Street type classification method based on Gaussian clustering mixture model

Abstract

This invention discloses a street type classification method based on a Gaussian mixture clustering model. The method comprises the following steps: 1. collecting and intelligently identifying street design attribute data; 2. determining a set of street type attribution indicators; and 3. constructing a street type classification algorithm model. The method implements street classification by quantitatively analyzing street types based on multi-source data, such as streetscape image attributes, and a Gaussian mixture clustering model algorithm.

Description

Technical Field

[0001] The present invention relates to the field of computer classification technology, and in particular to a street type classification method based on a Gaussian clustering mixture model. Background Art

[0002] Currently, many cities around the world have entered an era of stock planning. The focus of urban construction has gradually shifted from large-scale "city building" to small- and medium-scale "creation" and "renewal." In this context, people's attention to street space will gradually shift from a "form-oriented" to a "life-oriented" approach. Street renewal and management will also shift from a "two-dimensional" perspective to a more refined "three-dimensional space." This, in turn, creates a more urgent need for human-oriented analysis and design. In this era of stock planning, street space requires more refined measurement to support quality-oriented spatial design.

[0003] Streets are urban public spaces that are closely linked to people's production and daily lives. As a component of urban infrastructure, they possess both uniform and general characteristics. Different types of streets have distinct functional emphases and differing quality requirements.

[0004] Scholars and experts in various countries and cities have proposed various street classification methods. For example, the Los Angeles Street Design Guidelines categorize streets into main streets, motorized streets, bus-pedestrian streets, bicycle boulevards, festival streets, and shared spaces; the Chicago Complete Streets Guidelines categorize streets into pedestrian streets, service roads, community streets, main streets, connecting roads, and arterial roads; the New York Street Design Manual categorizes streets into general vehicle lanes, image-lined boulevards, public transportation lanes, community slow lanes, and fully pedestrianized streets; and the Shanghai Street Design Guidelines, which comprehensively considers factors such as street-side activities, streetscape characteristics, and transportation functions, divides streets into five major types: commercial streets, life service streets, landscape and leisure streets, transportation streets, and comprehensive streets. These street classification methods often use qualitative analysis.

[0005] Different types of streets can provide functions with different focuses. We cannot adopt the same standards and measures in urban renewal and quality improvement. We need to conduct differentiated analysis of quality improvement measures for different types of streets.

[0006] In order to subsequently propose scientific, effective, and targeted measures to improve street design quality, it is urgent to scientifically and rationally define street types. If there is a technology that can quantitatively analyze and judge street types using highly digital methods based on existing traditional qualitative street classification methods, it can provide certain prior knowledge for subsequent differentiated street quality improvement measures.

[0007] Therefore, how to implement street classification based on multi-source data such as street view image attributes has become a technical problem that technicians in this field urgently need to solve. Summary of the Invention

[0008] In view of the above-mentioned defects of the prior art, the present invention provides a street type classification method based on a Gaussian clustering mixture model, the purpose of which is to achieve street classification by quantitatively analyzing street types based on multi-source data such as street view image attributes and a Gaussian mixture clustering model algorithm.

[0009] To achieve the above object, the present invention discloses a street type classification method based on a Gaussian clustering mixture model, comprising the following steps:

[0010] Step 1: Collection and intelligent identification of street design attribute data;

[0011] Step 2: Determine the street type attribution index set;

[0012] Step 3: Construct a street type classification algorithm model.

[0013] Preferably, step 1 comprises the following steps:

[0014] Step 1.1: Obtain the spatial morphological data of the street, as follows:

[0015] Use a pan-tilt camera to take pictures of city streets, and then correct the pictures of city streets through field surveys to build a street space morphology database;

[0016] The database elements of the street space morphology database include land use nature, land use boundary and land use area;

[0017] Step 1.2: Obtain street view image data of the street, as follows:

[0018] Then, based on a fully convolutional complex neural network and deep learning methods, an intelligent recognition program is used to perform computer intelligent recognition of the lane elements, sidewalk elements, and motor vehicle elements in the acquired street view image data, and statistical analysis is performed on the pixel data occupied by each element to obtain the number and proportion of pixels of each type of element;

[0019] Step 1.3: Determine the street design space quality evaluation indicators and screen them from the objective evaluation dimension and subjective evaluation dimension respectively;

[0020] Among them, the objective evaluation dimensions include travel feasibility, road network accessibility and facility convenience;

[0021] The pedestrian feasibility is evaluated by the pedestrian access index (SFI), which is the ratio of the number of pixels of pedestrian paths and vehicle paths in all street scene images. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows:

[0022]

[0023] Among them, w n is the total number of pixels occupied by walking space in the street view image numbered n; R n is the number of pixels occupied by the total vehicle space in the street view image;

[0024] The road network accessibility is evaluated by the road network density index (RDI), which is the ratio of the number of pixels occupied by the roadway in the street view image to the total number of pixels in the image. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows:

[0025]

[0026] Among them, R n A is the number of pixels occupied by the total vehicle space in the street view image, n is the total number of pixels in the street view image, that is, the sum of all pixels in the face area of ​​the image;

[0027] The convenience of the facilities is evaluated by the service facilities satisfaction index (PSI), which is the ratio of the number of pixels occupied by the service facilities in all street street view images to the total number of pixels in the image. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows:

[0028]

[0029] Among them, p n A is the number of pixels occupied by service facilities in the street view image numbered n, that is, the sum of the pixels of the area of ​​i facilities in the image; n is the total number of pixels in the street view image, that is, the sum of all pixels in the face area of ​​the image;

[0030] The subjective evaluation dimensions include walking safety, space comfort and space friendliness;

[0031] The pedestrian safety is evaluated by the vehicle interference index VII and the traffic sign index ITI, both of which are negatively correlated with the street design space quality evaluation index;

[0032] The motor vehicle interference index VII is the sum of the number of pixels of motor vehicle patches in all the street scene images and the number of pixels of motor vehicle lanes in the entire image. The specific formula is as follows:

[0033]

[0034] Among them, C n represents the pixels of the motor vehicle patches identified in the street view image, R n The total number of pixels in the motor vehicle lane in the entire image. The higher the vehicle interference index, the greater the proportion of motor vehicles in the street space, and the lower the sense of security.

[0035] The traffic sign index (ITI) is the ratio of the number of pixels of traffic lights and signs in the street view image to the total number of pixels in the image. The specific formula is as follows:

[0036]

[0037] Among them, T n is the pixel area of ​​traffic lights and traffic signs in the street view image numbered n, that is, the sum of the area pixels of i doors and windows in the image; R n is the total number of pixels in the street space in the image, that is, the sum of the pixels in the i-th vehicle and pedestrian area in the image; the larger the ITI value, the more traffic facilities there are, the more complex the traffic conditions in the area, and the lower the walking safety; conversely, the smaller the ITI value, the higher the walking safety;

[0038] The spatial comfort is evaluated by the planar visual index (PVI) and the longitudinal visual index (DVI), both of which are positively correlated with the street design space quality evaluation index;

[0039] The Planar Visual Index (PVI) is the ratio of the number of pixels occupied by trees and vegetation in all street view images to the area pixels of the entire image. The specific formula is as follows:

[0040]

[0041] Among them, P n A is the number of pixels occupied by trees and vegetation in the street view image numbered n, that is, the sum of the pixels of the i vegetation areas in the image; n It is the sum of all face pixels in the street view image. The plane visual index is positively correlated with comfort, that is, the higher the plane visual index, the higher the comfort.

[0042] The vertical visual index (DVI) is the ratio of the surface area pixels of the sky visible within the human eye's visual range to the surface area pixels of the entire street view image. The specific formula is as follows:

[0043]

[0044] Among them, DVI is the vertical sky visual degree of each street view image, v i is the number of pixels of the i-th sky area in the picture; a i is the sum of the pixels in the i-th face area in the street view image;

[0045] The space friendliness is evaluated by the crowd attraction index (CCI) and the commercial facilities satisfaction index (CSI), both of which are positively correlated with the street design space quality evaluation index;

[0046] The Crowd Attraction Index (CCI) is the ratio of the sum of the pixels of the crowd patches in all the street view images to the sum of the pixels of all elements in the entire image. The specific formula is as follows:

[0047]

[0048] Among them, P n is the sum of the pixels of all the crowd panels in n in the image, R n is the sum of the pixels of all elements in image n;

[0049] The commercial facility satisfaction index (CSI) is the sum of the pixels of commercial service facilities in all street view images divided by the pixel count of the entire image. The specific formula is as follows:

[0050]

[0051] Among them, C n A is the number of pixels occupied by commercial service facilities in the street view image numbered n, that is, the sum of the pixels of the area of ​​i facilities in the image; n is the total number of pixels in the street view image, that is, the sum of all face pixels in the image.

[0052] More preferably, step 2 comprises the following steps:

[0053] Step 2.1: Based on the results of step 1.3, the streets to be classified are divided into four categories using a Gaussian mixture clustering model. The pedestrian traffic index (SFI), the road network density index (RDI), the service facility satisfaction index (PSI), the motor vehicle interference index (VII), the traffic sign index (ITI), the planar visual index (PVI), the longitudinal visual index (DVI), the crowd attraction index (CCI), and the commercial facility satisfaction index (CSI) are calculated for each category. The specific calculation method is as follows:

[0054] The algebraic symbols are defined as follows:

[0055] x j represents the jth observation data, j = 1, 2…∞;

[0056] k is the number of sub-Gaussian models in the mixture model, k = 1, 2...K;

[0057] α k is the probability that the observed data belongs to the kth submodel, α k ≥0, and

[0058] is the Gaussian distribution density function of the kth sub-model, The expanded form is the same as that of the single Gaussian model;

[0059] γ jk represents the probability that the j-th observation data belongs to the k-th sub-model;

[0060] The probability distribution of the Gaussian mixture model is:

[0061]

[0062] Among them, the parameters That is, the expectation, variance or covariance of each sub-model, and the probability of occurrence in the Gaussian mixture model;

[0063] First, the likelihood function:

[0064]

[0065] Among them, Z, L, and P are all random vectors;

[0066] After taking the logarithm of equation 2:

[0067]

[0068] Since the log function satisfies the concave function property, we can get:

[0069]

[0070] when When is a constant, the equality sign holds, so E-step:

[0071] Q i (z j )=p(z j |x i ;θ) Equation 5

[0072] Q i (z i ) represents the expected value of the j-th sample for the i-th Gaussian model, and the M-step is the value of the parameter θ when the expectation of the above function value is maximized on the E-step:

[0073] θ:=argmaxl(θ) Equation 6

[0074] The next step is to substitute GMM into the EM algorithm step:

[0075] 1) Contribution rate of the i-th sample to the j-th Gaussian distribution:

[0076] 2) Estimate μ and ∑ according to Q in the E-step, and the Gaussian distribution density function of the i-th sub-model is as follows:

[0077]

[0078]

[0079] For the probability of event j occurring, P(G j ) is calculated as follows:

[0080]

[0081] Step 2.2: Define the street type belongingness index. Based on the aggregated meaning of the index, construct the street belongingness index for the four street types, including transportation belongingness, landscape and leisure belongingness, life belongingness, and commercial belongingness.

[0082] The traffic attribution degree includes the motor vehicle interference index, traffic sign index, road network density and crowd attraction index. The specific formula is as follows:

[0083] Traffic belonging degree = VII + ITI + RAI - CCI Formula 9

[0084] The landscape leisure attribution degree includes the plane visual index, the vertical visual index and the crowd attraction index. The specific formula is as follows:

[0085] The landscape leisure belonging degree = PVI + DVI + CCI formula 10

[0086] The life belonging degree includes the walkability index, service facility satisfaction index and crowd attraction index. The specific formula is as follows:

[0087] The life belonging degree = SFI + PSI + CCI formula 11

[0088] The commercial attribution degree includes the commercial facilities satisfaction index and the service facilities satisfaction index, and the specific formula is as follows:

[0089] The commercial attribution degree = CSI + PSI (Formula 12).

[0090] More preferably, step 3 is as follows:

[0091] Calculate the degree of belonging of each type of street to different street types, and classify and define each type of street according to the size of the category index value, and take the one or two items with larger values ​​in the category index as the two items with the largest index mean.

[0092] Beneficial effects of the present invention:

[0093] The present invention implements street classification by quantitatively analyzing street types based on multi-source data such as street view image attributes and a Gaussian mixture clustering model algorithm.

[0094] The concept, specific structure and technical effects of the present invention will be further described below in conjunction with the accompanying drawings to fully understand the purpose, characteristics and effects of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Figure 1 FIG. 1 shows an execution flow chart of an embodiment of the present invention.

[0096] Figure 2 A representative cross section showing excellent and poor street quality ratings for a first major street type according to one embodiment of the present invention.

[0097] Figure 3 A representative cross section showing excellent and poor street quality ratings for the second major street type in accordance with an embodiment of the present invention.

[0098] Figure 4 A representative cross section showing excellent and poor street quality ratings for the third major street type in accordance with an embodiment of the present invention.

[0099] Figure 5 A representative cross section showing excellent and poor street quality grades for the first major street type in one embodiment of the present invention. DETAILED DESCRIPTION

[0100] Example 1

[0101] like Figures 1 to 5 As shown in FIG, the street type classification method based on the Gaussian clustering mixture model includes the following steps:

[0102] Step 1: Collection and intelligent identification of street design attribute data;

[0103] Step 2: Determine the street type attribution index set;

[0104] Step 3: Construct a street type classification algorithm model.

[0105] In practical applications, the present invention firstly uses a fully convolutional complex neural network and a deep learning method to perform computer intelligent recognition and semantic segmentation on street view image data using an intelligent recognition program; secondly, the street space quality measurement system and its elements are summarized to determine nine types of quality indicators; thirdly, based on the nine types of indicators for evaluating urban street design quality, a Gaussian mixture clustering model is used to divide urban streets into four categories, and constructs street belongingness indicators to four types of streets, specifically including traffic belongingness, landscape and leisure belongingness, life belongingness, and commercial belongingness; finally, each type of street is classified and defined according to its belongingness to different types of streets.

[0106] In some embodiments, step 1 includes the following steps:

[0107] Step 1.1: Obtain the spatial morphological data of the street, as follows:

[0108] Use a pan-tilt camera to take pictures of city streets, and then correct the pictures of city streets through field surveys to build a street space morphology database;

[0109] The database elements of the street space morphology database include land use nature, land boundary and land area;

[0110] Step 1.2: Get the street view image data of the street, as follows:

[0111] Then, based on a fully convolutional complex neural network and deep learning methods, an intelligent recognition program was used to perform computer intelligent recognition of lane elements, sidewalk elements, and motor vehicle elements in the acquired street view image data. Statistical analysis was performed on the pixel data occupied by each element to obtain the number and proportion of pixels of each type of element.

[0112] In actual applications, the number and proportion of pixels of each type of feature in street view images are as follows:

[0113]

[0114] Step 1.3: Determine the street design space quality evaluation indicators and screen them from the objective evaluation dimension and subjective evaluation dimension respectively;

[0115] Among them, objective evaluation dimensions include travel feasibility, road network accessibility and facility convenience;

[0116] The pedestrian feasibility is evaluated by the Walkability Index (SFI), which is the ratio of the number of pixels of pedestrian paths to vehicle paths in all street streetview images. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows:

[0117]

[0118] Among them, w n is the total number of pixels occupied by walking space in the street view image numbered n; R n is the number of pixels occupied by the total vehicle space in the street view image;

[0119] Road network accessibility is evaluated using the road network density index (RDI), which is the ratio of the number of pixels occupied by the roadway to the total pixel count in all street view images. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows:

[0120]

[0121] Among them, R n A is the number of pixels occupied by the total vehicle space in the street view image, n is the total number of pixels in the street view image, that is, the sum of all pixels in the face area of ​​the image;

[0122] The convenience of facilities is evaluated by the service facilities satisfaction index (PSI), which is the ratio of the number of pixels occupied by service facilities in all street street view images to the total number of pixels in the image. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows:

[0123]

[0124] Among them, p n A is the number of pixels occupied by service facilities in the street view image numbered n, that is, the sum of the pixels of the area of ​​i facilities in the image; n is the total number of pixels in the street view image, that is, the sum of all pixels in the face area of ​​the image;

[0125] Subjective evaluation dimensions include walking safety, spatial comfort, and spatial friendliness;

[0126] Pedestrian safety was evaluated using the Motor Vehicle Interference Index VII and Traffic Sign Index ITI, both of which were negatively correlated with the street design space quality evaluation index;

[0127] The motor vehicle interference index VII is the sum of the number of pixels of motor vehicle patches in all street view images and the total number of pixels of motor vehicle lanes in the entire image. The specific formula is as follows:

[0128]

[0129] Among them, C n represents the pixels of the motor vehicle patches identified in the street view image, R n The total number of pixels in the motor vehicle lane in the entire image. The higher the vehicle interference index, the greater the proportion of motor vehicles in the street space, and the lower the sense of security.

[0130] The Traffic Sign Index (ITI) is the ratio of the number of pixels of traffic lights and signs in the entire street view image to the total number of pixels in the image. The specific formula is as follows:

[0131]

[0132] Among them, T n is the pixel area of ​​traffic lights and traffic signs in the street view image numbered n, that is, the sum of the area pixels of i doors and windows in the image; R n is the total number of pixels in the street space in the image, that is, the sum of the pixels in the i-th vehicle and pedestrian area in the image; the larger the ITI value, the more traffic facilities there are, the more complex the traffic conditions in the area, and the lower the walking safety; conversely, the smaller the ITI value, the higher the walking safety;

[0133] Spatial comfort is evaluated by the planar visual index (PVI) and the longitudinal visual index (DVI), both of which are positively correlated with the street design space quality evaluation index;

[0134] The Planar Visual Index (PVI) is the ratio of the number of pixels occupied by trees and vegetation in all street view images to the total number of pixels in the entire image. The specific formula is as follows:

[0135]

[0136] Among them, P n A is the number of pixels occupied by trees and vegetation in the street view image numbered n, that is, the sum of the pixels of the i vegetation areas in the image; n It is the sum of all face pixels in the street view image. The plane visual index is positively correlated with comfort, that is, the higher the plane visual index, the higher the comfort.

[0137] The vertical visual index (DVI) is the ratio of the surface area pixels that are visible to the sky within the human eye's visual range to the surface area pixels of the entire image in all street view images. The specific formula is as follows:

[0138]

[0139] Among them, DVI is the vertical sky visual degree of each street view image, v i is the number of pixels of the i-th sky area in the picture; a i is the sum of the pixels in the i-th face area in the street view image;

[0140] Spatial friendliness is evaluated through the crowd attraction index (CCI) and the commercial facilities satisfaction index (CSI), both of which are positively correlated with the street design space quality evaluation index;

[0141] The Crowd Attraction Index (CCI) is the ratio of the sum of the pixels of crowd patches in all street view images to the sum of the pixels of all elements in the entire image. The specific formula is as follows:

[0142]

[0143] Among them, P n is the sum of the pixels of all the crowd panels in n in the image, R n is the sum of the pixels of all elements in image n;

[0144] The Commercial Facilities Satisfaction Index (CSI) is the sum of the pixels of commercial service facilities in all street view images divided by the total number of pixels in the entire image. The specific formula is as follows:

[0145]

[0146] Among them, C n A is the number of pixels occupied by commercial service facilities in the street view image numbered n, that is, the sum of the pixels of the area of ​​i facilities in the image; n is the total number of pixels in the street view image, that is, the sum of all face pixels in the image.

[0147] In practical applications, the evaluation indicators of objective material space dimensions are quantified as follows:

[0148] The material composition and environmental characteristics of street space serve as a vehicle for people and their activities. The objective physical space of a street is characterized by three dimensions: walkability, road network accessibility, and facility convenience. These are quantified as the pedestrian accessibility index, road network density, and service facility satisfaction index.

[0149] (1) Walking feasibility

[0150] Spatial feasibility index (SFI)

[0151] In the construction of modern urban road networks, pedestrian space often serves as an adjunct to motor vehicle traffic, occupying a relatively small proportion. Excessively narrow sidewalks have become a factor affecting street space quality, presenting a significant challenge for planners and designers. Therefore, this paper uses the ratio of pedestrian to roadway areas in streetscape images as the spatial feasibility index (SFI) of street pavement space to reflect the walkability of street surface space. A higher SFI indicates greater capacity for accommodating people and activities.

[0152]

[0153] w nis the total number of pixels occupied by walking space in the street view image numbered n; R n The number of pixels occupied by the total vehicle traffic space in the street view image. A larger SFI value indicates a larger walkable area; conversely, a smaller SFI value indicates a smaller walkable area.

[0154] (2) Road network accessibility

[0155] Road-network density RDI (Road-network density)

[0156] One of the essential functions of streets is to enable pedestrians to travel from one location to another. This includes connecting locations within the street as well as areas beyond. Street network accessibility, or rather, street network density, measures the ease with which pedestrians can use the street space to reach their destinations. The higher the street network density, the more convenient it is for pedestrians to travel from surrounding areas to the local area for activities and social interactions. This paper uses the ratio of the number of pixels occupied by lanes in street view images to the total image size as the road-network density (RDI) of the street surface space.

[0157]

[0158] R n A is the number of pixels occupied by the total vehicle space in the street view image, n is the total number of pixels in the street view image, that is, the sum of all face pixels in the image.

[0159] (3) Facility convenience

[0160] Public-facility satisfied index (PSI)

[0161] Street space service facilities serve as the foundation for engaging in various activities, and the convenience of these facilities significantly impacts the use and evaluation of street space. Most of people's ongoing social activities occur near walls, windows, and other sidewalk amenities, including vending machines, gardens, and benches. These street facilities not only fulfill their specific functions and facilitate their use, but also enhance the richness of street space. Generally speaking, street space amenities include municipal facilities, road traffic facilities, artistic landscapes, and event facilities. This paper uses the Public Facility Satisfaction Index (PSI) to reflect the convenience of facilities. This index quantifies the ratio of pixels occupied by service facilities to the total pixel count in a street view image.

[0162]

[0163] p n A is the number of pixels occupied by service facilities in the street view image numbered n, that is, the sum of the pixels of the area of ​​i facilities in the image; n is the total number of pixels in the street view image, that is, the sum of all face pixels in the image.

[0164] 3.3.2.2 Quantification of evaluation indicators for subjective psychological feelings

[0165] (1) Walking safety

[0166] For street spaces, various factors within the street environment influence both physical and perceived safety, and the sense of safety also influences how people use their surroundings. The two are interrelated and mutually influential. When pedestrians walk through street spaces, their safety must be ensured. This article examines pedestrian safety primarily through the Motor Vehicle Interference Index and the Traffic Sign Index. The Motor Vehicle Interference Index is a negative indicator; a higher value indicates a lower sense of safety. The Traffic Sign Index is a positive indicator; a higher value indicates a higher sense of safety.

[0167] ①Vehicle interference index VII

[0168] Numerous studies domestically and internationally on the relationship between environmental factors and safety have shown a correlation between parking space and perceptions of safety. While various regions are creating walkable and readable street spaces, dedicating more space to pedestrians, some streets still retain significant space occupied by motor vehicles, creating safety hazards for pedestrians. The large number of vehicles and parking spaces reduces users' sense of security, making street spaces less safe and accessible. This article uses the number of pixels in the motor vehicle patches and the total number of pixels in the motor vehicle lanes in the entire image as a quantitative indicator of the Vehicle Interference Index (VII). This index is negative; a higher VII indicates a greater proportion of motor vehicles in the street space, leading to a lower sense of safety.

[0169]

[0170] C n represents the pixels of the motor vehicle patches identified in the street view image, R n The total number of pixels in the motor vehicle lane in the entire image. A higher vehicle interference index indicates a greater proportion of motor vehicles in the street space, leading to a lower sense of security.

[0171] ②Interface transparency index (ITI)

[0172] Traffic lights and traffic signs are crucial ancillary facilities in modern transportation systems, serving to strengthen urban traffic management, facilitate transportation, and maintain traffic safety. Generally speaking, areas with more traffic lights and signs experience more complex traffic conditions, indicating greater traffic disruption and a sense of psychological insecurity. Therefore, this paper uses the Interface Transparency Index (ITI), which measures the presence of traffic lights and signs in street spaces, as another indicator of street safety.

[0173]

[0174] T n is the pixel area of ​​traffic lights and traffic signs in the street view image numbered n, that is, the sum of the area pixels of i doors and windows in the image; R n The total number of pixels in the street space in the image is the sum of the pixels in the i-th lane and pedestrian area in the image. A larger ITI value indicates more traffic facilities, more complex traffic conditions in the area, and lower pedestrian safety. Conversely, a smaller ITI value indicates higher pedestrian safety.

[0175] (2) Spatial comfort

[0176] People's basic needs for comfort, including those for the natural environment, take precedence over higher-level needs such as a sense of belonging, cognition, and aesthetics. Among the many factors influencing spatial comfort, horizontal visual reach and the openness of a space have the most direct and pervasive impact on the comfort of the walking experience. Therefore, this paper selects the horizontal and vertical visual indices of streets as evaluation indicators of comfort.

[0177] ① Plane visual index PVI (Plane visual index)

[0178] The most common source of comfort for the human eye's two-dimensional vision comes from abundant green vegetation. Trees and vegetation provide shade in the summer and shelter from the wind and rain on rainy days, enhancing a comfortable environment. Green is a fundamental color of nature, evoking a sense of peace and tranquility. This paper uses the "green view ratio" measurement method to extract the ratio of the number of pixels occupied by trees and vegetation to the total pixel area of ​​the image as the Plane Visual Index (PVI).

[0179]

[0180] P n A is the number of pixels occupied by trees and vegetation in the street view image numbered n, that is, the sum of the pixels of the i vegetation areas in the image; nis the sum of all face pixels in the street view image. The plane vision index is positively correlated with comfort, that is, the higher the plane vision index, the higher the comfort.

[0181] ②Diagonal visibility index (DVI)

[0182] The expansive view of the sky, offered by distant views, can effectively alleviate the stress of work and life brought on by the fast-paced, high-intensity of modern urban life. The expansive view of the sky fully extends the human eye's field of vision, while the sky's color wavelengths can help relax the eye's adjustment, relieving eye soreness and fatigue, and providing excellent protection and repair for vision and the nervous system. Therefore, this article uses the ratio of the number of pixels visible to the sky within the human visual range (approximately 120 degrees) to the total number of pixels in the image as the diagonal visibility index (DVI).

[0183]

[0184] DVI is the vertical sky visual degree of each street view image, v i is the number of pixels of the i-th sky area in the picture; a i is the sum of pixels in the i-th face region in the street view image.

[0185] (3) Facility friendliness

[0186] As a vital part of public space, streets are primarily used for socializing. This use imbues them with unique characteristics. Streets allow people to fully integrate into the urban environment, connecting with shops, residences, and the natural environment, and engaging with others. Social interaction within street spaces can be categorized into two main types: interaction between pedestrians and facility users, and interaction between pedestrians and users of adjacent buildings. Based on this, this article evaluates the facility-friendliness of street spaces from two perspectives: crowd aggregation and interactivity within street-facing interfaces.

[0187] ① Crowd concentration index (CCI)

[0188] A good street provides a high-quality environment, which inevitably attracts more people to engage in healthy social activities. Therefore, the degree of crowd attraction is, to a certain extent, an important reflection of the quality of the street space. This paper uses the ratio of the sum of the pixels of the crowd patches in the street view image to the sum of the pixels of all elements in the entire image as the crowd concentration index (CCI).

[0189]

[0190] P n is the sum of the pixels of all the crowd panels in n in the image, R n is the sum of the pixels of all elements in image n.

[0191] ② Commercial-facility satisfied index (CSI)

[0192] Street interfaces are the intersection of public and private spheres, and the functions of street-facing buildings significantly influence the behavior of street users. Commercial services offer diverse services, attracting more people and stimulating social interaction. Street-level commercial facilities serve as destinations, attracting people and fostering various social activities, such as shopping and window-shopping. To attract more people, commercial services incorporate spatial details into their building interfaces and front areas. For example, transparent display windows allow pedestrians to access more information, while rest areas in front of stores provide a place for pedestrians to rest. Therefore, this paper uses the Commercial Facility Satisfaction Index (CSI), which is the sum of the pixels of commercial service facilities in street view images and the ratio of the pixels in the entire image, as another evaluation dimension of facility friendliness.

[0193]

[0194] C n A is the number of pixels occupied by commercial service facilities in the street view image numbered n, that is, the sum of the pixels of the area of ​​i facilities in the image; n is the total number of pixels in the street view image, that is, the sum of all face pixels in the image.

[0195] The specific evaluation indicators for urban street design quality are shown in the following table:

[0196]

[0197]

[0198] In some embodiments, step 2 includes the following steps:

[0199] Step 2.1: Based on the results of step 1.3, use the Gaussian mixture clustering model to divide the streets to be classified into four categories. Calculate the pedestrian access index (SFI), road network density index (RDI), service facility satisfaction index (PSI), vehicle interference index (VII), traffic sign index (ITI), horizontal visual index (PVI), vertical visual index (DVI), crowd attraction index (CCI), and commercial facility satisfaction index (CSI) for each category. The specific calculation method is as follows:

[0200] The algebraic symbols are defined as follows:

[0201] x j represents the jth observation data, j = 1, 2…∞;

[0202] k is the number of sub-Gaussian models in the mixture model, k = 1, 2...K;

[0203] α k is the probability that the observed data belongs to the kth submodel, α k ≥0, and

[0204] is the Gaussian distribution density function of the kth sub-model, The expanded form is the same as that of the single Gaussian model;

[0205] γ jk represents the probability that the j-th observation data belongs to the k-th sub-model;

[0206] The probability distribution of the Gaussian mixture model is:

[0207]

[0208] Among them, the parameters That is, the expectation, variance or covariance of each sub-model, and the probability of occurrence in the Gaussian mixture model;

[0209] First, the likelihood function:

[0210]

[0211] Among them, Z, L, and P are all random vectors;

[0212] After taking the logarithm of equation 2:

[0213]

[0214] Since the log function satisfies the concave function property, we can get:

[0215]

[0216] when When is a constant, the equality sign holds, so E-step:

[0217] Q i (z j )=p(z j |x i ;θ) Equation 5

[0218] Q i (z i ) represents the expected value of the j-th sample for the i-th Gaussian model, and the M-step is the value of the parameter θ when the expectation of the above function value is maximized on the E-step:

[0219] θ:=argmaxl(θ) Equation 6

[0220] The next step is to substitute GMM into the EM algorithm step:

[0221] 1) Contribution rate of the i-th sample to the j-th Gaussian distribution:

[0222] 2) Estimate μ and ∑ according to Q in the E-step, and the Gaussian distribution density function of the i-th sub-model is as follows:

[0223]

[0224]

[0225] For the probability of event j occurring, P(G j ) is calculated as follows:

[0226]

[0227] In practical applications, a Gaussian mixture model is a linear combination of multiple Gaussian distribution functions. It is a parameterized probability density function and the weighted sum of a set of Gaussian density functions. The Gaussian mixture model can be estimated using the iterative EM algorithm or the maximum a posteriori probability method. It is often used to address situations where the same data set contains multiple distinct distributions. The Gaussian mixture model is used here because the Gaussian distribution has excellent mathematical properties and computational performance.

[0228] EM is essentially Maximum Likelihood Estimation (MLE). Probability models sometimes contain both observed and latent variables. If the variables in the probability model are all observed variables, as long as the measured data are available, the model parameters can be estimated directly using Maximum Likelihood Estimation or Bayesian Estimation. However, when the model contains latent variables, these estimation methods cannot be simply used. The EM algorithm is a Maximum Likelihood Estimation method for the parameters of probability models containing latent variables.

[0229] The mean values ​​of street scene recognition indicators in cluster analysis are shown in the following table:

[0230]

[0231] Step 2.2: Define the street type belongingness index. Based on the aggregated meaning of the index, construct the street belongingness index for the four street types, including transportation belongingness, landscape and leisure belongingness, life belongingness, and commercial belongingness.

[0232] The traffic attribution index includes the motor vehicle interference index, traffic sign index, road network density, and crowd attraction index. The specific formula is as follows:

[0233] Traffic belonging degree = VII + ITI + RAI - CCI Formula 9

[0234] The landscape leisure attribution degree includes the plane visual index, vertical visual index and crowd attraction index. The specific formula is as follows:

[0235] Landscape leisure ownership degree = PVI + DVI + CCI Formula 10

[0236] The degree of belonging to life includes the walkability index, service facility satisfaction index and crowd attraction index. The specific formula is as follows:

[0237] Life belongingness = SFI + PSI + CCI Formula 11

[0238] The commercial belonging degree includes the commercial facilities satisfaction index and the service facilities satisfaction index. The specific formula is as follows:

[0239] Commercial ownership = CSI + PSI Formula 12.

[0240] The details are shown in the following table:

[0241]

[0242]

[0243] As can be seen from the table above, categories 0 and 3 both have the characteristics of transportation-oriented streets, category 1 has the characteristics of a landscape and leisure street, and category 2 has the characteristics of a life service street. Notably, category 3 has the characteristics of both transportation and commercial streets. Based on this quantitative street type classification method, the street type of each street section can be quantitatively defined.

[0244] In some embodiments, step 3 is as follows:

[0245] Calculate the degree of belonging of each type of street to different street types, and classify and define each type of street according to the size of the category index value, and take the one or two items with larger values ​​in the category index as the two items with the largest index mean.

[0246] Example 2

[0247] Based on the exploration of street types and design elements in a certain city, the design quality of different types of streets was evaluated by combining the identification results of street elements. Figures 2 to 4 The Street Design Quality Matrix contains representative cross-sections of excellent and poor street quality ratings for four major street types.

[0248] By comparing representative sections of excellent and poor street quality grades for each major street type, we obtain design factors that have a significant impact on street quality. Then, combining quantitative and qualitative analysis, we provide experience for improving street design quality and planning practice.

[0249] The following table shows the identification of excellent design elements of living streets:

[0250]

[0251]

[0252] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A street type classification method based on a Gaussian clustering mixture model; characterized in that: The steps include: Step 1: Collection and intelligent identification of street design attribute data; Step 2: Determine the street type attribution index set; Step 2 includes the following steps: Step 2.1: Use the Gaussian mixture clustering model to divide the streets to be classified into four categories. Calculate the pedestrian access index (SFI), road network density index (RDI), service facility satisfaction index (PSI), vehicle interference index (VII), traffic sign index (ITI), horizontal visual index (PVI), vertical visual index (DVI), crowd attraction index (CCI), and commercial facility satisfaction index (CSI) for each category. The specific calculation method is as follows: The algebraic symbols are defined as follows: x j represents the jth observation data, j = 1, 2…∞; k is the number of sub-Gaussian models in the mixture model, k = 1, 2...K; α k is the probability that the observed data belongs to the kth submodel, α k ≥0, and is the Gaussian distribution density function of the kth sub-model, The expanded form is the same as that of the single Gaussian model; γ jk represents the probability that the j-th observation data belongs to the k-th sub-model; The probability distribution of the Gaussian mixture model is: Among them, the parameters That is, the expectation, variance or covariance of each sub-model, and the probability of occurrence in the Gaussian mixture model; First, the likelihood function: Among them, Z, L, and P are all random vectors; After taking the logarithm of equation 2: Since the log function satisfies the concave function property, we get: when When is a constant, the equality sign holds, so E-step: Q i (z j ) = p(z j |x i ; θ) Equation 5 Q i (z i ) represents the expected value of the j-th sample for the i-th Gaussian model, and the M-step is the value of the parameter θ when the expectation of the above function value is maximized on the E-step: θ:=argmaxl(θ) Equation 6 The next step is to substitute GMM into the EM algorithm step: 1) Contribution rate of the i-th sample to the j-th Gaussian distribution: 2) Estimate μ and ∑ according to Q in the E-step, and the Gaussian distribution density function of the i-th sub-model is as follows: For the probability of event j occurring, P(G j ) is calculated as follows: Step 2.2: Define the street type belongingness index. Based on the aggregated meaning of the index, construct the street belongingness index for the four street types, including transportation belongingness, landscape and leisure belongingness, life belongingness, and commercial belongingness. Step 3: Construct a street type classification algorithm model.

2. The street type classification method based on the Gaussian clustering mixture model according to claim 1 is characterized in that: Step 1 includes the following steps: Step 1.1: Obtain the spatial morphological data of the street, as follows: Use a pan-tilt camera to take pictures of city streets, and then correct the pictures of city streets through field surveys to build a street space morphology database; The database elements of the street space morphology database include land use nature, land use boundary and land use area; Step 1.2: Obtain street view image data of the street, as follows: Then, based on a fully convolutional complex neural network and deep learning methods, an intelligent recognition program is used to perform computer intelligent recognition of the lane elements, sidewalk elements, and motor vehicle elements in the acquired street view image data, and statistical analysis is performed on the pixel data occupied by each element to obtain the number and proportion of pixels of each type of element; Step 1.3: Determine the street design space quality evaluation indicators and screen them from the objective evaluation dimension and subjective evaluation dimension respectively; Among them, the objective evaluation dimensions include travel feasibility, road network accessibility and facility convenience; The pedestrian feasibility is evaluated by the pedestrian access index (SFI), which is the ratio of the number of pixels of pedestrian paths and vehicle paths in all street scene images. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows: Among them, w n is the total number of pixels occupied by walking space in the street view image numbered n; R n is the number of pixels occupied by the total vehicle space in the street view image; The road network accessibility is evaluated by the road network density index (RDI), which is the ratio of the number of pixels occupied by the roadway in the street view image to the total number of pixels in the image. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows: Among them, R n A is the number of pixels occupied by the total vehicle space in the street view image, n is the total number of pixels in the street view image, that is, the sum of all pixels in the face area of ​​the image; The convenience of the facilities is evaluated by the service facilities satisfaction index (PSI), which is the ratio of the number of pixels occupied by the service facilities in all street street view images to the total number of pixels in the image. It is positively correlated with the street design space quality evaluation index. The specific formula is as follows: Among them, p n A is the number of pixels occupied by service facilities in the street view image numbered n, that is, the sum of the pixels of the area of ​​i facilities in the image; n is the total number of pixels in the street view image, that is, the sum of all pixels in the face area of ​​the image; The subjective evaluation dimensions include walking safety, space comfort and space friendliness; The pedestrian safety is evaluated by the vehicle interference index VII and the traffic sign index ITI, both of which are negatively correlated with the street design space quality evaluation index; The motor vehicle interference index VII is the sum of the number of pixels of motor vehicle patches in all the street scene images and the number of pixels of motor vehicle lanes in the entire image. The specific formula is as follows: Among them, C n represents the pixels of the motor vehicle patches identified in the street view image, R n The total number of pixels in the motor vehicle lane in the entire image. The higher the vehicle interference index, the greater the proportion of motor vehicles in the street space, and the lower the sense of security. The traffic sign index (ITI) is the ratio of the number of pixels of traffic lights and signs in the street view image to the total number of pixels in the image. The specific formula is as follows: Among them, T n is the pixel area of ​​traffic lights and traffic signs in the street view image numbered n, that is, the sum of the area pixels of i doors and windows in the image; R n The total number of pixels in the street space in the image, that is, the sum of the pixels in the i-th vehicle and pedestrian area in the image; the larger the ITI value, the more traffic facilities there are, the more complex the traffic conditions in the area, and the lower the walking safety; conversely, the smaller the ITI value, the higher the walking safety; The spatial comfort is evaluated by the planar visual index (PVI) and the longitudinal visual index (DVI), both of which are positively correlated with the street design space quality evaluation index; The Planar Visual Index (PVI) is the ratio of the number of pixels occupied by trees and vegetation in all street view images to the area pixels of the entire image. The specific formula is as follows: Among them, P n A is the number of pixels occupied by trees and vegetation in the street view image numbered n, that is, the sum of the pixels of the i vegetation areas in the image; n It is the sum of all face pixels in the street view image. The plane visual index is positively correlated with comfort, that is, the higher the plane visual index, the higher the comfort. The vertical visual index (DVI) is the ratio of the surface area pixels of the sky visible within the human eye's visual range to the surface area pixels of the entire street view image. The specific formula is as follows: Among them, DVI is the vertical sky visual degree of each street view image, v i is the number of pixels of the i-th sky area in the picture; a i is the sum of the pixels in the i-th face area in the street view image; The space friendliness is evaluated by the crowd attraction index (CCI) and the commercial facilities satisfaction index (CSI), both of which are positively correlated with the street design space quality evaluation index; The Crowd Attraction Index (CCI) is the ratio of the sum of the pixels of the crowd patches in all the street view images to the sum of the pixels of all elements in the entire image. The specific formula is as follows: Among them, P n is the sum of the pixels of all the crowd panels in n in the image, R n is the sum of the pixels of all elements in image n; The commercial facility satisfaction index (CSI) is the sum of the pixels of commercial service facilities in all street view images divided by the pixel count of the entire image. The specific formula is as follows: Among them, C n A is the number of pixels occupied by commercial service facilities in the street view image numbered n, that is, the sum of the pixels of the area of ​​i facilities in the image; n is the total number of pixels in the street view image, that is, the sum of all face pixels in the image.

3. The street type classification method based on Gaussian clustering mixture model according to claim 2 is characterized in that: The traffic attribution degree includes the motor vehicle interference index, traffic sign index, road network density and crowd attraction index. The specific formula is as follows: Traffic belonging degree = VII + ITI + RAI - CCI Formula 9 The landscape leisure attribution degree includes the plane visual index, the vertical visual index and the crowd attraction index. The specific formula is as follows: The landscape leisure belonging degree = PVI + DVI + CCI formula 10 The life belonging degree includes the walkability index, service facility satisfaction index and crowd attraction index. The specific formula is as follows: The life belonging degree = SFI + PSI + CCI formula 11 The commercial attribution degree includes the commercial facilities satisfaction index and the service facilities satisfaction index, and the specific formula is as follows: The commercial attribution degree = CSI + PSI (Formula 12).

4. The street type classification method based on the Gaussian clustering mixture model according to claim 3 is characterized in that: Step 3 is as follows: Calculate the degree of belonging of each type of street to different street types, and classify and define each type of street according to the size of the category index value, and take the one or two items with larger values ​​in the category index as the two items with the largest index mean.

Images