Urban Ecosystem

Intelligent Identification of Building Patches and Assessment of Roof Greening Suitability in High-density Urban Areas: A Case Study of Chengdu

  • LUO Luhua , 1, 2, 3 ,
  • CHEN Mingjie 4 ,
  • DONG Lulu 5 ,
  • SU Wei , 6, * ,
  • LI Xin 7, 8 ,
  • HU Xiaodong 8 ,
  • ZHANG Xin 9 ,
  • LI Chen 10 ,
  • CHENG Weiming 7 ,
  • SHI Hanning 1, 2, 3 ,
  • LUO Jiancheng 9
  • 1. Faculty of Geomatics, Lanzhou Jiaotong University, Lanzhou 730070, China
  • 2. National-Local Joint Engineering Research Center of Technological and Application for National Geographic State Monitoring, Lanzhou 730070, China
  • 3. Gansu Provincial Engineering Laboratory for National Geographic State Monitoring, Lanzhou 730070, China
  • 4. College of Geoscience and Surveying Engineering, China University of Mining & 00026; Technology (Bejing), Beijing 100083, China
  • 5. School of Earth Science and Engineering, Hebei University of Engineering, Handan, Hebei 056000, China
  • 6. Tianjin University Urban Planning and Design Institute Co. Ltd, Tianjin 300072, China
  • 7. State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
  • 8. The ImageSky International Co. Ltd, Suzhou, Jiangsu 215000, China
  • 9. Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, China
  • 10. School of Architecture, Tianjin University, Tianjin 300072, China
* SU Wei, E-mail:

LUO Luhua, E-mail:

Received date: 2021-02-09

  Accepted date: 2021-10-24

  Online published: 2022-03-09

Supported by

The China Postdoctoral Science Foundation(2019M650830)

The National Key Research and Development Program of China(2016YFC0502903)

The National Key Research and Development Program of China(2017YFB0504201)

The Seed Foundation of Tianjin University(2021XSC-0036)

The Natural Science Foundation of Tianjin(19JCYBJC22400)


With the expansion of a city, the urban green space is occupied and the urban heat island effect is serious. Greening the roof surfaces of urban buildings is an effective way to increase the area of urban green space and improve the urban ecological environment. To provide effective data support for urban green space planning, this paper used high-resolution images to (1) obtain accurate building spots on the map of the study area through deep learning assisted manual correction; and (2) establish an evaluation index system of roof greening including the characteristics of the roof itself, the natural environment and the human society environment. The weight values of attributes not related to the roof itself were calculated by Analytic Hierarchy Process (AHP). The suitable green roof locations were evaluated by spatial join, weighted superposition and other spatial analysis methods. Taking the areas within the Chengdu city’s third ring road as the study area, the results show that an accurate building pattern obtained by deep learning greatly improves the efficiency of the experiment. The roof surfaces unsuitable for greening can be effectively classified by the method of feature extraction, with an accuracy of 86.58%. The roofs suitable for greening account for 48.08%, among which, the high-suitability roofs, medium-suitability roofs and low-suitability roofs represent 45.32%, 38.95% and 15.73%. The high-suitability green buildings are mainly distributed in the first ring district and the western area outside the first ring district in Chengdu. This paper is useful for solving the current problem of the more saturated high-density urban area and allowing the expansion of the urban ecological environment.

Cite this article

LUO Luhua , CHEN Mingjie , DONG Lulu , SU Wei , LI Xin , HU Xiaodong , ZHANG Xin , LI Chen , CHENG Weiming , SHI Hanning , LUO Jiancheng . Intelligent Identification of Building Patches and Assessment of Roof Greening Suitability in High-density Urban Areas: A Case Study of Chengdu[J]. Journal of Resources and Ecology, 2022 , 13(2) : 247 -256 . DOI: 10.5814/j.issn.1674-764x.2022.02.008

1 Introduction

“Urbanization in China 2.0” predicts that China’s urban ization rate will rise to 75% by 2030. With the rapid development of urbanization, the number of new buildings in cities is increasing, which means that the green space in cities and suburbs is decreasing (Hejl et al., 2020) and this highly “compressed” type of urbanization process has seriously affected the environmental quality of people’s living space (Zhang and Chen, 2010), which may also cause a series of problems such as urban heat island (UHI), and air and water pollution (Grimmond, 2007). Urban greening can alleviate the pressure of rapid urban development on the environment (EC, 2015; Kabisch et al., 2017), but it requires scientific methods on green planning for the effective use of existing urban space. However, the planned expansion of urban space is often limited by the existing urban development, the available urban space is often compact, high-density building spaces, which also forces urban planners to combine planning with urban infrastructure such as buildings to begin the planning (Langemeyer et al., 2020). In this context, covering the roofs of buildings with green vegetation (roof greening) is an effective way to provide a reasonable planning solution for the high-density urban space. The green roof strategy has gained more and more impetus, has gradually become the focus of people’s attention, and has already been widely used in many countries (Shafique et al., 2018). The rise of urban roof greening can not only expand urban space and increase the area of urban green space, but also improve the urban microclimate, increase the aesthetic and economic benefits and improve people’s quality of life (Han and Liu, 2014). Many researchers have also demonstrated that green roofs reduce air pollution, reduce storm runoff (Rowe, 2011), reduce noise (Connelly et al., 2013), protect roof materials, and extend the lives of the roofs (Tsang et al., 2011).
Green roofs have produced various benefits to the environment, society and economy. Among them, the land surface temperature (LST) is a key indicator for measuring the intensity of the urban heat island effect. The urban surface is covered by a large number of buildings, hardened surfaces and roads, which causes large amounts of latent heat and high temperatures in the urban area. Greening the roof can effectively reduce the heating effect and provide a clear and pleasant outdoor living environment.
Although roof greening has many advantages, it is difficult to comprehensively evaluate the suitability of roofs for greening on a large scale. From the perspective of the implementation of roof greening, the shapes and materials of the roofs of urban buildings, the heights of buildings and the positions of buildings will all result in the differences of suitability among individual buildings for roof greening (Xu et al., 2020).
Due to the influence of direct sunlight, the solar radiation is strong on the highest floor, which is unfavorable to the growth of vegetation. Roof greening is a complex system, which needs to consider the load, irrigation and drainage system of the roof. Because of the low water pressure and weak irrigation system caused by the height of the roof floor, overly tall buildings are not conducive to the implementation of roof greening. For irregular shapes, the material is not suitable for providing a green roof surface. For example, sloped roofs, glass roofs, colored steel sheet roofs, etc., are not suitable for plant growth and on-going maintenance management. Therefore, the characteristics of the floor height and roof surface should be taken into account before the suitability evaluation.
The suitability assessment of roof greening can specifically and accurately assess the roofs that are suitable for greening according to the existing roof morphological characteristics, reducing blindness in planning and improving the efficiency of the practice.
In the research on roof greening, Anahita et al. (2020) and others used the city’s 2D and 3D data to comprehensively consider the roof floor height, water distance and other factors, but they did not consider the influence of the roof’s shape on its suitability for roof greening. Santos et al. (2016) combined LiDAR data with ultra-high resolution remote sensing data to build 3D models to extract roofs with good greening potential, but this method only identified roof surfaces suitable for greening and did not rank the roof suitability priority. Marinos et al. (2016) used ultra-high-res olution images to assess the potential of green roofs and quantify their impact on the Thessaloniki area in Greece. They considering only flat roofs, and the effect of the height and location of the building on the implementation of the roof greening scheme was neglected. In the research of Grunwald et al. (2017), the authors calculated the slope of the roof, taking into account the four factors of air quality, rainwater retention, biodiversity and urban thermal climate, and measured the roof greening Priority, but their evaluation scale was rough (500 m). In urban planning and development, the need to make full use of urban effective space combined with urban building infrastructure and to carry out rapid, accurate and objective evaluation of roof greening suitability for buildings over a large area is an urgent problem that needs to be solved.
High-resolution remote sensing images can accurately distinguish key features for such an assessment, such as the size and shape of buildings. With the development of deep learning, the basic data from the images can be quickly and accurately obtained (Liu et al., 2019). In this paper, the convolutional neural network model (CNN) combined with the artificial correction method is used to quickly and accurately obtain the precise locations in the study area on the GF-2 image to provide effective data support for subsequent roof suitability assessment. The roof evaluation index system includes seven indexes: roof material, shape, building height, building density (B2), population density (B1), road density (B3) and land surface temperature (A1). The weight values of building density, population density, road density and land surface temperature are obtained by AHP, and the suitability of roofs in the Third Ring Road of Chengdu is evaluated by spatial connection, weighted superposition and other spatial analysis methods for green roofs (Fig. 1). This paper aims to provide effective data support and decision support for urban spatial planners.
Fig. 1 Technical flow chart of the roof greening suitability assessment system used in this study

2 Methods

2.1 Data sources

The study area in this paper is the area within the third ring road of Chengdu, which is also the central city of Chengdu. The urban areas within the third ring road include Chenghua District, Jinniu District, Jinjiang District, Wuhou District, Qingyang District and Longquanyi District (Fig. 2). The area has a high population density and dense building groups. It is one of the representatives of the high-density urban areas and also a demonstration area that urgently needs to increase the green area.
Fig. 2 Overview map of the study area
This article uses multi-source remote sensing images as the basic research data, including 0.8 m resolution GF-2 images (from the Geographical Information Monitoring Cloud Platform) and 30 m resolution Landsat8 (from the United States Geological Survey) satellite image data. Ancillary population and area data were obtained from the Chengdu Statistical Yearbook. Building height data and road network data were obtained by programming crawlers on a map by Josiah Goddard.

2.2 Data processing

The preprocessing of remote sensing images includes radiometric calibration, atmospheric correction, mosaic, fusion and cropping; road network data is superimposed on vector data after deduplication and matching; and building density and road density are calculated by spatial statistical analysis tools. In the overlay analysis, the spatial scale of land surface temperature (LST) is 30 m, which is different from the other raster data because LST is retrieved from the Landsat8 image. Therefore, it is necessary to resample the raster data of population density, building density and road density to a 30 m pixel resolution.
The distribution of population density is shown in Fig. 3. There are obvious spatial differences in the distribution of population density in the third ring area of Chengdu. The population density in the southwestern area is obviously higher than that in the northeastern area. In order to obtain a more detailed explanation of the population density in the study area, the population density values of different Districts in the study area are calculated, as shown in Table 1.
Fig. 3 Population density distribution map
Table 1 Population density parameter table
District Population density (person km-2)
Chenghua District 7186
Jinniu District 7090
Jinjiang District 9788
Longquanyi District 1284
Qingyang District 10713
Wuhou District 10594
Road density and building density are shown in Fig. 4 and Fig. 5. The road network is concentrated in Xiyuhe Street, Yanshikou Street and Duyuan Street in the First Ring Road, Guixi Street and South Railway Station Street. The distribution of buildings is highly concentrated within and around a ring road.
Fig. 4 Road density distribution map
Fig. 5 Building density distribution map
The surface temperature affects the outdoor environment where people live. The surface temperature is affected by the surface material (such as lawn, bare land, cement ground, or asphalt ground). At present, the method for obtaining the surface temperature over a large range mainly relies on the method of remote sensing inversion. Because Landsat8 has rich spectral characteristics, it is widely used in experiments on surface temperature inversion. In this paper, the surface temperature at noon on August 11, 2019, obtained by the inversion of the segmentation window algorithm, is shown in Fig. 6.
Fig. 6 Land surface temperature distribution (taking August 11, 2019 as an example)

2.3 Building vector extraction

Based on the GF-2 satellite image data, this paper uses the method of deep learning to assist with the manual correction to obtain building map spots, and the specific network structure adopted is D-LinkNet (Zhou et al., 2018). The process of extraction is divided into five steps: 1) Sample selection and sketch: Select 72 building sample points in the research area, cut and draw a sample size of 1000×1000 pixels, then sketch the surface sample. 2) Building model training: The building model is trained by drawing the building sample. During the training of the building model, the iteration number of the model was adjusted repeatedly, and the model was found to reach the best precision when the iteration number reached 5000 times, the training time of the model is long and the precision has no obvious change. 3) Model prediction: The trained model is used for prediction. If the model effect does not reach the required precision, samples will need to be added and the model will need to be re-trained based on the previous model; if the predicted model is better, the next step can be carried out. 4) Vectorization: Convert the binary image generated by the model prediction into a vector. The result of model prediction is used to generate black and white binary raster data, which needs to be converted into vector data in batches through vectorization. 5) Manual correction: Correct the architectural patterns that are omitted or incorrectly rendered.

2.4 Suitability assessment for roof greening

2.4.1 Feature selection of roof attributes

Referring to the “Chengdu Technical Guidelines for roof greening and vertical greening”, this paper spatially correlates the floor height data with the building vector and selects the roof surface that accords with the greening height (height: 15 stories, height less than or equal to 45 m), and the vector of roof building with the attribute information of floor height is obtained (Dong et al., 2018).
Due to the different shapes and materials of the roofs themselves, not all roof surfaces are suitable for greening. Therefore, it is necessary to obtain data on the roof surfaces with regular shapes and suitable materials through classification. The object-oriented classification method is a method of image classification based on establishing the corresponding discriminant rules according to the attribute characteristics of the image objects. It also makes full use of the spatial information, including shape, texture, size, area and other factors. This classification method is suitable for high-resolution satellite images with texture features.
Therefore, this article uses the object-oriented classification method, taking the building vector as the mask data, according to the texture, spectrum and shape characteristics of the building roof surface. Firstly, 342 building vectors that are not suitable for greening are selected as training samples, and the partition scale is set between 90-100. The features of the roofs not suitable for greening are shown in Table 2, and the kappa coefficient is used to measure the accuracy of the classification results.
Table 2 Features of roofs not suitable for greening
Name Type
Irregular shape
Material mismatch
Sloping roof

2.4.2 Weight calculation and comprehensive evaluation

The weight values for population density, road density, building density and land surface temperature are calculated by constructing an AHP model. Since the main benefit of green roofs is to expand urban green space and improve the ecological environment of high-density urban areas, when using AHP analytic hierarchy process, the importance of the four indicators is sorted as B1>A1>B2>B3. Figure 7 is a hierarchy diagram, where T1, T2, and T3 are alternatives for decision-making.
Fig. 7 AHP hierarchy diagram
For evaluating suitability, the evaluation index system of the suitability of roof greening (Table 3) is first constructed, then the importance of the two indexes is judged by the method of five grades and nine grades. The judgement matrix A = (aij) is formed by the comparison of the same-grade indexes, the consistency of the judgment matrix is checked, and the index weight value (Fu and Zhang, 2011) is obtained by solving the eigenvalues of matrix A.
Table 3 Index system of roof greening
Sub-objective Indicators
Roof attribute characteristics Floor height
Roof shape
Roof material
Natural environment
Land surface temperature (A1)
Characteristics of the humanistic social environment Population density (B1)
Building density (B2)
Road density (B3)
The consistency of the matrix is then checked. In general, if the consistency ratio (CR) value is less than 0.1, then the judgment matrix satisfies the consistency test. For calculating the CR, the consistency index (CI) calculation formula is shown as equation (1), the random consistency index (RI) value is obtained from the table of random consistency and ${\lambda _{{\rm{max}}}}$ is the maximum eigenvalue.
$CI=\frac{{{\lambda _{{\rm{max}}}} - n}}{{n - 1}}$
$CR = \frac{{CI}}{{RI}}$
Finally, through the comprehensive weighted superposition, the suitability grade of the green building roof surface is evaluated. In constructing the model of weighted superposition, the input factors are population density, building density, road density and land surface temperature. The input factors were set to the corresponding reclassification values. For example, in the raster data of land surface temperature, the higher the temperature, the higher the suitability value. The weighted values of the four factors obtained by AHP hierarchy analysis were taken as the weighted values of the factors in the weighted superposition model. Finally, the priorities of suitable green roof surfaces were evaluated, and the suitable roofs were divided into three categories, namely, high fitness, medium fitness and low fitness.

3 Results and analysis

Through deep learning, building vectors in the study area can be quickly and accurately obtained. In this paper, a total of 40817 building map spots in Chengdu’s Third Ring District were extracted, and buildings with a floor height of less than or equal to 15 m were screened out through the floor height of the spatial connection buildings. The further object-oriented classification was done, and the accuracy of the classification results were evaluated accordingly. A total of 19623 architectural map spots were screened out in line with the greening height and regular shape criteria. The classification result Kappa was 0.6649, and the overall classification intensive reading was 86.58%. The weight values of building density, population density, road density and land surface temperature were obtained by constructing the AHP analytic hierarchy process, as shown in Table 6.
Table 4 The average random consistency index RI
Order 3 4 5 6 7
RI 0.52 0.89 1.12 1.26 1.36
Table 5 Summary of results of the consistency testing
Variable ${\lambda _{{\rm{max}}}}$ CI RI CR
Value 4.061 0.020 0.890 0.023
Table 6 Table of indicator weights
Indicator Indicator weight
A1 0.3207
B1 0.3763
B2 0.1881
B3 0.1149
According to the method described in the third section, the greening potential of urban roofs was evaluated, and the results showed that the numbers of roofs in the third ring area of Chengdu for buildings with high-suitability (level 1), moderate-suitability (level 2), and low-suitability (level 3) for greening are 7683, 6603 and 5337, respectively. The proportions are 45.32%, 38.95%, and 15.73%, respectively. Among them, Wuhou District has the largest number of suitable green roofs within the Third Ring Road of Chengdu, with a total of 5122 (26.10%). The numbers of suitable green roofs in Jinniu District, Chenghua District, Qingyang District, Jinjiang District and Longquanyi District are 4277, 4093, 3385, 2676 and 70, respectively; accounting for 21.80%, 20.86%, 17.25%, 13.64%, and 0.36%, respectively.
The distribution of roofs suitable for greening in different areas and levels is heterogeneous (Fig. 9). Among them, the first level green roofs are mainly distributed in Jinyang Street, Shuangnan Street, Fangcao Street and Yulin Street in Wuhou District; Caotang Road Street, Guanghua Street and Wangjiagai Street in Qingyang District; Shuyuan Street, Lianxin Street, Chenglong Street and Taoxi Street in Jinjiang District; Xian Street, Fuqin Street and Sumaqiao Street in Jinniu District; and Taoxi Street, Xinhong Street and Shuangshuangqiao Street in Chenghua District. The secondary green roofs are mainly distributed in Hongpailou Street and Shuangnan Street in Wuhou District; Funan Street, Guanghua Street, Dongpo Street and Shaocheng Street in Qingyang District; Huangzhong Street, Chadianzi Street, Fuqin Street, Jiulidi Street and Xian Street in Jinniu District; and Shuangshuangqiao Street, Mengchouwan Street and Xinhong Street in Chenghua District. The distribution of the third-level roofs is more uniform, and they are distributed in each street.
Fig. 8 Extraction results and local maps of architectural map spots
Fig. 9 Distribution map of buildings suitable for greening
Fig. 10 Statistical distribution of suitability for greening of the buildings by district

4 Discussion

(1) Data extraction: Using a deep learning method to extract roof surfaces can effectively reduce the workload of manual extraction of roof surfaces that was necessary in the past. However, due to the limitations of data sources, there are some problems in the extraction process, such as incomplete extraction, wrong extraction and so on. Therefore, this study only serves as a reference for macro data analysis. In future research, we will try to obtain higher resolution images to improve the experimental classification accuracy.
(2) Suitability evaluation: In the construction of the roof greening evaluation index system, only seven indexes are considered, but the influences of other factors, such as building information, green space around buildings, regional economy, government policies and other factors in the experimental area, are not comprehensively considered. In the follow-up study, the construction of the roof greening index system needs to be further improved.

5 Conclusions

With the rapid urbanization process, a series of problems have appeared in the urban ecological environment. Especially in high-density urban areas, the serious damage to the ecological environment and the expanding urban areas lead to a situation in which the construction land tends to be saturated and the ecological environment is difficult to expand. Based on the concept of urban regeneration, the urban infrastructure and the available space resources are fully exploited to improve the urban ecological environment. “Green space” is the goal of ecological space planning.
Based on the high score satellite data, a building extraction model is established, and all the architectural spots in the third ring District of Chengdu are extracted. Compared with the traditional manual interpretation, obtaining the building spots through in-depth learning greatly improves the efficiency of the experiment. Based on the accurately extracted spots, this paper establishes a potential index system of roof greening in high density urban areas from the two levels of roof properties and environmental improvement demand. Through the AHP and weighted superposition, the evaluation of the potential implementation of greening on the roof surfaces in the research area is carried out. In the planning practice of the third ring District of Chengdu, the statistics show that there were 40817 roof surfaces of buildings in 2019, and 19623 of the building spots are suitable for greening. Among them, the first, second and third grade green roofs accounted for 45.32%, 38.95% and 15.73%, respectively, indicating that the first-grade green roofs accounted for the highest proportion.
This study uses deep learning to obtain accurate map spots in the study area, which can provide accurate data support for subsequent spatial planning research. The evaluation results for the suitability of roof greening provide a reference and basis for spatial planning for both the subsequent development of the ecological space network optimization of high-density urban areas, and also for improving urban green space and urban environmental quality.

This work was partially funded by LZJTU EP 201806 and the Jiangsu Shuangchuang Project.

Anahita A, Hossein A, Hafez F. 2020. Simulation of green roofs and their potential mitigating effects on the urban heat island using an artificial neural network: A case study in Austin, Texas. Advances in Space Research, 66(8): 1846-1862.


Connelly Maureen, Hodgson Murray. 2013. Experimental investigation of the sound transmission of vegetated roofs. Applied Acoustics, 74(10): 1136-1143.


Dong J, Zuo J, Li C, et al. 2018. Research on ecological spatial planning method in high-density area under the urban regeneration vision-A case study of a three-dimensional greening plan on Xiamen Island. Acta Ecologica Sinica, 38(12): 4412-4423. (in Chinese)

EC (European Commission). 2015. Towards an EU research and innovation policy agenda for nature-based solutions & re-naturing cities. Final report of the Horizon 2020. DOI: 10.2777/479582.


Fu Y, Zhang Y R. 2011. Performance evaluation and reveations of the poverty alleviation in Lanzhou City. Journal of Hunan Agricultural University (Social Sciences), 12(5): 25-30. (in Chinese)

Grimmond S. 2007. Urbanization and global environmental change: Local effects of urban warming. Geographical Journal, 173(1): 83-88.


Grunwald L, Heusinger J, Weber S. 2017. A GIS-based mapping methodology of urban green roof ecosystem services applied to a Central European city. Urban Forestry & Urban Greening, 22: 54-63.

Han L F, Liu Z Y. 2014. Research on urban green roof planning and design-Illustrated by the case of Beijing. Chinese Landscape Architecture, 31(11): 22-26. (in Chinese)

Hejl M, Martin M, Lukáš B. 2020. Modular green roofs and their usage in the world. International Journal of Engineering Research in Africa, 47: 103-108.


Kabisch N, Korn H, Stadler J, et al. 2017. Nature-based solutions to climate change adaptation in urban areas-Linkages between science, policy and practice. Theory and Practice of Urban Sustainability Transitions. Cham: Springer International Publishing: 1-11.

Liu H, Luo J C, Huang B, et al. 2019. Deep encoding network for building extraction from high-resolution remote sensing imagery. Remote Sensing, 11(20): 2380. DOI: 10.3390/rs11202380.


Langemeyer J, Wedgwood D, McPhearson T, et al. 2020. Creating nature-based solutions where they are needed-A spatial ecosystem service-based decision analysis of green roofs in Barcelona, Science of the Total Environment, 707: 135487. DOI: 10.1016/j.scitotenv.2019.135487.


Marinos K, Ifigeneia T, Giorgos M, et al. 2016. Towards a green sustainable strategy for Mediterranean cities: Assessing the benefits of large-scale green roofs implementation in Thessaloniki, Northern Greece, using environmental modelling, GIS and very high spatial resolution remote sensing data. Renewable and Sustainable Energy Reviews, 58: 510-525.


Rowe D Bradley. 2011. Green roofs as a means of pollution abatement. Environmental Pollution, 159(8-9): 2100-2110.


Santos T, Tenedório J, Gonçalves J. 2016. Quantifying the city’s green area potential gain using remote sensing data. Sustainability, 8(12): 1247-1253.


Shafique M, Kim R, Rafig M. 2018. Green roof benefits, opportunities and challenges-A review. Renewable and Sustainable Energy Reviews. 90: 757-773.


Tsang S W, Jim C Y. 2011. Game-theory approach for resident coalitions to allocate green-roof benefits. Environment and Planning A: Economy and Space, 43(2): 363-377.


Xu N, Luo J C, Zuo J, et al. 2020. Accurate suitability evaluation of large-scale roof greening based on RS and GIS methods. Sustainability, 12(11): 4375. DOI: 10.3390/su12114375.


Zhang J X, Chen H. 2010. China’s compressed urbanization and urban planning responses. Urban Planning Forum, (6): 10-11.

Zhou L C, Zhang C, Wu M. 2018. D-LinkNet: LinkNet with pretrained encoder and dilated convolution for high resolution satellite imagery road extraction. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, Salt Lake City, USA: 182-186. DOI: 10.1109/CVPRW.2018.00034.