The natural resource data were used along with the social economy data and Point Of Interest (POI) data to quantify the spatial patterns of Yantai’s six typical landscape functions (Biodiversity maintenance, Carbon sequestration, Soil retention, Crop production, Residential support, Economic activity support) using InVEST model, CASA model, Universal soil loss equations, kernel density analysis, and other methods. The village-level management unit was selected as the basic spatial unit to identify multi-functional landscape areas through the spatial superposition method as well as hot spot analysis. Meanwhile, the trade-offs and coordination between various landscape functions were explored by Spearman's correlation coefficient analysis. Finally, based on the second-order clustering method, the functional clustering of the landscape was conducted and the planning and zoning of territorial space in Yantai City was carried out. The protection and development strategies of various functional zoning were proposed. Results showed that 35.5% of village-level management units are multi-functional landscape hot spots, most of which locate in the contiguous mountain forest in the middle of Yantai City, namely, the junction of various cities. The other 24.1% of village-level management units are hot spots of two landscape functions, indicating a good landscape functional diversity of Yantai City. Meanwhile, the significant correlation between landscape functions shows a synergistic effect of the natural landscape functions. However, there is a significant spatial conflict between the residential and economic support functions. Based on the clustering results of landscape functions at village-level management units, Yantai City was divided into ecological protection areas, agricultural and rural development areas, urban functional development areas, and urban core areas, whose area proportions are 30%, 55%, 11% and 4%, respectively. There was a strong spatial consistency and coordination between the planning division and the current management boundary, indicating that under the support of big data of natural resources, the planning and zoning of territorial space based on landscape function clustering analysis is quite accurate and practical.

  • HU Zui, MIN Qingwen
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Agricultural areas are preserved well in most countries and are famous for the long history, deep cultural deposit, and various types. In the historical process, they have formed the unique cultural landscapes, land use, and food production systems, which are fit for the local geographical environments. Agricultural areas are rich in traditional cultural customs, such as production, sacrifices, prayers, and religions. They are characterized as the values of history, culture, art, research, education, and social economy. The Food and Agricultural Organization (FAO) of the United Nations launched Globally Important Agricultural Heritage Systems (GIAHS) initiative in 2002. This means that the agricultural heritage systems, as one type of special cultural heritages, are formally confirmed and are attracting widespread attention worldwide. Although the cultural heritage conservation has inevitably been involved with the digital tide, there is not enough attention on agricultural heritage digitalized conservation. To explore this issue, this paper proposes a conceptual framework for the digitalized conservation of agricultural heritage systems and mainly focuses on the pertinent context, objects, and key technologies. In this work, the authors thinks the food and livelihood security, agricultural diversity, local and traditional knowledge systems, culture, values and social organizations, landscapes and seascape features as the body of context of agricultural heritage digitalized conservation. In this paper, the conceptual framework is composed of five parts; namely, the elements of agricultural heritage systems, date types and capture methods or obtained ways, database construction and data organization, sub-systems or application program interfaces, and the main goals as well as the application fields. This paper details every part of the conceptual framework from the bottom up. Based on the above results, this paper further briefs the crucial technologies from the perspective of application development. These key technologies mainly include the analysis and process of multiple granularity spatiotemporal objects, cloud computing and services, big-data analysis, and artificial intelligence. It is especially worthy to note that multiple granularity spatiotemporal objects are the theoretical foundation of the next generation GIS. In work, GIS is treated as the key support to the conceptual framework of agricultural heritage digitalized conservation. Finally, this paper analyzes the corresponding significances when carrying out the agricultural heritage digitalized conservation. Agricultural heritage digitalized conservation can not only help the management departments to improve their work and make decisions but also improve the protection work and sustainable development. Agricultural heritage digitalized conservation is very helpful to promote the understanding of related knowledge and the popularization of sciences and education. In a whole, agricultural heritage digitalized conservation is the inevitable results of the combination of information sciences and traditional agricultural knowledge in depth. The authors strongly believe that current research work is of great significances to forward the implementation of agricultural heritage digitalized conservation and the implementation of the smart agricultural heritages since this paper provides good advices.

  • FENG Tianshi, PANG Zhiguo, JIANG Wei, QIN Xiangdong, FU Jun'e
    Download PDF ( ) HTML ( )   Knowledge map   Save

    With the increasing application requirements of water conservancy industry in the new era, hyperspectral remote sensing has shown great potential in water ecology and water environment due to its unique advantages such as high spectral resolution and real time. It also has great potential in applications in water disaster and water resources. This paper introduces the principle of hyperspectral imaging, summarizes the development of hyperspectral payload at home and abroad, and focuses on the application of hyperspectral remote sensing to solve specific water problems, including monitoring of water bloom and aquatic plants, accurate discrimination of water bloom and water grass, inversion of chlorophyll concentration, quantitative estimation of suspended matter concentration and sediment content, real-time and large-scale flood disaster emergency monitoring, quantitative inversion of land surface hydrological parameters, estimation of evapotranspiration, and so on. In addition, the bottleneck problems in the application of hyperspectral remote sensing in water conservancy industry are summarized and analyzed. With the rapid development of hyperspectral imaging technology and increasing availability of hyperspectral data sources, the application of hyperspectral remote sensing in water conservancy will also enter a new stage. It is necessary to confront the needs of water conservancy industry at the new stage. We should enhance the research on multi-platform hyperspectral water conservancy elements stereoscopic monitoring and integration, construction of the standard spectrum database for typical feature elements in water conservancy, and new methods or new theories for water conservancy hyperspectral remote sensing information intelligent mining.

  • YANG Lianbing, CHEN Chunbo, ZHENG Hongwei, LUO Geping, SHANG Baijun, Olaf Hellwich
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Random Forests Regression (RFR) is often used to inverse Soil Salinity Content (SSC)nowadays. However, the most important impact factors on the model accuracy such as the synchronization optimization of the inversion parameters subset and the model parameters have not been studied carefully in the applications of RFR. In this study, we selected Weiku Oasis and Qitai Oasis as experiment areas. The inversion parameters were constructed based on remote sensing data, including Landsat-5 TM, SRTM, and MOD11A2.006. Firstly, we applied Elastic Net (EN) to select a subset of the inversion parameters, developed Genetic Algorithm (GA) and Bayesian Optimization Algorithm (BOA) to optimize RFR, and established RFR models (EN-GA-RFR, EN-BOA-RFR) for stepwise optimization of inversion parameters subset and model parameters. Then we used GA and BOA to simultaneously optimize the inversion parameters subset and model parameters based on the combination methods of RFR, including GA-RFR and BOA-RFR methods. Furthermore, in each experiment area, we compared the prediction accuracy of EN-GA-RFR, EN-BOA-RFR, GA-RFR, and BOA-RFR. In this way, the spatial distributions of various saline soils in each experiment area were analyzed. The inversion parameters of the two experiment areas were also compared and analyzed. The results show that the order of model prediction accuracy in each study area from high to low is BOA-RFR>GA-RFR>EN-BOA-RFR=EN-GA-RFR. Overall, BOA had a better optimization performance than GA. Finally, the results show that the types of saline soils with the largest area in Ku Oasis and Qitai Oasis are saline soil and moderate saline soil, respectively. The inversion parameters have spatial differentiation in the characterization ability of SSC.

  • ZHOU Xinxin, WU Yanlan, LI Mengya, ZHENG Zhiteng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    With the improvement of the spatial resolution of remote sensing images, the high-precision extraction of vegetation information is of great significance for understanding the changing laws of surface vegetation and evaluating ecological regions. Aiming at the problem that the existing vegetation extraction methods are difficult to extract the yellow vegetation information and it is difficult to realize the vegetation cross-season extraction, this paper proposes a deep learning semantic segmentation network of vegetation extraction method based on the feature separation mechanism using the GaoFen-2 satellite data. The network adds a feature separation mechanism that combines separable convolution and atrous spatial pyramid on the basis of Densenet. The atrous spatial pyramid effectively reduces the loss of information while acquiring spatial features of different scales. This network takes the high-level semantic information of vegetation into account in complex background. The feature information is enhanced while the accuracy of the model is improved. In order to reduce the calculation amount and the parameter amount of the atrous spatial pyramid, a separable convolution layer is used to replace its original convolution layer. In this paper, we constructed a high-precision cross-season vegetation sample database. Using the method proposed in this article, vegetation information is extracted from remote sensing images, which solves the problem that it is difficult to effectively extract the yellow vegetation information. This paper selects overall accuracy, F1 score, and intersection over union as evaluation indicators to compare and analyze the accuracy of different traditional methods and deep learning methods. The experimental results show that the method proposed in this paper is better than traditional vegetation extraction methods and other deep learning methods according to the three evaluation indicators. The F1 score reaches 91.91%, the overall accuracy reaches 92.79%, and the intersection ratio reaches 85.10%. The general verification experiment of the different vegetation types in the remote sensing image of GaoFen-2 has been carried out. The experimental results show that the method in this paper can completely extract the vegetation types of woodland, arable land, and grassland in the image. The generalization of vegetation extraction is verified on the remote sensing images of GaoFen-1, GaoFen-6, and SuperView-1.The results show that the method proposed in this paper has a certain general ability. It can realize the automatic and high-precision extraction of vegetation from high resolution remote sensing images. The results of this paper can provide data reference for urban ecological environment evaluation and vegetation application research.

  • LI Guoqing, BAI Yongqing, YANG Xuan, CHEN Zhengchao, YU Haikun
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Land cover change refers to the areal change and type transformation between vegetation cover and non-vegetation cover caused by climate change and human activities, which is closely related to human survival and development, ecological environment evolution, and material energy cycle. Accurate classification is the basis of land cover change while land cover change is the core of global change research. The rapid development of high-resolution remote sensing technology poses a dual challenge to the speed and accuracy of land surface classification. In recent years, the development of new artificial intelligence technology has realized the automatic segmentation of natural scene images. Intelligent image processing technology has become an important force to promote the improvement of remote sensing information service level in the era of big data. The deep learning method represented by the convolution neural network also has significant advantages in the field of remote sensing image classification. To compare the impacts of deep learning model design on the classification results of high-resolution remote sensing images, this paper takes Gaofen-1 images of Zhengzhou City in Henan Province in 2019 as an example. This paper compares and studies the differences of four diverse deep learning network models, improved based on UNet model, in the application of automatic land cover classification of high-resolution images. Furthermore, this paper discusses the influence mechanism of encoder and decoder architecture settings, such as residual network, multi-scale loss function, skip-layer connection, and attention mechanism module, on classification accuracy. The results show that the MS-EfficientUNet model with multi-scale loss function, skip-layer connection, and attention mechanism module is the best for Zhengzhou City land cover classification, with an overall classification accuracy, based on pixel evaluation, of 0.7981. By introducing multi-scale loss function into the decoder, the classification accuracy of the forest, water, and other categories can be effectively improved. Moreover, by improving the encoder, adding skip-layer connection and attention mechanism, the classification accuracy of grassland, water, and other categories can be further improved. The results show that the powerful generalization ability of deep learning technology can effectively break through the spectral and time-phased differences between images, and realize the feature adaptive enhancement and intelligent decision-making, which has great potential in the field of high-resolution remote sensing image segmentation. However, further improvement of classification accuracy and multi-level and large-scale fine classification method are still the focus of the next step. At the same time, the unification of image sequence and the expansion of training samples are also the key factors to further improve the classification accuracy.

  • BIAN Meng, GUO Shuyi, WANG Wei, OUYANG Yuhui, HUANG Yinqin, FEI Teng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    There has been an increase in the area and quality of vegetation coverage in many cities of China, which leads to a concomitant increase in the risk of allergenic pollen that affects human health. However, there is still limitation in the accuracy and regional applicability of pollen forecasting services, partly because pollen concentration predictors are more focused on meteorological observations rather than phenological observation of plants. For the seasonal trend of allergenic pollen concentration, phenological observations of vegetation may be an important indicator as well as meteorological factors, because the characteristics of vegetation phenology are directly correlated with pollen release. In this study, the time series Leaf Area Index (LAI) for tree and grass covers that reflects vegetation growth processes was derived from remote sensing techniques and represented as one of the predictors of pollen concentration in air. By combining the derived LAI information with the daily meteorological data, Nonlinear Autoregressive Neural Networks with External Input (NARXnet) combined with stepwise regression were employed to predict the pollen concentration in air of the next day in Beijing. The results show that (1) the three-day moving average of daily temperature, the cumulative temperature, LAI, and the first-order derivatives of LAI were key predictors of the next-day pollen concentration for the spring season, while the mean daily temperature, mean wind speed, the minimum daily air temperature, the three-day moving average of daily temperature, the cumulative temperature, and the LAI were key predictors of the next-day pollen concentration for the fall season; (2) in Beijing, the inclusion of remotely sensed phenological information could significantly improve the prediction accuracy of the pollen concentration for both the spring and autumn seasons from NARXnet model. According to the results, we conclude that, in combination with the meteorological factors, vegetation phenology information such as LAI obtained from remote sensing is an effective predictor of the next-day pollen concentration.