Construction of community life circle database based on high-resolution remote sensing technology and multi-source data fusion

To provide supporting tools for the community life circle scienti ﬁ c planning and e ﬃ cient management, the paper takes the urban central area as the research object, and constructs the community life circle database based on multi-source data fusion. High-resolution remote sensing technology and mobile internet complement traditional planning institute data and provide e ﬀ ective ways to obtain multi-source data. When multi-source data describes the same feature, the data content could be repetitive or con ﬂ icting, which causes the data to be inaccurate. Multi-source data fusion coordinates repetitive or con ﬂ icting data contents, and forms a uni ﬁ ed, accurate and useful description on the same feature. The results: (1) Land use, building, supporting facility, and road are important construction elements of community life circle. (2) The planning institute, Baidu place API interface and high-resolution remote sensing technology are the main data acquisition channels. The Baidu POI data is an e ﬀ ective supplement to planning institute facility data. (3) Based on the high-resolution remote sensing images change detection and vector extraction technology, the paper updates planning institute building data. The new information technology has enriched the data sources, the paper contributes to take full advantage of multi-source data and improve the reality, completeness and accuracy of database.


Introduction
In the 1950 s, only 30% of the world's population lived in cities.By 2014, the level of urbanization had reached 54%.The United Nations estimates that by 2050, the global urbanization rate will reach 66% (United Nations, 2014).Rapid urbanization creates new challenges and issues.Mexico City has encountered a serious increase in gas emissions and particulate matter (Calderon-Garciduenas et al., 2015), 493 city areas in the USA had high traffic congestion (Schrank et al., 2011).Urban system is under tremendous pressure from rapid urbanization (Bibri & Krogstie, 2017).In response to problems such as the excessive concentration of resources, unbalanced development, and environmental pollution in the industrialization and urbanization process, the Japanese government proposed the concept of a wide-area life circle in 1965.In 1975, Japan's third National Comprehensive Development Plan proposed the construction of a model settlement circle to promote harmony between man and nature (Lu, 2011).In 2016, Several Opinions of the Central Committee of the Communist Party of China and the State Council on Further Strengthening the Management of Urban Planning and Construction proposed the construction of 15 minutes-community life circle.From 1 December 1 2018, the implemented Urban Residential Area Planning and Design Standards (GB50180-2018) (hereinafter referred to as the Standard) put forward to replace the "residential area, residential group" classification model with the concept of "life circle".It highlights the importance for residents to obtain living services within a suitable walking time.
Community life circle has received growing research attention over recent years.It focused on three aspects: spatial definition, service evaluation and planning strategy.In different administrative systems and spatial scales, the spatial definition of the life circle is diverse (Chai et al., 2015;T.-B. Liu & Chai, 2015).Some researchers re-examined and evaluated the relationship between residents, land use and space organization from the aspects of current construction, residents' daily activities, etc. (D.-S.Sun et al., 2016;Lu et al., 2018;Xiao et al., 2018;Y.-Y. Zhao et al., 2018).Based on evaluation results, researchers proposed planning strategies on housing, employment, service, transportation, leisure, etc. (Cheng, 2018;Li, 2017;Lu et al., 2018).The Standard put forward clear requirements for the spatial definition and evaluation content of the life circle, and the existing researches are different from the new content.
In the field of community intelligent construction, the research focuses on two main areas: the first is the target dimension to provide community residents with more convenient, complete and high-quality public services.The second is the method dimension.It takes use of information technology to support community managers to make scientific and reasonable decisions, improve community management efficiency, and foster new community management and service models.So it is necessary to systematically master multi-source data, accurately understand the current construction of community life circle, and make scientific planning and efficient management.The application of high-resolution remote sensing technology provides detailed spatial scenes for sustainable urban construction and development.Previous studies on high-resolution remote sensing technology application cover the following aspects: 1) environmental monitoring.Remote sensing data were applied to retrieve urban surface temperature (Li et al., 2007), explore the cooling impact scope and intensity of urban parks (Feng & Shi, 2012), analyse the spatiotemporal variations of the urban heat island (Zhang & Cheng, 2019) and the like.2) vector extraction of urban elements such as buildings, roads and green space, etc.The emergence of commercially available high-resolution remote sensing images with multispectral channels has provided more potential for large-scale building extraction with higher precision (Gu et al., 2018;Zhang et al., 2018) divided the existing road extraction methods based on high-resolution satellite remote sensing images into three categories: based on pixel, object-oriented and deep learning according to the realization form.Chen et al. rely solely on remote sensing data to conduct urban green space mapping with physical features (e.g., shrubs, trees) and social functions (e.g., public parks, green buffers).3) urban function and expansion measures.Many researchers took advantage of remote sensing data to carry out urban expansion analysis (Deng et al., 2018;Li & Ye, 1997;S.-L. Sun et al., 2008), urban functional zone fine division (Zhou et al., 2020), land use characteristics investigation (Lin & Wu, 2019) and so on.Prior empirical studies represent that new information technology application has led to the trend of diversified data acquisition channels.When the multi-source data describes the same feature, the data content is duplicated or different, which causes the data to be inaccurate.However, the existing research of database construction focus on the data integration, adopts the latest surveying, mapping, remote sensing, and 3 S technologies to build the basic geographic information data system, and integrates thematic information such as resource, tourism, planning, transportation, population and economic information on this basis (Du & Zhang, 2015;He & Yang, 2016;Xi & Zhang, 2014).It lacks the discussion of the relationship between remote sensing data, mobile internet data and traditional planning data, the advantages of multi-source data are not obvious.
Urban central area is the region with a high density of multiple elements.It is also the core area with the most dense population and construction and the most active change cycle (Guan & Yang, 2019).Taking the urban central area as the research object, the paper selects Hedong District, one of the administrative divisions of Tianjin downtown area, as a typical research area.The study explores the method of constructing a community life circle database based on multi-source data fusion, and provides supporting tools for community life circle scientific planning and efficient management.The research focuses on the relevant content of the Standard.The key problem is how to coordinate multi-source data relationships, utilize multi-source data advantages, coordinate diverse descriptions of the same feature, improve the reality, completeness and accuracy of the database, and transform the advantages of multi-source data into effective productivity of planners and efficient decision-making power of managers.

Research area profile
Tianjin is one of the four major municipalities in China, located at north latitude 38-40 degree, and east longitude 116-118 degree.Hedong District is one of the administrative divisions of Tianjin, with a total size of 42 square kilometres (Figure 1).Hedong District has 12 streets: Changzhou Road, Lushan Road, Chunhua, Tangjiakou, Xiangyanglou, Dongxin, Dawangzhuang, Shanghang Road, Dazhigu, Zhongshanmen, Fumin Road and Erhaoqiao (Figure 2).The average street area has 3.5 square kilometers.Hedong District is an administrative district closest to the airport, Binhai New Area, and Tianjin rail station in the six districts of Tianjin downtown area.Eleven cross-river bridges connect Hedong District and Heping, Hexi District.Jinbin avenue, Weiguo road, Dongzi express and other rapid systems connect the Tianjin downtown area.
In 2016, the urban construction land in Hedong District was 37.62 square kilometres, accounting for 90%.In 2017, the number of permanent residents in Hedong District was 972,800, with a population density of 23,200 people per square kilometre.In the 10 years from 2008 to 2017, Hedong District's resident population increased by 152,000, ranking third in the six districts in the city.The growth rate is 18.52%, ranking second in the six districts in the city.Hedong District has the typical characteristics of high urban construction and population concentration, and the research results could be available to other urban central area.The POI data has the advantages of high coverage of facilities types and wide coverage of data range (Xiao et al., 2018), it could also reflect the information of convenience facilities at the bottom of the building, which is not reflected in the supporting facilities from planning institutions (Table 1).

Research method
The method for constructing a community life circle database based on multi-source data fusion, that is, aiming at the characteristics of multi-source data, such as diverse semantic expressions, storage formats and coordinates, specification processing unifies data attribute field codes and data classification methods.
Multi-source data integration eliminates data format and coordinate differences, multi-source data matching establishes connections and detects duplicates and differences between different data sets.Multi-source data fusion integrates the advantages of data to achieve a unified, accurate, and useful description of the same feature, and improves the reality, completeness and accuracy of data.The change detection and vector extraction based on high-resolution remote sensing images constitute important content of multi-source data matching and fusion.The research designs data classification from the perspective of supply and

R E T R A C T E D
demand, inputs data by category, constructs the community life database, and provides an evaluation, optimization, and decision-making platform for effective planning and efficient management of community life circles.

Data specification processing
Unify Data Attribute Field Codes.When two types of data describe the same element, the attribute field codes should be unified to form a complete and standardized directory (Table 2), which provides the basis for multi-source data matching and fusion.
Unify Data Classification Methods.Land use and building data are classified according to relatively mature standards.Supporting facility is classified according to the five categories in the new Standard (Table 3).Road data is divided into the traffic path and obstruction path according to the impact on residents' walking.The traffic path includes normal walking speed and certain hindering parameter (Table 4).

Multi-source data fusion
The poor interconnection between multi-source data and duplication or conflict of data content leads to the practical problem that the advantages of multi-source data are not obvious.The paper proposes to integrate data advantages based on collaborative multiple descriptions, make up for the lack of content, improve the reality, completeness and accuracy of data, and transform multi-source data into effective productivity of planners and efficient decision-making power of managers.
Multi-source data integration and fusion are not two separate processes.Multi-source data integration is the basis of fusion, and fusion is the further development based on integration (Tang, 2009).Fusion takes advantages of different data to derive new data that is more usable than the original data (Cobb et al., 1998).The specific steps of multi-source data fusion are as follows: data integration, data matching, and data fusion (Chen et al., 2014).
Multi-source data integration.The data acquisition methods and platforms are different.It leads to the existence of multi-semantic, multi-temporal and multi-scale data (Cui & Guo, 2007).The data format is mostly Shp, Dwg, Excel, Jpg, etc. Data coordinates include local coordinates (Tianjin 90 coordinates, etc.), WGS84 coordinates, Baidu coordinates, etc. Multi-source data integration achieves spatially consistent processing through data format conversion, spatial coordinates unification, spatial correction and geographic registration.
Multi-source data matching.Multi-source data matching identifies the same features, establishes connections between them, and detects repetitive or conflicting content in different data sets (Xu et al., 2009).For different spatial forms such as feature points (supporting facility), lines (road), and areas (building and land use), different matching methods are adopted.The article involves multisource data matching based on geometric features and topological features.Common geometric features include distances between geographical elements, shape descriptions, direction trends, etc.Most point entities are matched using spatial distance as a measure.Topological matching is based on the similarity of topological features and is usually used in combination with geometric matching.The topological relationship is the basic spatial relationship among the built-up elements, and has the characteristics of not changing with the geometric changes.

R E T R A C T E D
Multi-source data fusion.Multi-source data integration realizes the unification of data space logic.Multisource data matching establishes a connection between the same name features, and detects duplicates and differences in different data sets which describes the same feature.Multi-source data fusion simplifies repeated data, complements differential data, and integrates the advantages of multi-source data to form a unified, accurate, and useful description of the same feature.Multi-source data fusion establishes the fusion principle.First, legal data are used as the background data.Second, if the data does not involve legal data, the reality and accuracy are given priority.The data with higher reality and accuracy are used as background, and other data sets describing the same feature are fused.

Database structure design
The spatial mismatch between the facilities supply and the actual demand has caused the higher time cost when using public services, and it has brought a lot of social and economic costs (Zheng et al., 2017) The analysis method based on the relationship between supply and demand takes into account multiple factors such as service population, service scope, and facility scale.The quantitative difference between supply and demand can directly reflect the contradiction between construction elements supply and resident demand.Therefore, in order to facilitate the analysis of service supply and service demand in the community life circle, and effectively meet the needs of community life circle planning, management, and service, the paper puts forward that the community life circle database should divide the data into three categories of service demand, service path and service supply, and form the database structure integrating "data categoriesdata classesdata subclassesaccuracy descriptionformat description"(Figure 3).

Result analysis
Multi-source data integration results analysis Baidu POI and planning institute facility data have some duplicate information Based on the integration results of the supporting facility (Figure 5) and statistics on the types and number of facilities, we could find that: (1) The public service facility accounts for the highest proportion of the planning institute facility data.The community service facility accounts for the highest proportion in Baidu POI (Figure 6).The commercial service facility data in Baidu POI, such as catering facility and banking outlets, is effective supplements to the planning institute facility data (Figure 7).(2) The paper further studies the public service facility, takes the primary school facility as an example, and finds that Baidu POI and the planning institute facility have repeat points (Figure 8).
Planning Institute Building Data and High-resolution Satellite Remote Sensing Image Reflect Different Construction Information High-resolution satellite remote sensing image and building data both reflect the construction information of Hedong District, but the building data's reality and completeness are not enough (Figure 9).Taking the Hedong District Experimental Primary School plot as an example, the remote sensing image reflects that the plot has been completed, but the building data reflects that the plot is blank (Figure 10).
Therefore, further research focuses on identifying which supporting facilities are duplicated and which buildings' information are differential based on multisource data matching, simplifying duplicate data and complementing differential data based on multisource data fusion, and then coordinating multiple source data to describe the same feature.

Multi-source data matching
Identifying duplicate data of supporting facilities based on geometric feature matching Taking primary school facility data as an example, the paper uses the distance analysis tool of the ArcGIS spatial analysis module, calculates the Euclidean distance by making the planning institute "primary school" facility as the source data, assuming that the coordinates of two points A and B are (x 1 ; y 1 ), (x 2 ; y 2 ), and making use of the following formula to calculate the Euclidean distance: The Euclidean distance output result is superimposed with the POI primary school facility (Figure 11).The study sets thresholds according to the requirements of the service radius of various facilities (such as the primary school facility threshold is set to 300 m).When the distance is less than a certain threshold, the two data are considered to describe the same feature, that is, they are regarded as duplicate data of the database.12).The buildings and remote sensing images in the same unit describe the same plot (Figure 13, Figure 14).The next step is to determine the unit whose buildings are different from the remote sensing images by change detection technology.Due to the imaging principle of high-resolution remote sensing images, there will be noise information, which will cause false change detection.As shown in Figure 15, compared with the 2015 remote sensing image, 2019 remote sensing image has the ground reflection phenomenon, the unit that has not changed will be identified as a change (different) unit.Therefore, the paper proposes a method for eliminating noise based on multitemporal remote sensing images, that is, downloading remote sensing image data in five phases of 2015/2016/ 2017/2018/2019 on a yearly cycle.The data information of multiple phases confirms each other.As shown in Figure 16, from 2015 to 2018, the building information of the plot has not changed.So there is noise information reflected in the 2019 remote sensing image, and it will be not determined as a change unit in the difference recognition.Based on the five temporal remote sensing image information, the paper detects the 2280 basic units in Hedong District, and obtains a total of 46 change units (Figure 17).Complementary difference data and make building space layout more closer to the current The method of building vector extraction is to eliminate the ground features with obvious differences in shape, texture and spectral characteristics with buildings, construct the building template according to building typical shape, perform convolution calculation on the changing unit image to extract the building area, and calculate the edge detection and refinement to achieve building vector extraction in changed units (Wu et al., 2012;M. Zhao et al., 2013).According to the data matching results (Figure 17), Hedong District involves two changed types: (1) The building layout changes and the land use boundary remain unchanged (Figure 18), with a total of 38 units.Aiming at this change type, we takes the current land use boundary as a constraint, extracts the building patches, fills and replaces the original building patches (Figure 19).(2) The building layout changes, and the land use boundary merges (without crossing the road) (Figure 20), with a total of 8 units.Aiming at this change type, we takes the merged land use boundary as a constraint, extracts the building patches, fills and replaces the original building patches (Figure 21).
The multi-source building data fusion in the urban central area realizes the updating of the planning institute building data (Figure 22).The study operates dynamically in a certain cycle to maintain the reality and accuracy of building data (Figure 23).
Simplify duplicate data and make the facility space layout more complete Based on the raster output results of Euclidean distance calculation of primary school, combined with Baidu coordinate picking system, a total of 13 primary

R E T R A C T E D
school overlapping points are found, and then the study deletes the duplicated primary school facility data in Baidu POI (Figure 24).But the attribute description of POI facility data is more abundant than the planning institute facility, for example, one of the POI primary school is described as Hedong    District Experimental Primary School, but the planning institute facility is just described as primary school, which is relatively vague and does not reflect the quality of the primary school.Therefore, while deleting the duplicate POI data, its attribute information is integrated into the planning institute facility data.The same method is adopted to achieve the geometric and attribute features a fusion of repeated data in various supporting facilities, and optimize the completeness and accuracy of the supporting facility data (Figure 25).

Enter data according to the data category
In order to facilitate the spatial analysis and optimization, the paper inputs data according to categories (Table 5).Service demand refers to the spatial location and quantity of demand, service path refers to the urban road network required by residents to obtain services, and service supply refers to the spatial location and quantity of services.

Database application
Based on the input of three major categories data, the community life circle database supports community life circle planning, construction, implementation and fine governance (Figure 26).The specific data content and evaluation standards for different application requirements are different, but the operation logic is similar.Therefore, the article takes the optimization of the supporting facilities layout in the community life circle as an example to explore the logic and process of the database to support the community life circle intelligent construction.The database includes four functional sections: spatial layout assessment, priority level delineation, community management optimization, and spatial layout optimization.It's work processes are shown in Figure 27.

Main function II: management and decision
The spatial analysis results include three levels: Residents Committee-Space Units (Docking Streets)-Administrative Districts.Each level uses the residential quarters as the basic evaluation unit to calculate the percentage of residential quarters that have completed the supporting facilities (the percentage = the number of residential quarters that meet the standards in the evaluation unit/the total number of residential quarters) (Figure 31, Figure 32), and forms a horizontal visualization of the comparison results at each level.Managers can judge the priority level of the upgrading unit, draw up a plan to improve the supporting facility of Hedong community life circle, and carry out the improvement and upgrading work of supporting facility.

Conclusion and discussion
Based on the relevant content of Urban Residential Area Planning and Design Standards, with the vision of enhancing residents' happiness and livability, the article establishes an interconnected community life circle database, integrates the advantages of multisource data and synergistically describes multiple descriptions on the same feature to improve the reality, completeness and accuracy of the data.Research indicates that:     The study provides methods and processes for the subsequent processing and entry of richer, multi-source data, and provides basic data support for scientific planning, efficient management and effective services of the community life circle.Based on the base of this article, the possibility of database application can be explored in the following areas: research on the optimization of supporting facilities in community life circles, the fairness of park green space, improving the vitality of streets, and dynamic monitoring of life circle construction.Follow-up studies could enrich data acquisition channels, expand database content, extend the application export of community life circle database, and improve the multiple targets oriented community life circle database content to promote the community life circle scientific planning, efficient management and effective service.

Disclosure statement
No potential conflict of interest was reported by the authors.

R E T R A C
T E D Standard contents, the paper regards land use, building, supporting facility and road as the core construction elements of community life circle.The basic data acquisition methods are divided into three aspects.The first is the acquisition of planning institute basic data, including statutory confidential data such as topographic maps and land use.They reflect the spatial distribution information of urban land use, buildings, supporting facilities and roads.Second, with the development of high-resolution satellite remote sensing technology, Institute of Remote Sensing and Digital Earth Chinese Academy of Sciences research team has mature technology products such as basic remote sensing image synthesis and building vector extraction.The third is to obtain the POI information through the Baidu map Place API interface.
. Song et al. regarded the residential population as the demand side, regarded the facility capacity as the supply side, and analyzed the supply-demand relationship between primary schools and residential population (song et al., 2014).Liu et al. used facilities effective supply rate, compliance rate, effective operating rate, and demand rate to quantify facilities supply and resident demand (Y.-T.Liu & He, 2016).

Figure 4 .
Figure 4. Community life circle database construction.

Figure 6 .
Figure 6.The current facility data composition of the planning institute (left) and the Baidu POI (right).

Figure 5 .
Figure 5. Integrate planning institute facility data with Baidu POI data.

Figure 7 .
Figure 7. Number comparison of Partial facility.

Figure 8 .
Figure 8.Primary School facilities have repeat points.

Figure 9 .
Figure 9. Overlaying remote sensing image and building data.

Figure 10 .
Figure 10.Remote sensing image and building data have different descriptions on Hedong District Experimental Primary School plot.

Figure 11 .
Figure 11.Superimposing the raster output result with the POI primary school facilities.

Figure 12 .
Figure 12.Cutting remote sensing images with current land use boundaries.

Figure 15 .
Figure 15.Remote sensing image in 2015 (left) and remote sensing image in 2019 (right).

Figure 17 .
Figure 17.Building data change detection result.

Figure 18 .
Figure 18.Building layout changes and land use boundary is unchanged.

Figure 19 .
Figure 19.Vector extraction of building based on the current land use boundary.

Figure 20 .
Figure 20.Building layout changes, land use boundary mergers.

Figure 21 .
Figure 21.Vector extraction of building based on the merged boundary.

Figure 23 .
Figure 23.Multi-source building data fusion based on multitemporal change detection and high-resolution satellite remote sensing image vector extraction.

Figure 24 .
Figure 24.Deleting primary school facility duplicate data.

Figure 26 .
Figure 26.Community Life Circle Database Application.

Figure 27 .
Figure 27.Database supports the community life circle construction process.

Figure 28 .
Figure 28.Spatial analysis, scientific visualization and iterative optimization function.

Figure 29 .
Figure 29.Layout evaluation of the supporting facilities.
Land use, building, supporting facility, and road are important construction elements of community life circles.The supporting facilities data comes from the two channels of Baidu Map (POI) and the Planning Institute, and the building data comes from the Planning Institute and the remote sensing agency.(2) Based on the multi-source data integration results, it is found that there are differences and duplicate information in the building and supporting facility data.The change detection technology based on high-resolution remote sensing images is a process of multi-source data matching based on topological features.It found that there are 46 land units that have different building data, which are divided into two types: the same land use boundary with different building layouts and different land use boundaries with different building layouts, respectively, 38 and 8 land units.The repeated contents determining the result of the facility data takes junior high school and primary school facilities as examples, and the repetition rates are 35% and 56%, respectively.Multisource data fusion is to simplify the repeated content of facility data and complement the differences in building data to achieve a unified and accurate description of the same facility or the same plot.Based on the perspective of supply and demand, a data classification structure is designed, and the data of each element is entered into the database by category.

Figure 30 .
Figure 30.Comparison of the current situation and optimization plan.

Figure 31 .
Figure 31.Priority upgrade level of supporting facilities at the space unit level.

Table 1 .
Multi-source data content, format, advantages and disadvantages.
• Short data update time • Low labor cost It needs human-computer interaction to make up for some errors Road Baidu Map Place API interface Supporting facility Vector (Excel) • High coverage of facilities types • Wide coverage of data range -224 J. ZUO ET AL.

Table 2 .
Attribute field code directory.

Table 3 .
Unify the classification methods of supporting the facility.

Table 4 .
Unify the classification methods of road.