A Study on the Cognition of Urban Spatial Image at Community Scale: A Case Study of Jinghu Community in Zhengzhou City

Abstract

The community is the basic spatial unit for urban residents to live and rest. It is a crucial direction of city image research to explore people’s cognitive characteristics of community space image. Aiming at the lack of cognitive quantification of community spatial images, a new method that can quantify community spatial data into cognitive results is proposed. By employing spatial analysis tools, eleven spatial indicators from the perspective of community spatial form and spatial services are selected, and an image structure is constructed based on the characteristics of the indicator results. The results of multiple indicators are organized through the improved technique for order preference by similarity to ideal solution (TOPSIS) and overlay analysis method to produce a spatial image map of the community. The study displays that the spatial image characteristics of the community scale can be comprehensively expressed through three types of elements: district, path (edge), and node (landmark). These three types of elements constitute the image structure at the community scale and present apparent elements’ characteristics. This scrutiny is also aimed to demonstrate the construction and use process of the methodology and to provide new ideas for the cognitive research of urban spatial image at the community scale.

1. Introduction

1.1. Research Background and Progress

As the space where people live, urban space is not only a place that includes clothing, food, housing, transportation, and other activities of people but also a place that is closely related to people’s behavior. We can not only think of the city as a collection of existing things, but we should understand it as a city that the citizens living in it feel. These feelings represent a set of cognitive results of urban space, which are comprehensively expressed as a city image [1,2]. However, due to the large size of modern cities, urban observers tend to follow the law of neighborhood organization, putting together some parts with short distances or close to each other to form a whole [3,4]. Citizens living in urban spaces are accustomed to the community they live in as an organically connected whole and shape a corresponding community image [5]. Therefore, the cognition of residents at the community scale deserves more attention, and the cognition results achieved in the community space can better reflect the spatial characteristics of various parts of the city [6,7].

As one of the methods to examine the urban spatial structure and cognitive results, city image has quickly become the mainstream of city research after being proposed by Lynch [8], and geographers have also begun relevant empirical research under its influence [9,10,11,12,13,14]. City image obtains the cognitive results of individual or group residents in the city on urban space through questionnaire interviews and cognitive maps and summarizes the cognitive results into five categories of elements: path, edge, district, node, and landmark. These five elements randomly appear and are regularly connected, and they will also be transformed into each other under various circumstances, and finally integrated to form an image map in people’s minds [8].

With the continuous deepening of research on the city image, while the theories such as the content and formation process of city image have been improved, the scale and method of research have presented apparent changes. First, the scale of city image research has gradually expanded from macro to micro. Researchers continue to explore microscopic areas such as villages, gardens, campuses, and squares, providing ideas for small-scale imaging research. The research is mostly performed on the aspects of image space structure and its physical constituent elements in the microscopic area [15], the spatial cognition process and influential factors [16], and the evaluation of the image perception situation [17,18]. Second, the research method of city image is gradually shifting from qualitative analysis to quantitative analysis [19,20]. The related research methods mostly include space syntax [21,22], city image identification [23,24], deep learning algorithm [25,26,27], sentiment analysis [28,29], and employing quantitative methods to introduce big data, such as street view photos [30,31] and various social media data [32,33,34], into the study of image and explore the structure and characteristics of the urban spatial image.

Liu et al. took the five elements of the city image as the entry point and took the city image with geographic tags as the breakthrough point. The deep learning algorithm of “scene recognition” was exploited to obtain high-dimensional spatial information in the image, and the image elements of each city were then statistically analyzed. By taking type characteristics and spatial distribution characteristics, the correlation and difference of urban image types in various cities were explored, and the public’s image perception of the city was detected in a quite efficient way [23]. Wilkowski et al. implemented computer vision and machine learning methods, attempting to connect urban form with human perception. To this end, a trained image classification model was employed to identify Lynch’s five elements of city image, and Lynch’s original Boston photos and charts were adopted to construct labeled training data and the same location. The new images were also utilized to quantify the human perception of urban form through the new city image [35]. Based on the online photos of 24 cities in China, Cao et al. constructed a research model centered on the composition of city image elements, dominant directions, characteristics, and image similarity between cities, and the city image in a quantitative way was comprehensively recognized [33]. By comparing 784 pictures posted on Instagram and 10,590 pictures posted by tourists, Agustí examined the results of urban spatial perception based on Barcelona and analyzed the regional size and process characteristics of residents’ spatial cognition based on the discrepancies between social pictures [31].

By reviewing the progress of city image research, two problems were identified by our research group: (1) At present, the research scale of city image has begun to focus on the micro-scale urban space, but most studies still focus on public service areas such as scenic spots, gardens, and shopping malls. There are still few studies on spatial imagery. (2) Although many quantitative analysis methods have been used in the current research on urban intentions, it is worth considering that these methods are mostly applicable to the image research of the entire urban space at the urban scale. On the contrary, in the study of spatial images at the community scale, the experimental data should be more refined, and there will be problems with high complications in obtaining data or large gaps that make it difficult to implement the quantitative approaches that are commonly used in imagery research. Let us take Weibo data commonly exploited in city image research as an example. These data can cover an extensive range of urban spaces, but in the case of focusing on a small space such as a community it can be concluded that the available data are often too small for effective analysis.

1.2. The Main Goal of the Paper

As one of the most important living spaces in the city, the results of people’s perception of its space provide crucial support for the exploration of the urban space [36]. Therefore, the research method used for the city image cognition at the community scale has also become an attractive research topic.

In order to unlock the problem that the quantification methods, which are commonly implemented in image research, are difficult to apply in community space; hence, the spatial analysis method is chosen as the research tool in this scrutiny [37,38,39]. On the one hand, it is not difficult to obtain spatial data in the case of employing spatial analysis methodologies. It can be obtained through image vectorization or directly downloaded from open-source databases, and there is usually no problem of missing data (except for some specific areas that do not disclose image data). On the other hand, spatial analysis methodologies are applied at various scales. Compared with quantitative methods commonly used in image research, spatial analysis methodologies demonstrate strong applicability and advantages in urban micro-space.

The main research goal of this paper is to propose a newly suitable method for exploring spatial image cognition at the community scale and to supplement and describe the specific operation of the method in a case study. The research part of this paper reveals the idea of constructing the methodology and the process of using the methodology to obtain the spatial image results of the community. By examining the geographic data of the community space, the research displays the information contained in the space through an image map. Such an image map can intuitively display the results of residents’ perception of community space, helping city managers to understand residents’ feelings about the exploitation of community space. During managing and updating community space, different solutions are proposed for areas with different levels of cognition, assisting community space to become a more preferred living space for residents.

2. Materials and Methods

2.1. Description of Understudied Area

Jinghu community is placed under the jurisdiction of Renhe Road Street, Erqi District, Zhengzhou City, Henan Province, China. This is the only provincial smart community pilot selected in Zhengzhou (see Figure 1). China is actively running the construction of smart communities, fully applying modern information technology in the community, integrating various service resources in the community, and improving the level of community governance and services. By establishing a smart community pilot in the city, it is possible to summarize the successful experience of building a community and then promote the level of community building in a scientific and systematic manner.

It has seven courtyards, 45 buildings, and about 13,000 residents. The constituent courtyards are adjacent to the community, with strong spatial connectivity. Each courtyard is different in shape and spatial layout, with clear boundaries and access control equipment. It is a typical gated community [40,41]. The architectural form in the Jinghu community presents prominent characteristics of homogeneity, essentially consisting of high-rise residential buildings and low-rise shops. High-rise residential buildings maintain high consistency in shape, color, height, and orientation. Although there are differences in the orientation of low-rise commercial buildings, they all face the courtyard exterior. The community selected in the present investigation has a typical new urban community form and model, which can be employed as a community representative in the urban space [42,43,44,45].

2.2. Source of Data

The research needs to analyze the space of the urban community and obtain quantitative spatial image results. High-resolution remote sensing images and a large amount of open-source data provide strong support for the acquisition of experimental data.

The “Tianditu” website (https://www.tianditu.gov.cn/ (accessed on 10 February 2022)) is loaded with geographic information data covering the entire world, of which China’s data cover the whole territory of China from the macroscopic to the microscopic towns and villages. The geographic information service website with the most complete data resources in the region. This article downloads the high-resolution remote sensing image of the GCS_WGS_1984 coordinate system with a resolution of 0.6 m and an update date of February 2022. After vectorization, the community building outline, building and courtyard outline, and road data are appropriately attained. The building outline data are divided based on the function of the building, and the residential buildings and other ones in the community are distinguished through visual interpretation. The road data are divided into urban roads, community vehicle roads, and community pedestrian roads according to road characteristics. Urban roads include vehicular roads outside and through the community, vehicular roads of the community include those placed within the community that allow cars to pass through, and pedestrian roads within the community. The length, area, centroid, and other spatial feature information of vector data are obtained through geographic information systems (GIS) calculation geometry.

The AutoNavi Map Open Platform (https://lbs.amap.com/ (accessed on 17 February 2022)) provides a wealth of map products and services, which can supplement the spatial data not contained within the image. By invoking the open-source data interface made available by the AutoNavi map platform, the information on the number of buildings in the understudied area and all points of interest data (hereinafter referred to as POI) can be obtained, and the data in the AutoNavi map coordinate system can be converted to the coordinate system through QGIS software, the converted POI data can be accurately matched with the community buildings in the WGS_1984 coordinate system, and there is no spatial conflict with the road features.

To ensure the accuracy of the data, we conducted two field investigations in the under-researched area, took photos and recorded the community space, confirmed, and corrected the existing data, and supplemented and improved the missing data.

2.3. Build a Methodology

2.3.1. Selection and Calculation of Spatial Indicators

At present, in the study of the city image, the selection of influencing factors on the image is more focused on the individual factors of urban residents, such as the residents’ gender, age, education, and income [46], while the direction of urban spatial structure and spatial environment are less paid attention to. The influence of residents’ factors on forming urban spatial images is limited in terms of influence [47]. People’s perception and memory of the city still remarkably rely on the characteristics of the spatial environment itself as well as the urban spaces that people see and feel. The structure is of overriding importance. Therefore, in the study of community spatial image, this study is aimed to select 11 sub-categories from the perspectives of community spatial form and spatial service [48,49,50,51,52,53], respectively, reflecting the results of community spatial structure identification and community spatial environment vitality perception (see

Table 1. Community spatial evaluation indicators.

Evaluation Indicator	Indicator Decomposition	Indicator Meaning
Spatial Form	Plot Ratio	Building crowding
	Road Network Density	Ease of access
	Population Density	Degree of crowding
	Average Number of Floors	High density
	Building Coverage	Building density
	Per Capita Green Area	Degree of greenery
Space Service	Shape Index	The degree of regularity and the degree of closeness within the region
	Compactness	Dispersion degree and internal stability of ground objects
	POI Distribution	POI service scope
	POI Aggregation Degree	POI service capability
	Road Vitality	Road service capacity

).

The plot ratio, road network density, population density, average number of floors, building coverage, green area, and compactness per capita can all be evaluated based on general formulas. The calculation and analysis methods of the remaining indicators are briefly given in the following:

(1) The Boyce-Clark shape index based on the radial radius (hereinafter referred to as SBC) is extensively employed in shape analysis. The index was proposed by Boyce and Clark in 1964 [54]. The basic idea is to obtain a relative index by comparing the shape of the understudied area with the standard circle, which is used to express the regularity of the area and reflect the closeness of the internal connection.

The calculation formula is:

$$SBC={{\displaystyle \sum }}_{i=1}^n{\displaystyle \mid }\frac{r_i}{{{\displaystyle \sum }}_{i=1}^n{r}_i} \times 100\% -\frac{100}{n}{\displaystyle \mid },$$

(1)

in which n represents the number of radiations from the area’s center point to the area’s boundary, and $r_i$ denotes the length of the radiation from the center point to each boundary. If the graph gets closer to a circle, that is, the distance from each direction to the center gets shorter, and the SBC value of the graph becomes smaller. This means that the connection between the central point of the graph and each boundary becomes stronger [55,56].

For a community space with a combination of multiple buildings and courtyards, the discrepancy in the spatial form of the courtyards enhances the discrepancies in the spatial experiences of residents. By taking the activity center of the building as the center point, we made 32 radiations and evaluated the radioactivity index (SBC) of each building and courtyard in the Jinghu community.

(2) The distribution pattern of POIs can be made more clearly by constructing grid fishing nets in the community space. During the construction of a fishing net, the side length of the grid unit should not be greater than the minimum building width in the community, ensuring that the buildings in the community will not be included in a single grid unit. By counting the number of POIs in the fishnet cells, the spatial distribution characteristics of POIs in the community space can be intuitively revealed. The spatial distance of each cell reaching the nearest POI is extracted through spatial statistics, and the distance value is divided into various levels to display the service scope and intensity of POIs in the community space. Alternatively, the average value of the spatial distance values of all cells in a specific area to the nearest POI can be evaluated to obtain the average distance of the POI in the area, realizing the quantification of this indicator.

(3) The POI is essential in community spaces as facilities provide various services. The POI aggregates in the form of points to configure a linear or planar service space, which becomes the frequently visited area by residents in the community and promotes the formation of different image elements in the process of residents’ cognition [57]. The kernel density estimation method represents a specific approach to extracting the spatial aggregation characteristics of POI data. The basic idea is to exploit a smooth moving window to estimate the density of points or linear elements and to evaluate the data aggregation of the entire area according to the sample data to produce a continuous density surface [58].

The calculation formula reads:

$$F\left(x\right)=\frac{1}{nh}{{\displaystyle \sum }}_{i=1}^{n}K\left(\frac{{x - x}_i}{h}\right),$$

(2)

where n denotes the total number of sample points, K(x) represents the kernel function, and h is the bandwidth. The bandwidth is a parameter that defines the size of the smoothing window. The larger the bandwidth, the smoother the kernel density estimation result [59]. Regarding the kernel density estimation research of the comprehensive urban scale, the choice of kernel function has little impact on the result of kernel density estimation. In contrast, the choice of bandwidth will have a substantial effect on the estimation result. The bandwidth setting chiefly relies on the analysis scale and geographical characteristics. By comparing various bandwidths, the local and global characteristics of the data distribution can be better reflected. In order to better reproduce the spatial characteristics of POI in the Jinghu community and identify small-scale agglomeration areas, we selected multiple bandwidths for the experiment. We set the values 30, 40, 50, and 60 m as the community’s POI kernel density for the bandwidth calculation and employed the natural breaking point methodology to categorize the obtained results.

(4) As an important linear space for residents’ activities, in addition to the basic traffic function, the road also acts together with the POI along the street to provide communication services to community residents [60,61]. By calculating the POI density on different road sections, that is, road vitality, the service intensity of POI on the road in different spaces and the importance of different road sections can be clearly reflected. Use Thiessen polygons to calculate the POI 50 m coverage area in the community space, establish a 50 m buffer for urban roads and a 20 m buffer for community roads in the study area, and calculate the POI density on each road section to obtain the corresponding road vitality value.

2.3.2. Construct an Image Structure Based on Spatial Indicators

To construct the image structure at the community scale, the subsequent two steps are followed.

The first step is to determine what common features exist between the analysis results of the spatial indicators and those of the image elements. Through tight observation, we realized that the analysis results of spatial indicators and image elements are divided into three categories with the same morphological characteristics. On the one hand, based on the element composition of the city image theory and the spatial form of the constituent elements, it is further subdivided into three types of image elements: path (edge), district, and node (landmark). On the other hand, starting from the scope of action of 11 spatial indicators, according to the organizational form of the elements in the spatial analysis results of each indicator, it is summarized based on the planar, linear, and point features [62].

The second step is to link the analysis results of spatial indicators with image elements through common features. To this end, the following data and results should be appropriately organized: (i) the information exhibiting the area features in the spatial analysis results to match with district elements; (ii) the information presenting linear features in the spatial analysis results to match with path (edge) elements, (iii) metric calculation results, aiming at matching against node (landmark) elements.

Through the above two steps, the calculation and analysis results of spatial indicators are accordingly divided into three categories, which correspond to three categories of image elements, which constituted the image structure at the community scale (see Figure 2).

2.3.3. Organization of Multiple Indicators Results

Among the indicators introduced in Section 2.3.1, some indicators can be evaluated by formulas to obtain specific calculated values, and some others should be processed by spatial analysis tools to obtain geographic analysis results in the form of the vector or raster. For both types of results of various natures (computed values and layered results), a suitable way should be sought to organize them [63].

(1) Overlay analysis

The spatial analysis results pertinent to various indicators can be organized through overlay analysis. By performing overlay analysis, the data overlay of various levels and types is realized, logical operations on spatial data attributes are carried out, and multiple layers are stacked to configure a new data layer, leading to comprehending the combined analysis of the results of multiple indicators [64].

In the image structure at the community scale, the index results associated with path (edge) elements and node (landmark) elements are computed through spatial analysis tools. Therefore, overlay analysis tools are implemented to organize the index results of these two types of elements. The weight distribution of various layers should be taken into account during the superimposing. Generally, the weight value is determined by the influence of various elements on the community. This research is aimed to combine the spatial characteristics of the Jinghu community in the summary of Section 3.2.2 and Section 3.2.3. The organization process results of class features are also demonstrated in detail. It is worth mentioning that when the methodology is applied to other communities, there is no need to employ a unique weight value in the overlay analysis, but it is recommended to suitably set the weight of various elements based on the characteristics of the community space [65].

(2) TOPSIS model based on the weight improvement

The calculated values of different indicators correspond to different dimensions, and simple mathematical operations cannot realize the organization of these calculated values [66].

The technique for order preference by similarity to ideal solution (TOPSIS), also known as the distance method between superior and inferior solutions, can make full use of the information of the original data and has no restrictions on decision-making factors [67]. This method describes the closeness of the evaluation object to the ideal solution through the scores of each program and accurately reflects the gap between the evaluation programs.

The TOPSIS method is implemented by building a model and running the code, which chiefly consists of three steps. The first step is to forward the original matrix by converting all index types into unified types [68]. The second step is to normalize the matrix for eliminating the influence of various index dimensions [66]. The third step is to calculate the normalized score and calculate the score of each scheme to evaluate the discrepancy between the pros and cons of each scheme. The actual is to calculate the Euclidean distance between each index of the current evaluation object and the ideal solution [66,69].

However, in the distance calculation process in the third step, the model defaults to setting the weights of each index to be the same. This setting will cause the evaluation results to deviate significantly from the actual ones. Therefore, this paper adds appropriate weights to the distance calculation process for optimization. It makes the evaluation results more reasonable [70].

For each indicator value ($y_{ij}$, i = 1, 2, 3, …, n, j = 1, 2, 3, …, m) whose dimension influence has been eliminated, the maximum (the minimum) value set of all indicators is specified by $Y^{+}\left(Y^{-}\right)$.

The distance between the i-th number evaluation object and the maximum (minimum) value is defined by the following relations:

$$D_i^{+}=\sqrt{{{\displaystyle \sum }}_{j=1}^m{(Y_j^{+} {- y}_{ij})}^2},$$

(3)

$$D_i^{-}=\sqrt{{{\displaystyle \sum }}_{j=1}^m{{(Y}_j^{-} {- y}_{ij})}^2},$$

(4)

The weight term ${\varphi }_j$ is now introduced to Equations (3) and (4) to arrive at more precise distance values. Hence, the revised distance relation is given by:

$$D_i^{+}=\sqrt{{{\displaystyle \sum }}_{j=1}^m{\varphi }_j{{(Y}_j^{+}{ - y}_{ij})}^2},$$

(5)

$$D_i^{-}=\sqrt{{{\displaystyle \sum }}_{j=1}^m{\varphi }_j{{(Y}_j^{-}{ - y}_{ij})}^2},$$

(6)

The commonly utilized indicators for evaluating the results of community spatial cognition have significant differences in dimensions, and the number of evaluation objects has considerable uncertainty. The analytic hierarchy process (AHP) and the entropy weight method (EWM) are among the standard methods employed for the quantitative analysis of qualitative decision-making problems [71].

The AHP represents a weighting methodology with strong subjectivity in the implementation process [72,73]. It transforms human judgment into the comparison of the importance of several factors by establishing a multi-level structure. However, in the application of the AHP, the evaluation objects and indicators should not have too many factors. The EWM is a more objective weighting approach, which computes the weight according to the information of the data [67]. As a measure of the degree of disorder in the system, entropy is inversely proportional to the amount of information. According to this rule, the inherent information of each scheme in the evaluation is employed to obtain the information entropy of each index. The smaller the entropy value, the lower the disorder degree, the greater the utility value of the information, and the higher the influence of the index on the comprehensive evaluation.

For community space, the spatial scale is small, and each has its own characteristics. If the AHP is employed for empowerment, the difficulty and workload of constructing a target system and scoring community indicators will be considerable. Furthermore, the subjective evaluation system’s decision-making approach is inappropriate for the situation where the community evaluation indicator data are known. The EWM can determine the weight according to the indicator data which describe the community space. Therefore, the weighting result reflects the characteristics of the community space itself, which is more helpful for mining people’s cognitive results of the community space. As a result, the EWM is chosen to determine the weight ${\varphi }_j$ of each index.

2.3.4. Modules of Methodology

Figure 3 illustrates the modules of this methodology and the steps to apply it in the com-munity space. Module 1 denotes the preparation work for the experiment, and Module 2 and Module 3 represent the process of selecting and calculating spatial indicators. After achieving the results of the spatial indicators, the community spatial image structure is first constructed, and then the results of multiple spatial indicators are organized by the method in Module 4, and the results of three types of image elements of the community space are obtained respectively (Module 5). The final image map (Module 6) represents the output of the entire method.

3. Results

3.1. The Results of Calculation and Analysis of Spatial Indicators

3.1.1. Plot Ratio

According to the total aboveground construction area and planned land area of the Jinghu community, the plot ratio of this community is in the range of 1.991–2.433, neither exceeding the limit of 2.5 (see Figure 4a). The homogeneity of the distribution of buildings in the Jinghu community and the stability of the floor area ratio make the residents’ perception of the community space not significantly different. However, when comparing different spaces within the same community, the areas with the highest or lowest plot ratio exhibit differences in spatial cognition results.

3.1.2. Shape Index

The shape index of each courtyard is calculated using the formula. The index ranking result is positively correlated with the regularity and negatively correlated with the closeness of the connection between the seven courtyard activity centers and the area (see Figure 4b). The smaller the index is, the stronger the connection between the residents living in the building and the central point will be, the more frequent utilization of the central point, and the more accurate the spatial location cognition of the element by various residents. With the increase of the index, the connection between the dominant point and the boundary of the courtyard gradually weakens, and the probability of the central point becoming an image element in the residents’ cognitive results lessens.

3.1.3. Spatial Analysis of POIs

The spatial distribution of POIs in the Jinghu community demonstrates prominent linear characteristics. The coincidence of the POI distributed in the linear form with the path and edge elements in the spatial position deepens the residents’ awareness of the relevant elements. The kernel density estimation results of POIs reveal that the areas with high POI agglomeration are mostly the connections between courtyards, reflecting that the connections are often visited by community residents, which verifies the importance of the connections in community life services (see Figure 5).

The results of the comprehensive spatial analysis indicate the impact of different organizational forms of POIs on residents’ cognitive process: the outward-oriented service POI mainly denotes catering services, linearly distributed, and the storefronts are all facing the outside of the building, the provided service for the interior of the building is weak, the given service space is easy to form linear or planar image elements in the residents’ cognitive results (take the No. 1, No. 2, No. 3, and No. 6 courtyards as examples). The internal service POI essentially represents the life service, discretely distributed in the building, primarily serving the residents inside the building, and the service space is easy to form planar or point image elements (take the No. 5 and No. 7 courtyards as examples); residents are considerably more aware of the former than the latter.

3.1.4. Road Vitality

The result map of the road vitality of the community (see Figure 6) displays that the red and orange parts with high road vitality value are the road sections with bottom-level commercial layouts, which have strong service capabilities and exhibit a more substantial influence and spatial permeability on the community space [74]. The residents’ awareness of these road sections will be noticeably higher than other road sections and correspondingly become the path elements with higher frequency in the residents’ cognitive results.

3.2. The Results of Community Spatial Image Using Methodology

3.2.1. District Features

Among the eleven measurement indicators, the calculation results of the spatial form indicators are all the mean values of the corresponding regions. The spatial analysis results of the spatial service indicators are more likely to present planar features associated with the district elements. The seven buildings in the Jinghu community are taken as the corresponding district of spatial indicators, and the types of spatial morphological indicators are determined by qualitative analysis. The weights are calculated based on the EWM (see

Table 2. The indicator calculation results and weight of each courtyard.

District	Plot Ratio	Average Number of Floors	Building Coverage	Road Network Density	Population Density	Per Capita Green Area	Compactness	Average POI Distance
Court No. 1	2	10.84	0.216	19.886	5.029	15.598	0.875	22.252
Court No. 2	2.266	11	0.172	19.201	5.425	15.261	0.848	21.541
Court No. 3	2.306	15.112	0.228	20.257	5.394	14.306	0.88	15.012
Court No. 5	1.991	16.722	0.127	18.024	3.173	27.512	0.868	30.716
Court No. 6	2.433	13.947	0.232	17.993	5.759	13.330	0.833	21.994
Court No. 7	2.269	20.583	0.137	15.499	6.144	14.053	0.757	41.666
Court No. 8	2.251	13.8	0.231	40.628	4.207	18.281	0.875	39.351
Entropy Results	0.110	0.138	0.101	0.203	0.066	0.219	0.056	0.106

), and then these are input into the TOPSIS model to get the scores of each district. The score value has no strong significance, but the comparison of multiple data sets effectively reveals the discrepancy between the results and realizes the quantification of the cognition of district elements [75].

Residents in a small-scale community space have a high perception of the whole space. If the eight indicators are divided into very large and very small based on the law of benefits, the higher the comprehensive benefit of the indicators, the higher the regional score. Further, the scoring result reflects the discrepancy in the comfort level of community space, which is not wholly consistent with the spatial cognition degree. Since the Jinghu community has prominent homogeneity characteristics, the heterogeneity degree of the district better reflects the characteristics of spatial cognition. Therefore, the indicator classification standard is altered from the very large and the very small to the intermediate and the interval types so that the regional score is related to the spatial cognitive characteristics. For various indicators, the closer to the normal value, the higher the score, and the weaker the corresponding degree of spatial cognition (see Figure 7).

The elements with prominent planar characteristics are screened out by comparing POI kernel density estimation results with various bandwidths. Subsequently, these are superimposed with the district results of the courtyard to get the community district cognition result integrating multiple spatial indicators (see Figure 8). After superimposing, a unified symbol system is exploited for expression, and weights are assigned according to the influence degree of the two parts of the results on the community cognition. The area formed in the kernel density estimation result is based on the residents of the whole community, and its importance is higher than that of the courtyard residents. According to the cognitive results of the district elements of the Jinghu community, the POI in the inner community space has the highest degree of aggregation. It substantially impacts residents’ lives and forms the area with the highest degree of cognition in the community space. With the increase of the spatial distance, the POI’s spatial influence weakens, and residents’ cognition accordingly reduces. As for the building area in the community, although the overall architectural form exhibits prominent characteristics of homogeneity, the No. 8 courtyard, as a demonstration district of the community, exhibits complete intelligent facilities and intelligent display boards. There is only a single residential building, but the roads and greening in the courtyard are complete, and the residential functions of other courtyards are dissimilar. The tremendous difference prompts community residents to form a deeper cognition.

3.2.2. Path (Edge) Features

As a typical gated community, the Jinghu community has definite edges and access control equipment for courtyards in the community. Although residents feel the existence of edges in their daily activities in the community space, the residents weaken the edge elements and overlap with the path elements in the cognitive process due to the homogeneity of the community edge and the presence of the side-by-side path. Therefore, when conducting community-scale spatial image research, the edge elements that appear parallel to (or adjacent to) the path are merged into the path elements, and the cognition of the corresponding path elements is deepened.

The urban roads and community vehicle roads within the Jinghu community are taken as the path elements corresponding to the spatial indicators. Subsequently, the influence degrees of various indicators on the path elements are appropriately superimposed. The calculation results of the vitality of each path in the community are then classified based on the natural breaking point method, and the classification result represents the initial weight of each path. The weight is proportional to the cognition degree of the path in the community space. The weight value of the roadside by side with the boundary factor grows to superimpose the influence of the boundary factor on the cognitive result. By comparing the estimation results of the POI kernel density with various bandwidths, the weights of the road sections that overlap with the area space with linear features increase. Finally, based on the weight value of each path in the community, the unified symbol system is utilized to express, and the cognition result of the community path (edge) integrating multiple spatial indicators was obtained (see Figure 9). From the cognitive results of path (edge) elements in the Jinghu community, the north-south and east-west urban paths in the community space connect the whole community and provide many life services. These crucial elements play a pivotal role in enhancing the convenience of residents’ lives, leading to a deeper level of cognition in the community space, and the cognition level of path sections magnifies with the enhancement of service capabilities.

3.2.3. Node (Landmark) Elements

The nodes in the city image results do not have a specific shape and are transformed according to the size of the spatial scale. It is difficult to precisely express the distinction between nodes and districts in a community space with limited features and scope. Landmarks have distinctive features in the elements of city image; however, most community spaces in the city follow the neighboring organization’s law during construction. It is necessary to consider the connection between adjacent buildings and spaces in the community and maintain a relatively consistent style, as a result, few landmarks with apparent features have appeared in the community space. Therefore, the nodes and landmarks are integrated into examining the urban spatial image at the community scale. The node elements are defined as point feature elements with a high frequency of use in community space.

As a relatively independent living unit in an urban space, the community must provide residents with places for daily communication. Activity centers or fitness equipment locations often become places residents use frequently. The activity center is responsible for the internal communications and interactions of the community space [76,77]. Transportation is one of the vital ways for community residents to connect with the outside world. Traffic stations and significant road intersections also correspond to the high frequency of use and serve the residents of the entire community.

The shape index calculated for each building in the community represents the connection between the activity center and the radio residents, matching the importance of the activity center in the residents’ cognitive results. According to the spatial distribution as well as attribute information of traffic stations and road intersections, the spatial serviceability of the elements is judged, and the factors are classified into levels. The elements with prominent point characteristics in the kernel density estimation results are then superposed. Finally, the cognitive results of the community node (landmark) that integrated multiple spatial indicators are achieved (see Figure 10). When superimposing, a unified symbol system is employed for expression, and weights are assigned according to the influence of activity centers, traffic stops, and road intersections on community cognition. Traffic stations and road intersections are utilized by the residents of the whole community, which is more crucial in the cognitive results than the activity centers within the service community. Judging from the cognitive results of the node (landmark) elements in the Jinghu community, nodes with traffic service functions in the community space all correspond to higher cognitive levels. These nodes are essentially distributed in the middle and northwest of the community. The nodes in the middle of the community represent vital transportation hubs within the community, and the nodes in the northwest are the intersections and stations that community residents utilize most when entering or leaving the community.

3.2.4. Image Results and Characteristics of Community Space

The three parts of the spatial cognition results of the community space are superimposed and drawn. The map starts from the planar, linear, and point features of the spatial elements and states the cognition degree of various elements in the community space through the symbol system, forming three types of image elements. Figure 11 illustrates a community spatial image map based on the spatial analysis.

Judging from the spatial image map of the Jinghu community, these three elements can cover the entire community space and express the spatial characteristics of the Jinghu community more clearly. The three elements have similar properties to the city image elements: First, the cognitive results of various image elements in the community space have their characteristics. The dark red planar elements in the picture denote the districts with high cognitive levels, consisting of two parts. One is the No. 8 Courtyard, which is quite different from other courtyards in terms of spatial form and characteristics. It belongs to the heterogeneous community space. The other part is the district with many POIs distributed on the edge of the two columns of courtyards, which is the hot spot for residents’ activities in the community space. The dark blue and bold linear elements of the picture denote the paths with a high degree of cognition, mainly on urban roads in the community. Continuous shops on both path sections strongly influence the community and spatial permeability [78]. It is adjacent to the edges of courtyards in the community, which enhances the recognition of the residents and memory of the linear elements. The dark green point elements with prominent symbols in the picture are the nodes with high cognition levels, mainly traffic stops and road intersections with high service rates. These serve the residents of the whole community and are vital elements in the travel and daily activities of residents of the Jinghu community.

Second, various image elements in the community space are interconnected. The same data in the community space impact the cognitive results of these three types of elements and are transformed in the process of cognition of various elements. Furthermore, the elements with a higher cognitive level in the Jinghu community spatial image map always appear in combination. Districts with high cognition often include paths with a high frequency of use. Nodes with high cognition mostly appear near paths with high cognition. The highest cognition of the three types of elements in the whole community space is concentrated near the No. 3 Courtyard, which represents a “hot spot” within the community.

4. Discussion

By surveying the progress of urban image research in the literature, there are still few image studies on urban community space, and the quantitative methods commonly used in image research are difficult to apply at the community scale due to data limitations. There exists a large scientific gap in the quantitative research of image cognition. Therefore, this study initiates from the community space in the city, attempts to change the previous research approaches of the urban imagery, and aims to exploit a quantitative method to achieve the results of the spatial imagery of the community. The methodology and results of the whole scrutiny are discussed in the following.

4.1. Advantages of the Data and Quantification Methods

The city image research proposed by Lynch [8] requires numerous questionnaires in the understudied area, which costs researchers a lot of time and energy. Therefore, in acquiring experimental data, a moderately easy and efficient way is implemented using high-precision remote sensing images and open-source geographic databases as source of experimental data. Through quantified high-precision remote sensing images [79], spatial data in the community, mostly building and road elements, are achieved. At present, researchers can obtain high-precision imagery covering most cities in the world. Although image vectorization requires visual interpretation to guarantee the accuracy of the data, it takes far less time than a questionnaire. By downloading open-source geospatial data, the property information of POIs, buildings, and road elements in the community space can be appropriately extracted, and the spatial information that cannot be gained from remote sensing images is supplemented. Currently, the application programming interface of open-source websites has become one of the crucial channels for researchers to obtain data. Commonly, the data obtained on usual open-source websites have high accuracy and aims to enjoy fast update speed and data types. There are plentiful advantages of convenient data acquisition.

In the existing research on the urban micro-spatial image [80,81,82], researchers usually choose to analyze questionnaires and cognitive sketches to obtain spatial image results. However, relying only on questionnaires and cognitive sketches to describe the research objects is too subjective, and cognitive results lack quantitative analysis result support. Spatial analysis is a quantitative analysis approach for geospatial data and geospatial phenomena. This methodology digs out potential information of spatial objects by describing, analyzing, and modeling the spatial relationship of geographic elements, and finally outputs vector or raster layer to comprehend the visual expression of spatial information. In the examination of image cognition at the community scale, the spatial analysis approach is therefore exploited as a quantitative tool.

4.2. Analysis of Parameter Settings in Methodology

This study takes community spatial data as input, and after using the method, outputs an image map that describes the residents’ awareness of the space in the community. The key to the method is to organize the results of spatial indicators based on the community spatial image structure. In the process of organizing, the two methods of weight-based improved TOPSIS model and superposition analysis are employed. For each methodology application, the parameter settings should be paid attention to.

(1) Setting of weights and index properties in the TOPSIS model

By default, the TOPSIS model sets the weights of each indicator to be the same in the steps of calculating the score and normalizing, and such a setting causes the evaluation result to considerably deviate from the fact [70]. Therefore, when using the TOPSIS model, the EWM is selected that is more suitable for the community space to alter the weight settings of the indicators so that the weights assigned by different indicators conform to the characteristics of the community space.

After addressing the weight settings, the computed values of multiple spatial metrics with the improved TOPSIS model are organized. However, after analyzing the output results of the model, in the case of employing the TOPSIS model in the community space, if the indicators are divided according to the benefits law in the step of converting the indicator types, the obtained scoring results reflect the living comfort in various areas of the community. Further, the resulting discrepancy is not completely consistent with the degree of spatial cognition. When the classification criteria are replaced by with intermediate and interval types, the scoring results output by the model describes the degree of deviation of various areas from the normal, so that the model results can reflect residents’ awareness of different areas.

Taking the plot ratio as an example, it should be classified as a very small index according to the law of benefit. The specific performance is that the lower the floor area ratio, the higher the comfort level of residents, and vice versa. However, it is worth thinking about when the residents’ comfort level is higher (low), the residents’ awareness of the area is higher (low). This is apparently wrong. In the community space, when the conditions of various areas are very similar, the residents’ perception of different areas is not very different. When some areas have considerable deviations from other areas, it often leaves a deep impression on the residents [83]. Therefore, we convert the plot ratio index into an intermediate index and take the average plot ratio of different areas in the community as the intermediate value. The farther from the intermediate value, the greater the difference with other spaces in the community.

(2) Setting of weights in the superposition analysis

When performing overlay analysis on multiple layers, we should set weights for various layers, so that the combined results correctly reflect the overall cognitive results of the community. This paper does not offer specific weight settings when proposing the method set, because when the methodology is applied to other communities, due to the influence of cultural background or region, the weight of different layers should be reset by researchers according to the characteristics of the community space.

This point may limit the exploitation of the methodology, as the setting of the weights is essentially based on the researcher’s judgment that it should be employed by experts. However, we still hope that this methodology can be employed by more people because the set of weights is also a study worthy of discussion. For instance, when studying a specific community, the weights of different layers can be adjusted during the stacking process, conduct multiple sets of experiments, and draw some interesting conclusions by comparing different stacking results.

4.3. Significance and Application of the Results of Community Space Image Research

City-scale spatial image maps are difficult to reflect the characteristics of other cities due to their uniqueness. However, at the community scale, in the context of China’s rapid urbanization, there are obvious similarities in the development characteristics of cities, particularly in the real estate development field [44]. The new communities in the city exhibit great similarity in terms of spatial location, scale, and arrangement of buildings in the community. Therefore, the results of the research on the spatial image characteristics of the Jinghu community noticeably reflect the situation of similar communities in other Chinese cities.

The novel approach proposed in the present work quantifies the community spatial data into the spatial cognition results of the community and generates an image map. The spatial image map of the Jinghu community displays that this result is different from the image map of the city scale. At the community scale, dissimilar to the five types of urban image elements proposed by Lynch [8], the spatial image characteristics of the community are more comprehensively expressed through only three types of elements: district, path (edge), and node (landmark). This also proves that it is reasonable to use three types of image features as the image structure at the community scale.

Another crucial result of this exploration is to summarize the characteristics of these three categories of elements. Taking the node (landmark) feature as an example, in the case study results of the Jinghu community, the fitness place represents the node element that has a remarkable impact on the lives of residents. However, when exploring various communities, it is affected by regional characteristics, these nodes may be no longer the gyms mentioned in the text, but maybe churches, bookstores, or temples [84]. Although the specific expressions of these node elements alter, their characteristics do not change. These are all node elements that residents frequently visit and exploit in the community space.

Cognitive processes at the community scale are generated by daily life practices, represent discrepancies in the living spaces of residents, and thus guide practice. The community space image map apparently expresses the residents’ perception of the community space. This information is utilized as a reference for planners to assist them to understand the rationality of community space planning and promote better renewal and renovation of community space.

Taking the Jinghu community as an example, from the image map, the overall awareness of the Jinghu community presents the characteristics of “high in the west and low in the east” and “high in the inside and low in the outside”. On the one hand, the combination of image elements with higher cognitive levels appears, all concentrated in the surrounding area of the No. 3 courtyard on the west side. These areas are not only distributed with numerous external service shops and bus stops, but also contain crucial road intersections connecting community spaces. As a result, they have become “hot spots” in the entire community, leaving deep memories in the process of community residents’ cognition. On the other hand, the arrangement of the courtyards in the Jinghu community makes the connection within the community close, and the service level of the space inside the community is higher. The closer the area to the edge, the weaker the connection. The processes are easily overlooked. Through these results, the Jinghu community can be provided with two aspects of space renovation suggestions. The first is to pay attention to the space, that is, easy to leave deep memories in the cognitive process. These spaces play a vital role in residents’ community life. Residents have a high probability of using these spaces. Therefore, it is necessary to pay attention to the management of the area. By increasing street service facilities, expanding vegetation coverage, and beautifying the building interface, the community can be improved residents get a better user experience. The second is to improve the space where the cognitive processes are readily overlooked. These spaces are less dynamic and therefore need to be more attractive and can be made to play a greater role within the community by adding more shops and increasing access roads.

In future research works, we can replicate this methodology in community spaces in various regions, gain image maps of community spaces, and realize the residents’ awareness and characteristics of community spaces. This result will be employed as a reference for community planning and renovation to create a community environment that is more suitable for residents’ living.

5. Conclusions

The cognition of spatial units at various scales exhibits various spatial and temporal scale effects, and the quantification methods applicable to the study of images at the urban scale cannot be directly reflected in the community space. Therefore, aiming at the problem of insufficient quantitative research on community spatial image cognition, this paper is aimed to propose a methodology suitable for the exploration of spatial image cognition at the community scale and takes the Jinghu community as a case to demonstrate the application of this method set. After analyzing the experimental process and results, the major results are summarized as follows:

(1) By using this methodology, the community spatial data are quantified as the spatial cognition results of the community. From the spatial image map of the Jinghu community, the spatial image features of the community scale are expressed more comprehensively through three types of elements: district, path (edge), and node (landmark), while edge and landmark elements are not at the community scale, and their pertinent results are separately outlined in the following.

(2) In the image results at the community scale, the district elements with higher cognition are essentially composed of two parts; one is the heterogeneous area in the community space, and these areas exhibit a large degree of deviation from other areas in the community. The other part is the hot spot for residents’ activities in the community space. These areas often contain POIs, which provide residents in the community with services such as shopping, sports, and transportation.

(3) In the community-scale image results, paths with a large number of POIs distributed and arranged parallel to the edge tend to be linear elements with much residents’ cognition. In the community space, in addition to providing traffic functions, road elements also work together with POIs along the street to offer communication services to community residents, deepening residents’ awareness of the corresponding road sections; edge elements are clearly perceived but are not easily recognized as an image element alone, and the results of residents’ cognition of boundary elements will be superimposed on the roads arranged parallel to them.

(4) In the image results at the community scale, the point elements do not exhibit the crucial characteristics of the landmark elements. On the contrary, those non-salient elements that are inseparable from life are more likely to be utilized by residents in the community. Those node elements that are frequently accessed and used by residents in the community space often form a deeper level of cognition in the image map.

At present, the updated speed of geographic data in urban space is getting faster and faster, and the acquisition of open-source data is more convenient. Therefore, the method proposed in this paper will not be limited by experimental data in other spatial applications. We also hope to replicate the method in different neighborhoods and different cities to examine and compare its performance in different spaces. When this methodology is employed in other community spaces, the spatial structure and spatial environment of these communities are different from those of the Jinghu community. When using the superposition analysis approach to organize index results, the weight settings of different elements will be different from those used in the Jinghu community. This requires researchers to set the weights of various elements in combination with the characteristics of the community space, but the group of cognitive result weights that best reflect the community often needs to be given by experts [65].

In general, this paper establishes a methodology, which is different from the existing urban image research methods. The new method does not require much data acquisition and processing, nor does it require a lot of time and energy to conduct research work. By evaluating and analyzing open-source geospatial data, the spatial image map of the community is derived in a quantitative manner via this method, which cannot only directly reflect the residents’ awareness of the space in the community, but also provide suggestions for urban planners to “pay attention to the space that is easy to leave a deep memory in the cognitive process, and improve the space that is easy to be ignored in the cognitive process”, which offers new ideas for the planning and management of urban space.

But the current research on this method is not deep enough. We will follow two goals for our subsequent research works. On the one hand, we will replicate the methodology in multiple different urban spaces, check and compare its performance in different spaces, and combine the experimental results will supplement and improve this method. On the other hand, we will try to add other spatial indicators to observe the impact of changes in indicators on the image cognitive results presented by the community, and to analyze the structure and characteristics of urban micro-space.