Urban low-efficiency space identification and classification method and system

Abstract

The invention discloses an urban low-efficiency space identification and classification method and system, and the method comprises the steps: firstly obtaining the people flow space distribution data of a target space, building a square grid in the target space, and screening out a primarily selected urban low-efficiency space according to the people flow in the grid; secondly, acquiring live-action image data of the primarily selected urban low-efficiency space, identifying functions of the urban low-efficiency space, including an activity facility square grid, a greenbelt square grid and a pavement square grid, and obtaining the urban low-efficiency space according to a space area formed by adjacent connection of the grids; and finally, performing shape classification and function classification on the urban low-efficiency space. According to the method, accurate identification of the urban low-efficiency space and automatic classification in the plane form dimension and the space function dimension are realized, subjective judgment of people in urban low-efficiency space demarcation is avoided, and the reliability and usability of identification and classification results are improved.

Description

Technical Field

[0001] The present invention relates to the field of urban planning, and in particular to a method and system for identifying and classifying urban inefficient spaces. Background Art

[0002] Urban space is one of the important research contents of urban and rural planning. Good urban space quality can not only improve the comfort of urban residents' activities, but also enhance the vitality of urban space. However, with the growth and prosperity of urban economy, a series of urban space problems have also emerged, including the emergence of a large number of inefficient urban spaces. Urban inefficient space refers to the space with low vitality and imperfect layout of greening facilities in urban space. The number and distribution of urban inefficient space are related to the quality of urban space, which makes the accurate identification and classification of urban inefficient space particularly important. The high-precision identification and classification technology of urban inefficient space can collect detailed urban space information, accurately identify the vitality depression of urban space, and accurately present the distribution of urban inefficient space of various shapes and functions in urban space, which helps planners to accurately identify urban inefficient space and formulate targeted transformation plans to improve the quality of urban space environment.

[0003] At present, the common identification and classification methods of urban inefficient spaces are: one is traditional field research, which first observes and counts the flow of people in urban spaces at different times to obtain the distribution of urban space flow at different times; secondly, by taking on-site photos, the urban inefficient spaces are classified, and finally combined with the two-dimensional urban space morphological data to form the classification and identification of urban inefficient spaces. This method has problems such as large workload, large data errors, and the classification standard relies on subjective judgment, and it is impossible to accurately classify and identify urban inefficient spaces;. Another method is to classify and identify urban inefficient spaces by combining mobile phone signaling data with two-dimensional urban space data. This method has data timeliness errors and cannot accurately restore the three-dimensional morphology of urban inefficient spaces. The results lack scientificity and accuracy. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a method and system for identifying and classifying urban inefficient spaces, so as to achieve accurate identification of urban inefficient spaces and automatic classification of urban inefficient spaces in terms of plane form dimension and spatial function dimension.

[0005] Technical solution: The method for identifying and classifying urban inefficient spaces of the present invention comprises the following steps:

[0006] Collecting the spatial distribution data of the crowd flow in the target space, where the boundary of the target space is the land boundary between the park green space and the square land in the urban area;

[0007] A square grid is established in the target space, and the spatial distribution data of the flow of people is divided into each grid. The space corresponding to the grid whose flow of people is ranked lower than the first threshold is the inefficient space of the preliminary selected city.

[0008] Collect real-life image data of the inefficient spaces in the preliminary selected cities, and identify the functions of the inefficient spaces in the preliminary selected cities through image segmentation methods and feature point detection and description methods, including square grids of activity facilities, square grids of green spaces, and square grids of pavement;

[0009] The space corresponding to the grid where the square grids of activity facilities, green space and pavement can be connected to each other and the space area exceeds the second threshold is the urban inefficient space;

[0010] The square grids of the urban inefficient spaces are merged according to the adjacency relationship to obtain a composite urban inefficient space, and the urban inefficient spaces are classified according to the shapes of the composite urban inefficient spaces to obtain a shape classification result;

[0011] The inefficient urban spaces are classified according to their functions and the functional classification results are obtained.

[0012] Furthermore, the square grids of the urban inefficient space are merged according to the adjacency relationship to obtain a composite urban inefficient space, and the urban inefficient space is classified according to the shape of the composite urban inefficient space, and the shape classification results include:

[0013] According to the area, rectangularity and aspect ratio of the minimum circumscribed rectangle of the composite urban inefficient space, the urban inefficient space is divided into three categories: scattered urban inefficient space, linear urban inefficient space and surface urban inefficient space.

[0014] Furthermore, based on the area of ​​the urban inefficient space being less than the third threshold, the scattered urban inefficient space is screened out; based on the rectangularity of the urban inefficient space being less than the fourth threshold, the curved urban inefficient space is screened out; based on the aspect ratio of the minimum circumscribed rectangle of the urban inefficient space being not less than the fifth threshold, the straight urban inefficient space is screened out; the remaining urban inefficient space is surface urban inefficient space; both the curved urban inefficient space and the straight urban inefficient space belong to linear urban inefficient space.

[0015] Further, the urban inefficient spaces are classified according to their functions, and the functional classification results include: clustering the functions of the urban inefficient spaces using an agglomerative hierarchical clustering algorithm, and the functional classification results include space-poor urban inefficient spaces, greening-deficient urban inefficient spaces, facility-deficient urban inefficient spaces, and functional use-inefficient urban inefficient spaces;

[0016] Among them, the categories with low proportion of greening square grids and low proportion of activity facilities square grids are defined as spatially poor urban inefficient spaces, the categories with low proportion of greening square grids and high proportion of activity facilities square grids are defined as greening-deficient urban inefficient spaces, the categories with high proportion of greening square grids and low proportion of activity facilities square grids are defined as facility-deficient urban inefficient spaces, and the categories with high proportion of greening square grids and high proportion of activity implementation square grids are defined as functionally inefficient urban inefficient spaces.

[0017] Furthermore, real-life image data of the preliminarily selected urban inefficient spaces are collected, and the functions of the preliminarily selected urban inefficient spaces are identified through image segmentation methods and feature point detection and description methods, including: taking facilities, green spaces, and paved objects as foregrounds, and using the GrabCut algorithm to divide the preliminarily selected urban inefficient spaces into facility square grids, green space square grids, and paved square grids; extracting key feature points of each type of facility through feature point detection and description methods, and dividing the facility square grids into municipal facility square grids, transportation facility square grids, and activity facility square grids.

[0018] Furthermore, the method also includes displaying the shape classification results and function classification results of the urban inefficient space in an urban sandbox.

[0019] Furthermore, collecting the spatial distribution data of the flow of people in the target space includes: using the millimeter wave radar sensor to collect the spatial distribution data of the flow of people every day and every hour within a week, and the week is seven consecutive days excluding special holidays and extreme weather days.

[0020] The urban inefficient space identification and classification system of the present invention comprises:

[0021] A crowd flow spatial distribution collection unit is used to collect crowd flow spatial distribution data of a target space, where the boundary of the target space is the boundary between a park green space and a square land in an urban area;

[0022] A preliminary city inefficient space division unit is used to establish a square grid in the target space, divide the spatial distribution data of the flow of people into each grid, and the space corresponding to the grid whose flow of people is ranked lower than the first threshold is the preliminary city inefficient space;

[0023] An urban inefficient space identification unit is used to collect real-scene image data in the preliminarily selected urban inefficient space, and identify the functions of the preliminarily selected urban inefficient space through an image segmentation method and a feature point detection and description method, including square grids of activity facilities, square grids of green space, and square grids of pavement; the grid corresponding space formed by the spatial area exceeding the second threshold formed by the square grids of activity facilities, green space, and pavement being adjacently connected to each other is the urban inefficient space;

[0024] The urban inefficient space classification unit is used to merge the square grids of the urban inefficient space according to the adjacency relationship to obtain a composite urban inefficient space, classify the urban inefficient space according to the shape of the composite urban inefficient space to obtain a shape classification result; and classify the urban inefficient space according to the function of the urban inefficient space to obtain a function classification result.

[0025] The electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the method for identifying and classifying inefficient urban spaces is implemented.

[0026] The computer-readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor, the method for identifying and classifying urban inefficient spaces is implemented.

[0027] Beneficial effects: Compared with the prior art, the advantages of the present invention are:

[0028] 1. The present invention uses the GrabCut algorithm, feature point detection and description based on the spatial distribution data of human flow, real-life images, to perform multiple rounds of screening and identification of urban inefficient spaces, thereby excluding municipal administration, transportation facilities and small-area land, and achieving accurate identification of urban inefficient spaces. It combines human behavior patterns with the material space environment, avoids human subjective judgment in the delineation of urban inefficient spaces, improves the scientific nature of the delineation of urban inefficient spaces, and improves the work efficiency of urban planners.

[0029] 2. The present invention classifies the shape characteristics and functional characteristics of urban inefficient spaces through random forest algorithm and agglomerative hierarchical clustering algorithm, realizes the automatic classification of urban inefficient spaces in plane morphological dimension and spatial functional dimension, avoids the problem of arbitrariness in classification by planners, improves the reliability and availability of recognition results, and provides a more scientific basis for further analyzing the causes of the formation of urban inefficient spaces and optimizing and transforming urban inefficient spaces.

[0030] 3. The present invention displays the shape and real-life images of urban inefficient spaces through a holographic sand table, thereby improving the intuitiveness of the display of identification results, enabling planners to clearly observe the spatial distribution characteristics, shape characteristics, spatial constituent elements, etc. of various types of urban inefficient spaces, and combining with wearable three-dimensional motion capture equipment to improve the accuracy and efficiency of feedback information recording.

[0031] 4. The present invention establishes a square grid vector matrix covering the target space coordinate data range in the target space coordinate data range, and the side length of each square grid is not more than 100 mm and a corresponding data set is established, and a square grid vector matrix database is collectively established, which can realize the identification of urban inefficient spaces with decimeter-level accuracy and improve the granularity of identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flow chart of the method for identifying and classifying inefficient urban spaces of the present invention;

[0033] Figure 2 A schematic diagram of sites that are screened and not included in the urban inefficient space according to an embodiment of the present invention;

[0034] Figure 3 This is a schematic diagram of the classification results of the shape features of urban inefficient spaces according to an embodiment of the present invention;

[0035] Figure 4 Schematic diagram of the classification results of urban inefficient spatial functional characteristics according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0037] like Figure 1 As shown, the method for identifying and classifying inefficient urban spaces includes the following steps.

[0038] 1. Spatial vitality data collection.

[0039] Extract the target spatial coordinate data range of the land use boundaries of park green land and square land in the target urban area. Use millimeter wave radar sensors to build a radar detection matrix and fully cover the target spatial coordinate data range, and collect spatial distribution data of pedestrian flow every day and every hour within a week.

[0040] The land use boundaries of parks and green spaces and square land in the target urban area are the land use boundaries of parks and green spaces and square land in the urban land use planning map of the target urban area issued by the local government in accordance with the latest national "Urban Land Classification and Planning and Construction Land Standards".

[0041] The spatial distribution data of passenger flow is collected daily and hourly within the week, wherein a week is seven consecutive days excluding special holidays and extreme weather days.

[0042] 2. Preliminary selection of inefficient urban spaces.

[0043] In the target space coordinate data range, a square grid vector matrix is ​​established to cover the target space coordinate data range to establish a corresponding data set, and the side length of each square grid is not more than 100mm, to achieve decimeter-level accuracy, and a square grid vector matrix database is established. The collected spatial distribution data of the flow of people is cut according to the square grid and entered into the corresponding data set. The first threshold of this embodiment is 20%. When the average daily flow of people in a square grid data set ranks in the bottom 20% of all square grid data sets, the corresponding space of this square grid is determined to be the inefficient space of the primary city.

[0044] The square grid vector matrix can be established by using a fishing net tool of a geographic information processing system, and the square grids can be increased or decreased according to the target spatial coordinate data range.

[0045] 3. Real scene image data collection.

[0046] Extract the inefficient spatial coordinate data range of the primary selected cities, and use an aerial drone to collect real-scene image data within the coordinate data range. If the coordinate data range is in a drone-free zone, use a panoramic fisheye network camera to collect real-scene image data within the coordinate data range.

[0047] 4. Reselect inefficient urban spaces.

[0048] The GrabCut algorithm is used to identify and classify the real-life image data of urban inefficient spaces, and obtain facility-type square grids, green space-type square grids, and pavement-type square grids. The facility-type elements are further refined according to the key feature points of each type of facility through feature point detection and description (SIFT) to identify the image function category, and the facility-type square grids are identified to obtain municipal facility-type square grids, transportation facility-type square grids, and activity facility-type square grids. The municipal facility-type square grids and transportation facility-type square grids are excluded from the preliminary selection of urban inefficient spaces; the second threshold value of this embodiment is 10 m2. If the area of ​​space formed by the adjacent connection of the square grids in the remaining preliminary selection of urban inefficient spaces is less than 10 m2, the corresponding space of such square grids is excluded as urban inefficient space.

[0049] The use of the Grabcut algorithm to identify and classify urban inefficient space real-scene image data refers to using the GrabCut segmentation algorithm based on image theory to segment the foreground objects and background of urban inefficient space real-scene image data; defining one or more rectangles containing foreground objects in the acquired urban inefficient space real-scene image data, the foreground objects include facilities, green spaces, and pavement, the outer area of ​​the rectangle is automatically identified as the image background, and the inner area of ​​the rectangle can use the background data of the outer area to distinguish the foreground object from the background; using a Gaussian mixture model to classify the preliminarily classified background and foreground object data Modeling, undefined pixels are marked as possible foreground objects or backgrounds. Each pixel in the image is connected to adjacent pixels through virtual edges. After the connection is completed, if the two endpoints of a virtual edge are of different types, that is, one endpoint belongs to the foreground object and the other endpoint belongs to the background, the virtual edge between them is cut off. If they belong to the same type, the virtual edge is retained, thereby separating the foreground object from the background in the real-life image data of urban inefficient spaces. By identifying the foreground objects, the real-life image data of urban inefficient spaces is preliminarily identified as facility-type square grids, green space-type square grids, and pavement-type square grids.

[0050] The feature point detection and description (SIFT) method further refines the image function category recognition of facility elements according to the key feature points of various facilities, which refers to the feature point search of the classified facility real-life image data, and the identification of feature points with constant scale and rotation angle in the image by searching the image position at all scales in combination with the Gaussian differential function to obtain discrete spatial feature points, and then the fitting function is used to accurately determine the position and scale of the feature points, while removing feature points with unstable edges; the main direction matching of the feature points is based on the local gradient direction of the image, and a main direction is assigned to each feature point; thus, each feature point has three key information: position, scale, and direction, and the image area adjacent to the feature point is divided into blocks, The feature points are converted into 128-dimensional feature vectors, and clustering operations are performed on all feature vectors. The converged and unchanged cluster centers are obtained through iterative updates. The cluster centers are the corresponding visual words in the visual vocabulary model. All cluster centers constitute a set of visual words. The distance between each feature point in the image and each cluster center is counted, and the weight of the nearest cluster center in the visual vocabulary histogram is increased by 1. The image data corresponding to each square network can be represented by the visual vocabulary histogram. By statistics and classification of the distribution of the visual vocabulary histogram, a training database based on the visual vocabulary model is established, and the facility elements can be further classified and identified as municipal facility square grids, transportation facility square grids, and activity facility square grids.

[0051] 5. Classification of inefficient urban spaces.

[0052] (1) The square grids corresponding to the selected urban inefficient spaces are merged according to the adjacency relationship, and each merged result is used as an independent urban inefficient space for classification. According to the shape characteristics formed by the adjacency between the square grids in the real-life image data within the coordinate data range of the urban inefficient space, the random forest model is used to classify the shape characteristics of the urban inefficient space, and the scattered urban inefficient space, linear urban inefficient space, and surface urban inefficient space are obtained.

[0053] The use of the random forest model to classify the shape features of urban inefficient spaces refers to selecting shape feature parameters as features used by the decision tree in the random forest to classify the shape features of urban inefficient spaces. The shape feature parameters include the area, rectangularity, and aspect ratio of the minimum circumscribed rectangle of the urban inefficient space, and the urban inefficient space is divided into three categories: scattered urban inefficient space, linear urban inefficient space, and planar urban inefficient space. The method of using the area of ​​urban inefficient space for classification is: based on the area of ​​the urban inefficient space being less than a threshold, the scattered urban inefficient space is screened out; based on the rectangularity of the urban inefficient space being less than a threshold, the linear (curve) urban inefficient space is screened out; based on the aspect ratio of the minimum circumscribed rectangle of the urban inefficient space being not less than a threshold, the linear (straight) urban inefficient space is screened out; the remaining urban inefficient space is planar urban inefficient space; both the linear (curve) urban inefficient space and the linear (straight) urban inefficient space belong to the linear urban inefficient space. By screening the linear (curve) urban inefficient space, it is possible to avoid misjudging the curve as a planar space, thereby improving the accuracy of classification. The specific classification formula is as follows:

[0054] (1.1) The method of using area classification is:

[0055]

[0056] Where S i is the area of ​​the inefficient space i in the city, S0 is the area feature classification threshold generated after the decision tree training;

[0057] (1.2) The method of using rectangularity classification is:

[0058]

[0059] Where R i is the area of ​​the urban inefficient space i, R0 is the rectangular feature classification threshold generated after the decision tree training; the calculation formula of the rectangular degree is:

[0060]

[0061] Where S i is the area of ​​the urban inefficient space i, S mer It is the area of ​​the minimum circumscribed rectangle of the urban inefficient space.

[0062] (1.3) The classification method using the aspect ratio of the minimum enclosing rectangle is:

[0063]

[0064] AR i is the area of ​​the inefficient space i in the city, and AR0 is the aspect ratio feature classification threshold of the minimum bounding rectangle generated after the decision tree training.

[0065] The merged urban inefficient space was input into the geographic information processing system software, and the minimum enclosing rectangle of the urban inefficient space was generated. The area, rectangularity and aspect ratio of the minimum enclosing rectangle of the urban inefficient space were calculated. The shape feature parameter data set of the urban inefficient space, which included the square grid of the urban inefficient space, feature parameters and annotated shape feature labels, was constructed. The data set was divided into a training set and a test set. The data volume ratio of the training set and the test set was 8:2. The shape feature labels included scattered urban inefficient space, linear (straight) urban inefficient space, linear (curve) urban inefficient space and surface urban inefficient space. The bootstrap method was used to extract samples from the test set to obtain 10 sub-training sets. The sub-training sets were used to train the random forest model, and the test set was used to evaluate the performance of the trained model. When the accuracy of the classification result output by the model was less than 80%, the bootstrap method sampling and model training were performed again, otherwise the model training was completed. The data in the current urban inefficient space shape feature parameter database were input into the trained random forest model to classify the shape features of the urban inefficient space.

[0066] (2) Based on the number ratio of each type of square grid contained in the urban inefficient space in the image function category identification in step 4, the functional characteristic parameters of the urban inefficient space are clustered using the agglomerative hierarchical clustering algorithm to classify the functional characteristics of the urban inefficient space, and the urban inefficient space is divided into space-poor urban inefficient space, greening-deficient urban inefficient space, facility-deficient urban inefficient space, and functionally inefficient urban inefficient space.

[0067] The use of agglomerative hierarchical clustering algorithm to cluster the functional characteristic parameters of urban inefficient spaces and classify the functional characteristics of urban inefficient spaces refers to counting the number of activity facility-type square grids and green space-type square grids obtained in step 4 in the urban inefficient spaces, calculating the ratio of the above two types of square grids in the urban inefficient spaces, obtaining the functional characteristic parameters of urban inefficient spaces, storing the calculation results and constructing a database of the current functional characteristic parameters of urban inefficient spaces. The agglomerative hierarchical clustering algorithm was used to cluster the functional characteristic parameters of urban inefficient spaces. The number of clusters was set to 4, and four clustering results were obtained. The categories with low greening square grid proportions and low activity facility square grid proportions were defined as spatially poor urban inefficient spaces, the categories with low greening square grid proportions and high activity facility square grid proportions were defined as greening-deficient urban inefficient spaces, the categories with high greening square grid proportions and low activity facility square grid proportions were defined as facility-deficient urban inefficient spaces, and the categories with high greening square grid proportions and high activity implementation square grid proportions were defined as functionally inefficient urban inefficient spaces, thus realizing the classification of functional characteristics of urban inefficient spaces.

[0068] 6. Interactive display of classification results.

[0069] The shape feature classification and functional feature classification data of urban inefficient spaces in the target urban area are imported into the holographic sand table display device to display the spatial distribution of urban inefficient spaces of various shapes and functions in the target urban area, conduct human-computer virtual interaction and record feedback information.

[0070] The technical solution of the present invention will be described in detail below by taking the area along Central Road in Nanjing as an example.

[0071] (1) Spatial vitality data collection, including:

[0072] (1.1) Based on the land use boundaries of parks and squares in the urban land use planning map of the area along Central Road in Nanjing, which was published by the Nanjing Planning Department in accordance with the latest national "Urban Land Classification and Planning and Construction Land Standards", the target spatial coordinate data range of the land use boundaries of parks and squares in the area along Central Road in Nanjing was further extracted.

[0073] (1.2) In the area along Central Road in Nanjing, use millimeter-wave radar sensors with a detection angle of more than 120°, an effective detection distance of more than 150m, and a detection accuracy of more than 100mm to build a radar detection matrix with an interval of no more than 150m and fully cover the coordinate data range of the area along Central Road in Nanjing, and collect spatial distribution data of pedestrian flow every day and every hour within a week. Specifically, collect spatial distribution data of pedestrian flow every day and every hour for seven consecutive days, excluding special holidays and extreme weather days.

[0074] (2) Preliminary selection of urban inefficient spaces, including:

[0075] (2.1) Using the fishing net tool of ArcGIS geographic information processing system, a square grid vector matrix covering the target spatial coordinate data range is established in the area along Central Road in Nanjing. Each square grid has a side length of 100 mm and a corresponding data set is established. A square grid vector matrix database is established. A total of 45,110,223 square grids are obtained.

[0076] (2.2) The collected spatial distribution data of human flow are cut according to the square grid and entered into the corresponding data set. When the daily average human flow in a square grid data set ranks in the bottom 20% of all square grid data sets, the corresponding space of this square grid is determined to be the inefficient space of the primary city.

[0077] (3) Collect real-life image data of urban inefficient spaces, including:

[0078] (3.1) Extract the inefficient spatial coordinate data range of the primary selected cities, and use an aerial photography drone with an effective pixel of more than 12 million and a field of view of more than 120° to collect real-life image data within the coordinate data range.

[0079] (3.2) If the coordinate data range is within the drone no-fly zone, a panoramic fisheye network camera with an effective pixel of more than 12 million and a field of view of more than 180° is used to collect real-scene image data within the coordinate data range.

[0080] (4) The GrabCut algorithm is used to identify and classify the real-life image data of urban inefficient spaces, and square grids of facilities, green space, and pavement are obtained. The feature point detection and description (SIFT) is used to further refine the image function category recognition of facility elements according to the key feature points of each type of facility, and square grids of municipal facilities, transportation facilities, and activity facilities are identified. The corresponding space after excluding some square grids is the urban inefficient space, including:

[0081] (4.1) One or more rectangles containing foreground objects are defined in the acquired real-life image data of urban inefficient spaces. The foreground objects include facilities, green spaces, and pavements. The outer area of ​​the rectangle is automatically identified as the image background, and the inner area of ​​the rectangle can use the background data of the outer area to distinguish the foreground object from the background.

[0082] (4.2) A Gaussian mixture model is used to model the preliminarily classified background and foreground object data. Undefined pixels are marked as possible foreground objects or backgrounds. Each pixel in the image is connected to adjacent pixels through virtual edges. After the connection is completed, if the two endpoints of a virtual edge belong to different types, that is, one endpoint belongs to the foreground object and the other endpoint belongs to the background, the virtual edge between them is cut off. If they belong to the same type, the virtual edge is retained, thereby achieving the separation of foreground objects and backgrounds in the real-life image data of urban inefficient spaces.

[0083] (4.3) By identifying foreground objects, the real-life image data of urban inefficient spaces is preliminarily identified as facility-type square grids, green space-type square grids, and pavement-type square grids.

[0084] (4.4) Feature point search is performed on the classified facility-type real-scene image data. By searching the image positions at all scales, the Gaussian differential function is combined to identify the feature points in the image whose scale and rotation angle remain unchanged, and discrete spatial feature points are obtained. The fitting function is then used to accurately determine the position and scale of the feature points, while removing feature points with unstable edges.

[0085] (4.5) The main direction matching of feature points is based on the local gradient direction of the image. A main direction is assigned to each feature point, so that each feature point has three key information: position, scale, and direction. By dividing the image area adjacent to the feature point into blocks, the feature point is converted into a 128-dimensional feature vector. All feature vectors are clustered and the converged and invariant cluster centers are obtained through iterative updates.

[0086] (4.6) The cluster center of the feature vector is the corresponding visual word in the visual vocabulary model. All cluster centers constitute a set of visual words. The distance between each feature point in the image and each cluster center is counted, and the weight of the nearest cluster center in the visual vocabulary histogram is increased by 1. The image data corresponding to each square network can be represented by the visual vocabulary histogram. By statistics and classification of the distribution of the visual vocabulary histogram, a training database based on the visual vocabulary model is established, and the facility elements can be further classified and identified as municipal facility square grids, transportation facility square grids, and activity facility square grids.

[0087] (4.7) Exclude the municipal facilities square grid and the transportation facilities square grid from the preliminary urban inefficient space. If the area of ​​the space formed by the adjacent connection of the remaining preliminary urban inefficient space square grids is less than 10㎡, the corresponding space of such square grids is excluded as urban inefficient space. Figure 2 Shown is a schematic diagram of sites along Central Road in Nanjing that have been screened out and not included in the city’s inefficient spaces.

[0088] (5) Based on the current status and functional characteristics of the urban inefficient spaces obtained through multiple selection, the urban inefficient spaces are classified into the following categories:

[0089] (5.1) The square grids corresponding to the selected urban inefficient spaces are merged according to the adjacency relationship, and each merged result is used as an independent urban inefficient space for classification.

[0090] (5.2) Use the random forest model to classify the shape characteristics of urban inefficient spaces. Select shape characteristic parameters as the features used by the decision tree in the random forest to classify the shape characteristics of urban inefficient spaces. The shape characteristic parameters include the area, rectangularity, and aspect ratio of the minimum enclosing rectangle of the urban inefficient space. Input the merged urban inefficient space into the GIS software, generate the minimum enclosing rectangle of the urban inefficient space, calculate the area, rectangularity, and aspect ratio of the minimum enclosing rectangle of the urban inefficient space, and construct a database of the shape characteristic parameters of the current urban inefficient space.

[0091] (5.3) Construct a dataset of urban inefficient space shape feature parameters that includes urban inefficient space square grids, feature parameters, and annotated shape feature labels. Divide this dataset into a training set and a test set. The ratio of the data volume between the training set and the test set is 8:2. The shape feature labels include scattered urban inefficient space, linear (straight) urban inefficient space, linear (curved) urban inefficient space, and surface urban inefficient space. Use the bootstrap method to extract samples from the test set to obtain 10 sub-training sets. Use the sub-training sets to train the random forest model, and use the test set to evaluate the performance of the trained model. When the accuracy of the classification result output by the model is less than 80%, perform bootstrap sampling and model training again. Otherwise, complete the model training.

[0092] (5.4) Input the data in the current urban inefficient space shape characteristic parameter database into the trained random forest model to classify the urban inefficient space shape characteristics. The classification results are scattered point-shaped urban inefficient space, linear urban inefficient space, and surface urban inefficient space. Figure 3 Shown is a schematic diagram of the classification results of the shape characteristics of urban inefficiency spaces along Central Road in Nanjing.

[0093] (5.5) Use the agglomerative hierarchical clustering algorithm to cluster the functional characteristic parameters of urban inefficient spaces, classify the functional characteristics of urban inefficient spaces, count the number of square grids of activity facilities and green space obtained in step 4 in urban inefficient spaces, calculate the ratio of the above two types of square grids in urban inefficient spaces, obtain the functional characteristic parameters of urban inefficient spaces, store the calculation results and construct a database of the current functional characteristic parameters of urban inefficient spaces.

[0094] (5.6) Input the existing database of urban inefficient space function characteristic parameters into the agglomerative hierarchical clustering algorithm, cluster the urban inefficient space function characteristic parameters, set the number of clusters to 4, and obtain four types of clustering results. The categories with low greening square grid ratio and low activity facility square grid ratio are defined as spatially poor urban inefficient space, the categories with low greening square grid ratio and high activity facility square grid ratio are defined as greening-deficient urban inefficient space, the categories with high greening square grid ratio and low activity facility square grid ratio are defined as facility-deficient urban inefficient space, and the categories with high greening square grid ratio and high activity implementation square grid ratio are defined as functionally inefficient urban inefficient space, thus realizing the classification of urban inefficient space function characteristics. Figure 4 The figure shows the spatial distribution of urban inefficient spaces with space poverty, urban inefficient spaces with lack of green space, urban inefficient spaces with lack of facilities and urban inefficient spaces with inefficient functional use in the area along Central Road in Nanjing.

[0095] (6) Interactive display of classification results, including:

[0096] (6.1) The shape feature classification and functional feature classification data of urban inefficient spaces along Central Road in Nanjing will be imported into a holographic sand table display device equipped with a wearable 3D motion capture system and a large screen with a resolution of 8K or above.

[0097] (6.2) Display the spatial distribution of urban inefficient spaces of various shapes and functions along the Central Road area in Nanjing, conduct human-computer virtual interaction and record feedback information.

[0098] The urban inefficient space identification and classification system of the present invention comprises:

[0099] A crowd flow spatial distribution collection unit is used to collect crowd flow spatial distribution data of a target space, where the boundary of the target space is the boundary between a park green space and a square land in an urban area;

[0100] A preliminary city inefficient space division unit is used to establish a square grid in the target space, divide the spatial distribution data of the flow of people into each grid, and the space corresponding to the grid whose flow of people is ranked lower than the first threshold is the preliminary city inefficient space;

[0101] An urban inefficient space identification unit is used to collect real-scene image data in the preliminarily selected urban inefficient space, and identify the functions of the preliminarily selected urban inefficient space through an image segmentation method and a feature point detection and description method, including square grids of activity facilities, square grids of green space, and square grids of pavement; the grid corresponding space formed by the spatial area exceeding the second threshold formed by the square grids of activity facilities, green space, and pavement being adjacently connected to each other is the urban inefficient space;

[0102] The urban inefficient space classification unit is used to merge the square grids of the urban inefficient space according to the adjacency relationship to obtain a composite urban inefficient space, classify the urban inefficient space according to the shape of the composite urban inefficient space to obtain a shape classification result; and classify the urban inefficient space according to the function of the urban inefficient space to obtain a function classification result.

[0103] The electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the method for identifying and classifying inefficient urban spaces is implemented.

[0104] The computer-readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor, the method for identifying and classifying urban inefficient spaces is implemented.

[0105] The computer-readable storage media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, flash memory, or any other medium that can be used to store the desired program code in the form of instructions or data structures and that can be accessed by a computer.

[0106] The processor is used to execute the computer program stored in the memory to implement each step of the method involved in the above embodiment.

Claims

1. A method for identifying and classifying inefficient urban spaces, characterized in that: The steps include: Collecting the spatial distribution data of the crowd flow in the target space, where the boundary of the target space is the land boundary between the park green space and the square land in the urban area; A square grid is established in the target space, and the spatial distribution data of the flow of people is divided into each grid. The space corresponding to the grid whose flow of people is ranked lower than the first threshold is the inefficient space of the preliminary selected city. Collect real-life image data of the inefficient spaces in the preliminary selected cities, and identify the functions of the inefficient spaces in the preliminary selected cities through image segmentation methods and feature point detection and description methods, including square grids of activity facilities, square grids of green spaces, and square grids of pavement; The space corresponding to the grid where the square grids of activity facilities, green space and pavement can be connected to each other and the space area exceeds the second threshold is the urban inefficient space; The square grids of the urban inefficient spaces are merged according to the adjacency relationship to obtain a composite urban inefficient space, and the urban inefficient spaces are classified according to the shapes of the composite urban inefficient spaces to obtain a shape classification result; The inefficient urban spaces are classified according to their functions and the functional classification results are obtained.

2. The method for identifying and classifying inefficient urban spaces according to claim 1, characterized in that: The square grids of the urban inefficient space are merged according to the adjacency relationship to obtain a composite urban inefficient space, and the urban inefficient space is classified according to the shape of the composite urban inefficient space, and the shape classification results include: According to the area, rectangularity and aspect ratio of the minimum circumscribed rectangle of the composite urban inefficient space, the urban inefficient space is divided into three categories: scattered urban inefficient space, linear urban inefficient space and surface urban inefficient space.

3. The method for identifying and classifying inefficient urban spaces according to claim 2, characterized in that: According to the fact that the area of ​​the urban inefficient space is less than the third threshold, the scattered urban inefficient space is screened out. According to the fact that the rectangularity of the urban inefficient space is less than the fourth threshold, the curved urban inefficient space is screened out. According to the fact that the aspect ratio of the minimum circumscribed rectangle of the urban inefficient space is not less than the fifth threshold, the straight urban inefficient space is screened out. The remaining urban inefficient space is the surface urban inefficient space. Both the curved urban inefficient space and the straight urban inefficient space belong to the linear urban inefficient space.

4. The method for identifying and classifying inefficient urban spaces according to claim 1, characterized in that: The urban inefficient spaces are classified according to their functions, and the functional classification results include: The functions of urban inefficient spaces were clustered using an agglomerative hierarchical clustering algorithm, and the functional classification results were obtained, including space-poor urban inefficient spaces, green space-deficient urban inefficient spaces, facility-deficient urban inefficient spaces, and functional use-inefficient urban inefficient spaces. Among them, the categories with low proportion of greening square grids and low proportion of activity facilities square grids are defined as spatially poor urban inefficient spaces, the categories with low proportion of greening square grids and high proportion of activity facilities square grids are defined as greening-deficient urban inefficient spaces, the categories with high proportion of greening square grids and low proportion of activity facilities square grids are defined as facility-deficient urban inefficient spaces, and the categories with high proportion of greening square grids and high proportion of activity implementation square grids are defined as functionally inefficient urban inefficient spaces.

5. The method for identifying and classifying inefficient urban spaces according to claim 1, characterized in that: The functions of collecting real-life image data in the inefficient spaces of the primary selected cities and identifying the inefficient spaces of the primary selected cities through image segmentation methods and feature point detection and description methods include: Taking facilities, green space and paved objects as foreground, the GrabCut algorithm is used to divide the preliminary selected urban inefficient spaces into facility square grids, green space square grids and paved square grids. The key feature points of various facilities are extracted through feature point detection and description methods, and the facility square grids are divided into municipal facility square grids, transportation facility square grids, and activity facility square grids.

6. The method for identifying and classifying inefficient urban spaces according to claim 1, characterized in that: The method also includes displaying the shape classification results and function classification results of the urban inefficient space in an urban sandbox.

7. The method for identifying and classifying inefficient urban spaces according to claim 1, characterized in that: Collecting the spatial distribution data of the flow of people in the target space includes: using the millimeter wave radar sensor to collect the spatial distribution data of the flow of people every day and every hour within a week, and the week is seven consecutive days excluding special holidays and extreme weather days.

8. A system for identifying and classifying inefficient urban spaces, characterized in that: include: A crowd flow spatial distribution collection unit is used to collect crowd flow spatial distribution data of a target space, where the boundary of the target space is the boundary between a park green space and a square land in an urban area; A preliminary city inefficient space division unit is used to establish a square grid in the target space, divide the spatial distribution data of the flow of people into each grid, and the space corresponding to the grid whose flow of people is ranked lower than the first threshold is the preliminary city inefficient space; An urban inefficient space identification unit is used to collect real-scene image data in the preliminarily selected urban inefficient space, and identify the functions of the preliminarily selected urban inefficient space through an image segmentation method and a feature point detection and description method, including square grids of activity facilities, square grids of green space, and square grids of pavement; the grid corresponding space formed by the spatial area exceeding the second threshold formed by the square grids of activity facilities, green space, and pavement being adjacently connected to each other is the urban inefficient space; The urban inefficient space classification unit is used to merge the square grids of the urban inefficient space according to the adjacency relationship to obtain a composite urban inefficient space, classify the urban inefficient space according to the shape of the composite urban inefficient space to obtain a shape classification result; and classify the urban inefficient space according to the function of the urban inefficient space to obtain a function classification result.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, the method for identifying and classifying urban inefficient spaces according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for identifying and classifying urban inefficient spaces according to any one of claims 1 to 7 is implemented.

Images