Urban landscape bearing capacity evaluation method based on number of interest points

Abstract

A method for evaluating urban landscape bearing capacity based on the number of interest points belongs to the field of urban planning, and comprises the following steps: searching interest points and streetscape images through network crawling, carrying out matching screening and processing, and selecting a road with the interest points and the streetscape images as a research area; analyzing the streetscape image through deep learning and calculating the green vision rate of the streetscape image; quantizing the category of the urban landscape according to the interest points; training a training model obtained by deep learning; inputting a streetscape image for deep learning, and outputting a green landscape index of the streetscape image; calculating an evaluation index according to the quantized interest point and green vision rate; and the urban landscape bearing capacity is evaluated through big data calculation and analysis of different evaluation indexes. According to the method, the relation between the number of the interest points and the urban landscape bearing capacity can be rapidly and accurately calculated, whether the distribution trend of the urban landscape in the urban environment needs to be optimized or not can be truly evaluated, a multiple regression model can be built easily, and the method is suitable for big data analysis.

Description

Technical Field

[0001] The invention belongs to the technical field of urban planning, and in particular relates to an evaluation method for urban landscape carrying capacity based on the number of interest points. Background Art

[0002] The development of a city is not only reflected in the high-rise buildings and high-quality and efficient economic growth, but also in the blue sky and clear water brought by the construction of ecological civilization. The impact of the environment on people has been confirmed by scientific research. The urban landscape affects human subconsciousness and judgment from four aspects: air, light, heat and sound environment. The most important part is the green area in the city. They represent a small "slice" of nature in the city and are the natural colors that humans can touch in the sea of ​​steel and concrete in the city. From the perspective of urban construction, the goals of "ecology" and "livability" have also made the "green" in the city an object of constant discussion, calculation and evaluation; traditional evaluation of "green" mostly uses the NDVI index. NDVI is sensitive to green plants and has a linear relationship with plant distribution density. It is a good indicator of plant growth status and spatial distribution density, but there is expansion in low vegetation coverage areas and compression in high vegetation areas.

[0003] The green view rate, an indicator of green quantity evaluation, can well reflect the differences in the spectral characteristics of vegetation and soil. In addition to reflecting the green quantity, the green view rate also reflects the proportional relationship between greening and roads, which can better evaluate the quality of street space beyond traditional indicators. From a personal perspective, the green view rate that can be seen in the city in daily life is the dose of nature that can be touched, and it is an important criterion for measuring the dimension of urban space and evaluating the green quantity of urban streets; from the perspective of pedestrians, although urban landscape plays a vital role in improving the street environment, the benefits are not necessarily good when the location of urban landscape is behind a large tree canopy. Existing literature shows that there is an unclear relationship between the type and number of points of interest and the quantitative proportion of tree canopy shading in street scenes. POI is the abbreviation of "Point of Interest", which can be translated into "point of interest" in Chinese. In a geographic information system (GIS), a POI can be a house, a shop, a mailbox, a bus stop, etc. The traditional method of collecting geographic information requires map surveyors to use precise surveying instruments to obtain the longitude and latitude of a point of interest, and then mark it down. Because the collection of POIs is a very time-consuming and laborious task, for a geographic information system (GIS), the number of POIs represents the value of the entire system to a certain extent. Each POI contains four aspects of information: name, category, longitude, and latitude. Comprehensive POI information is essential information for enriching navigation maps. Timely POI points of interest can remind users of detailed information about road branches and surrounding buildings, and can also facilitate navigation to find the places you need and choose the most convenient and unobstructed road for route planning. Therefore, the number of POIs on the navigation map directly affects the usability of navigation. In the geographic information system (GIS), the application of POIs is multifaceted. The key is that they provide relevant information about specific locations, which is helpful for data analysis, decision support, planning and navigation.

[0004] Green view rate can quantify people's perceptual experience in the city, and some studies have begun to look for the relationship between the number of points of interest and the carrying capacity of urban landscapes. However, most of the current research on the calculation method of urban landscape carrying capacity focuses on questionnaires and historical data research, and lacks accurate and detailed research on landscape carrying capacity at the urban road level. This is mainly because most of the scholars currently engaged in related research come from the fields of ecology and economics, and do not have the research capabilities of related computer technology, so they cannot complete scientific green view rate identification technology. Summary of the invention

[0005] In order to solve the problem that the existing calculation methods of the quantitative carrying capacity of the types and numbers of interest points and the canopy shading in the street scene cannot obtain accurate results and cannot fully calculate the urban landscape carrying capacity of the number of interest points, the present invention uses computer crawler technology and deep learning to provide an evaluation method for the urban landscape carrying capacity based on the number of interest points. The present invention obtains urban interest points through computer crawler technology, and independently develops a green view rate recognition model through model training. It accurately quantifies the number of urban interest point types and the urban landscape green view rate carrying capacity through the fusion of computer science, ecology and economics knowledge, and uses it in an evaluation method for urban landscape carrying capacity based on the number of interest points.

[0006] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0007] The present invention provides an evaluation method for urban landscape carrying capacity based on the number of interest points, which mainly comprises the following steps:

[0008] Step 1: Collect points of interest and street view images through web crawling;

[0009] Step 2: Match, screen and process the points of interest and street view images, select roads with both points of interest and street view images as the research area, and retain and classify the grids in the area;

[0010] Step 3: Analyze street view images through deep learning and calculate the green view rate of street view images, and establish a large database on the green view rate of street views; quantify the category of each urban landscape according to the number of urban landscapes of commercial types of points of interest in each area, and screen them according to roads; train the training model obtained by computer deep learning; input street view images for computer deep learning, and output the green landscape index of street view images;

[0011] Step 4: Calculate the evaluation index based on the quantified points of interest and green view rate; evaluate the urban landscape carrying capacity through big data calculation and analysis of different evaluation indicators.

[0012] Furthermore, in step one, the network crawling and collecting method is: crawling and downloading points of interest from a map web page or APP application program interface with a map function; the map web page is Amap, Baidu Map and Google Map; the APP application is Amap, Baidu Map and Google Map.

[0013] Furthermore, in step one, the web crawling and collection method is: performing spatial analysis in a geographic information system, setting road random points, and exporting the coordinates of relevant road random points, and using the coordinates of the road random points to crawl and download street view images in a map web page or APP application program interface.

[0014] Furthermore, in step one, the network crawling and collecting method is: photographing street scenes in real streets and counting points of interest on both sides of the road.

[0015] Furthermore, in step 2, after matching, screening and processing the POIs and the street view images, the coordinates of each POI are assigned corresponding road attributes.

[0016] Furthermore, in step 2, when retaining and classifying the grids in the area, only grids with green viewing rates are considered, and other grids are filtered out, and the grids finally retained are valid grids.

[0017] Furthermore, in step three, the points of interest include five commercial types of urban landscapes, namely, catering, shopping, life services, entertainment and hotels.

[0018] Furthermore, in step 3, the specific method for constructing the training model is:

[0019] Collect at least 30,000 street view images at random locations for model training. Unify the format of all street view images and require them to have the highest dpi. Select a certain number of street view images to create pixel-level semantic labels for feature recognition.

[0020] Furthermore, in step three, image semantic segmentation is used to achieve pixel-level semantic labeling.

[0021] Furthermore, in step 4, the evaluation index includes correlation analysis and regression analysis, and the causal relationship between the number of urban landscapes and the green view rate is explored through correlation analysis and regression analysis.

[0022] The beneficial effects of the present invention are:

[0023] The present invention provides an evaluation method for the urban landscape carrying capacity based on the number of points of interest, which solves the problem of the inability to obtain accurate results and the difficulty in evaluation in the existing methods. The present invention adopts big data network crawlers and deep learning technology, uses the map application program interface to retrieve the quantitative results of the urban crawled points of interest, and crawls the street view images of the relevant road sections to analyze the street view images through deep learning to establish a large database on the street view green view rate. Through the calculation of big data, the urban landscape carrying capacity of the number of points of interest is objectively and accurately measured and calculated, and then it is used to evaluate whether the distribution trend of the urban landscape in the urban environment needs to be optimized. Compared with the existing technology, the present invention has the following advantages:

[0024] (1) The present invention can quickly and accurately calculate the relationship between the number of points of interest and the carrying capacity of urban landscape.

[0025] (2) The present invention can truly evaluate whether the distribution trend of urban landscape in the urban environment needs to be optimized.

[0026] (3) The present invention is conducive to constructing a multivariate regression model and is suitable for big data analysis.

[0027] (4) The indicator construction method of the present invention is flexible and has diverse applications and is suitable for various evaluation purposes. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A flow chart of a method for evaluating the carrying capacity of urban landscape based on the number of points of interest provided by the present invention. DETAILED DESCRIPTION

[0029] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0030] See also Figure 1 To illustrate, the present invention provides an evaluation method for urban landscape carrying capacity based on the number of points of interest, and its specific implementation process is as follows:

[0031] Step 1: Collect points of interest and street view images;

[0032] The present invention specifically adopts the following three methods to collect and crawl the number of points of interest (POIs) types and street view images, which are:

[0033] (1) Crawl and download points of interest from a map webpage or APP application programming interface (API) with map function.

[0034] Among them, the map web pages with map functions mainly include: Amap, Baidu Map and Google Map, etc., but are not limited to these.

[0035] Among them, the APP applications with map functions mainly include: Amap, Baidu Maps and Google Maps, but are not limited to these.

[0036] (2) Perform spatial analysis in a geographic information system (GIS), set random road points, and derive the coordinates of relevant random road points. Use the coordinates of the random road points to crawl and download street view images in a map web page or APP application program interface.

[0037] Among them, the map web pages with map functions mainly include: Amap, Baidu Map and Google Map, etc., but are not limited to these.

[0038] Among them, the APP applications with map functions mainly include: Amap, Baidu Maps and Google Maps, but are not limited to these.

[0039] (3) Photograph street scenes in real streets and count points of interest on both sides of the road.

[0040] Step 2: Processing of points of interest and street view images;

[0041] (1) Match, screen and process the POIs and street view images to meet the following requirements: Each POI coordinate is assigned a corresponding road attribute.

[0042] (2) The roads with both points of interest and street view images are selected as the study area. In the present invention, the selected study area is divided into 6037 grids with a side length of about 0.6 kilometers; then, the retention and classification process is applied to the grids in the area. In order to ensure the accuracy of the experiment, only grids with green view rate are considered, and other grids are filtered out. The grids finally retained are valid grids.

[0043] Step 3: POI and street view image analysis;

[0044] (1) Analyze street view images through computer deep learning and establish a large database on street view green view rate;

[0045] (2) The points of interest contain five types of commercial urban landscapes (catering, shopping, life services, entertainment, and hotels). The category of each urban landscape is quantified according to the number of the five types of commercial urban landscapes of the points of interest in each area and filtered according to the roads.

[0046] (3) A training model for image analysis obtained by pre-training computer deep learning, the specific construction method of which is as follows:

[0047] At least 30,000 street view images at random points are collected in cities in all provinces across the country for model training. These street view images need to be in a unified format and must have the highest dpi (dots per inch). At the same time, a certain number of street view images are selected to create pixel-level semantic labels for feature recognition.

[0048] In the present invention, image semantic segmentation is mainly used to realize pixel-level semantic labeling. In the field of computer vision, image semantic segmentation mainly includes two steps, namely: completing image recognition and target location detection at the pixel level: first understand the content and meaning of the image, and then segment different types of objects.

[0049] (4) The street view image is input for computer deep learning, the tree crown pixels are divided by the ratio of all pixels, and the green landscape index (GVI) is used to quantitatively evaluate the tree crown shading degree, and finally the green landscape index (GVI) of the street view image is output.

[0050] Step 4: Calculate evaluation indexes;

[0051] The evaluation index is calculated based on the quantified points of interest and green view rate obtained in step 3; wherein the evaluation index mainly includes correlation analysis to explore the causal relationship between the number of urban landscapes and the green view rate. In addition, regression analysis is also used to explore the causal relationship between the number of urban landscapes and the green view rate.

[0052] In the present invention, all statistical analyses were performed using SPSS (IBM Corporation, Armonk, NY, USA). Using the Shapiro-Wilk test, it was found that most of the data in the present invention did not follow a normal distribution (p<0.05). Therefore, the present invention used Spearman correlation analysis to investigate the relationship between the number of various urban landscapes and the green view rate. This non-parametric method is applicable to data that do not conform to a normal distribution and effectively depicts the distribution trend of urban landscapes in urban environments. In addition, regression analysis was used to explore the causal relationship between the number of urban landscapes and the green view rate, which is a key aspect of understanding urban planning and development strategies.

[0053] Through big data calculation and analysis of different evaluation indicators, the calculation results of the urban landscape carrying capacity of the final number of points of interest are obtained. The specific analysis of the calculation results is as follows:

[0054] (1) Display the correlation between the number of interest points and the green view rate value and find the best regression fitting curve. The relationship between the number of interest points and the green view rate can be fitted by a nonlinear model described by a positive skewness curve.

[0055] (2) In all POI categories, the number of POIs in different regions also varies significantly due to road levels.

[0056] (3) Each POI is predicted using the best regression fitting function based on the number of various POI types and green view rate.

[0057] (4) It makes it clearer that the actual sales in the area are compared with the expected sales. There is a nonlinear relationship between the green view rate and the number of urban landscapes, which is characterized by an initial increase followed by a decrease. This correlation suggests that both excessive and insufficient urban tree canopies may have a negative impact on the growth and development of sales.

[0058] The effect verification test is as follows:

[0059] Test Example 1:

[0060] Taking Changchun City, Jilin Province as an example, after filtering out other grids, a total of 1116 valid grids were retained. The research of the present invention shows that not all types of urban landscapes are affected by roadside tree canopy coverage. Too high (green view rate> 0.08) and too low (green view rate< 0.02) urban canopy levels have a certain negative impact on the growth and development of sales.

[0061] Test Example 2:

[0062] Taking Harbin, Shenyang and Changchun, the capital cities of the three northeastern provinces, as examples, in 2022, we crawled the points of interest around 106 communities of collapsed real estate companies in the three cities, and obtained 732,387 points of interest in 11 categories. The accuracy of the constructed multivariate regression model reached an extremely significant level, and by calculating the urban landscape carrying capacity of the number of points of interest, effective rectification suggestions were put forward.

[0063] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for evaluating the carrying capacity of urban landscape based on the number of points of interest, characterized in that: The following steps are involved: Step 1: Collect points of interest and street view images through web crawling; Step 2: Match, screen and process the points of interest and street view images, select roads with both points of interest and street view images as the research area, and retain and classify the grids in the area; Step 3: Analyze street view images through deep learning and calculate the green view rate of street view images, and establish a large database on the green view rate of street views; quantify the category of each urban landscape according to the number of urban landscapes of commercial types of points of interest in each area, and screen them according to roads; train the training model obtained by computer deep learning; input street view images for computer deep learning, and output the green landscape index of street view images; Step 4: Calculate the evaluation index based on the quantified points of interest and green view rate; evaluate the urban landscape carrying capacity through big data calculation and analysis of different evaluation indicators.

2. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 1, characterized in that: In step one, the network crawling and collecting method is: crawling and downloading points of interest from a map web page or APP application program interface with a map function; the map web page is Amap, Baidu Map and Google Map; the APP application is Amap, Baidu Map and Google Map.

3. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 1, characterized in that: In step one, the web crawling and collection method is: perform spatial analysis in a geographic information system, set road random points, and derive the coordinates of relevant road random points, and use the coordinates of the road random points to crawl and download street view images in a map web page or APP application program interface.

4. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 1, characterized in that: In step one, the network crawling and collection method is: photographing street scenes in real streets and counting points of interest on both sides of the road.

5. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 1, characterized in that: In step 2, after matching, screening and processing the POIs and street view images, the coordinates of each POI are assigned corresponding road attributes.

6. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 1, characterized in that: In step 2, when retaining and classifying the grids in the area, only grids with green view rates are considered, and other grids are filtered out. The grids that are finally retained are valid grids.

7. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 1, characterized in that: In step three, the points of interest include five commercial types of urban landscapes, namely, catering, shopping, life services, entertainment and hotels.

8. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 1, characterized in that: In step 3, the specific method for constructing the training model is: Collect at least 30,000 street view images at random locations for model training. Unify the format of all street view images and require them to have the highest dpi. Select a certain number of street view images to create pixel-level semantic labels for feature recognition.

9. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 8, characterized in that: In step three, image semantic segmentation is used to achieve pixel-level semantic labeling.

10. The method for evaluating the carrying capacity of urban landscape based on the number of points of interest according to claim 1, characterized in that: In step 4, the evaluation indicators include correlation analysis and regression analysis, and the causal relationship between the number of urban landscapes and the green view rate is explored through correlation analysis and regression analysis.

Images