Community life circle delimitation method and device based on multi-source big data

Abstract

The application relates to a community life circle demarcation method and device based on multi-source big data. The method automatically extracts residential communities, public service facilities and urban road network feature information around a target community from multi-source big data such as Baidu, Gaode and other Internet maps, population heat maps and national "three surveys" data to make a planning base map, determines the optimal predicted path between the residential area and the public service facility area, obtains the predicted heat map image of the path in the planning base map in combination with the visited frequency of the public service facility area obtained by means of mobile phone signaling data, inputs the difference area, the residential area and the public service facility area into a boundary demarcation model to output the planning base map after boundary demarcation, and embeds the high-latitude information of the difference area into the boundary demarcation model for training, so that the life circle boundary can be more accurately demarcated, and different types of life circle boundaries such as five minutes, ten minutes, fifteen minutes and half an hour can be selected according to the use demand of users.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, and in particular relates to a method and apparatus for delineating community living circles based on multi-source big data. Background Technology

[0002] Currently, communities are the basic units of residents' lives. The configuration of service facilities within communities directly affects residents' quality of life and lifestyles, making them important spatial units in the process of urban renewal and a key focus of community planning. Delineating the boundaries of living circles in community planning is a key research focus and challenge. Accurately delineating the boundaries of living circles is essential for the more rational allocation of community public service facilities and the improvement of residents' living experience. To delineate community living circles, existing technologies mainly rely on information about residential communities, facilities, and roads in vector maps, and on residents wearing GPS-enabled devices to track their daily activities, thereby obtaining the approximate range of the community living circle. However, these methods are costly, have small sample sizes, and low data quality. Furthermore, community living circle delineation needs to be based on diverse living scenarios, and these scenarios vary from community to community. Existing methods for delineating community living circles have poor generalization ability, making it difficult to generalize methods from limited scenarios to various types of community scenarios, resulting in low accuracy and poor practicality of the final delineation results. Therefore, improving the accuracy of community living circle delineation to achieve the rational allocation of community facilities has become an urgent problem to be solved. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method and apparatus for delineating community living circles based on multi-source big data, in order to solve the problem of unreasonable community facility configuration caused by the low accuracy of existing community living circle delineation.

[0004] In a first aspect, embodiments of this application provide a method for delineating community living circles based on multi-source big data, the method comprising:

[0005] Based on multi-source big data, a planning base map of the target community is obtained, and the center point representation of each residential area and each public service facility area in the planning base map is extracted, and the path representation of the road area in the planning base map is extracted.

[0006] Using a GIS traffic network analysis model and a heuristic search algorithm, combined with the central point representation and the path representation, the optimal predicted path between each residential area and each public service facility area is determined.

[0007] Based on the access frequency of each public service facility area, the pixel units of the optimal predicted path connected to the corresponding public service facility area are assigned values ​​on the planning base map. By traversing each public service facility area, a predicted heat map of the path in the planning base map is obtained.

[0008] Obtain the actual thermal image of each path in the road area, compare the predicted thermal image with the actual thermal image, and determine the area where the comparison result is greater than the difference threshold as the difference area;

[0009] The location of each residential area and each public service facility area in the planning base map, as well as the location of the differential area, are input into the trained boundary delineation model to generate a planning base map after delineating the boundaries of the living circle. The training set of the trained boundary delineation model consists of the known locations of differential areas, residential areas, and public service facility areas, and the planning base map with different types of living circle boundaries is marked.

[0010] Secondly, embodiments of this application provide a community living circle delineation device based on multi-source big data, the community living circle delineation device comprising:

[0011] The representation data extraction module is used to obtain the planning base map of the target community based on multi-source big data, extract the center point representation of each residential area and each public service facility area in the planning base map, and extract the path representation of the road area in the planning base map.

[0012] The optimal path prediction module is used to determine the optimal predicted path between each residential area and each public service facility area by using a GIS traffic network analysis model and a heuristic search algorithm, combined with the center point representation and the path representation.

[0013] The path heat map prediction module is used to assign values ​​to the pixel units of the optimal predicted path connected to the corresponding public service facility area on the planning base map according to the access frequency of each public service facility area, and traverse each public service facility area to obtain a predicted heat map image of the path in the planning base map.

[0014] The difference region determination module is used to acquire the actual thermal image of each path in the road area, compare the predicted thermal image with the actual thermal image, and determine the region where the comparison result is greater than the difference threshold as the difference region;

[0015] The living circle boundary delineation module is used to input the location of each residential area and each public service facility area in the planning base map, as well as the location of the differential area, into the trained boundary delineation model to generate a planning base map after delineating the living circle boundaries. The training set of the trained boundary delineation model consists of the known locations of differential areas, residential areas, and public service facility areas, and the planning base map with different types of living circle boundaries is marked.

[0016] The beneficial effects of this application embodiment compared with the prior art are as follows: This application directly extracts features of residential areas, public service facility areas, and road areas from the planning base map of the target community. Then, it uses a GIS traffic network analysis model and a heuristic search algorithm to determine the optimal predicted path between each residential area and each public service facility area. Combining the access frequency of the public service facility area, it assigns values ​​to the pixel units corresponding to the optimal predicted path connecting the public service facility area, thereby obtaining the predicted heat map of the path in the planning base map. It compares this with the actual heat map to find the difference areas. The difference areas, residential areas, and public service facility areas are used as inputs to the trained boundary delineation model, and then the planning base map after boundary delineation is output, realizing the delineation of the living circle boundary of the target community. By predicting the optimal path and the heat map of the path between the residential area and the public service facility area, and comparing it with the actual path heat map, the difference areas are obtained. These difference areas are used as high-dimensional information to embed into the boundary delineation model for training, resulting in a boundary delineation model that is more in line with reality. Thus, using the trained boundary delineation model, a more accurate living circle boundary delineation result can be obtained. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a method for delineating community living circles based on multi-source big data, provided in Embodiment 1 of this application.

[0019] Figure 2 This is a schematic diagram of a road binary map of a road area provided in Embodiment 1 of this application;

[0020] Figure 3 This is a flowchart illustrating a heuristic search process provided in Embodiment 1 of this application;

[0021] Figure 4 This is a schematic diagram of a predicted thermal image provided in Embodiment 1 of this application;

[0022] Figure 5 This is a schematic diagram of an actual thermal image provided in Embodiment 1 of this application;

[0023] Figure 6 This is a schematic diagram of the structure of a boundary delineation model provided in Embodiment 1 of this application;

[0024] Figure 7 This is a schematic diagram of a community living circle delineation device based on multi-source big data provided in Embodiment 2 of this application. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0026] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0027] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0029] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0030] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0031] The terminal device in this application embodiment may be a handheld computer, desktop computer, laptop computer, ultra-mobile personal computer (UMPC), netbook, cloud terminal device, personal digital assistant (PDA), etc. This application embodiment does not impose any restrictions on the specific type of terminal device.

[0032] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0033] Foundational technologies in artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, terminal device technology, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0034] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0035] To illustrate the technical solution of this application, specific embodiments are described below.

[0036] See Figure 1This is a flowchart illustrating a method for delineating community living circles based on multi-source big data, provided in Embodiment 1 of this application. The aforementioned method is applied to a terminal device, which connects to a corresponding database via the internet to obtain relevant planning base maps, facility access traffic, real-time path heatmaps, and other data, thereby achieving big data collection. Figure 1 As shown, the method for delineating the community living circle may include the following steps:

[0037] Step S101: Obtain the planning base map of the target community based on multi-source big data, extract the center point representation of each residential area and each public service facility area in the planning base map, and extract the path representation of the road area in the planning base map.

[0038] In this application, the aforementioned multi-source big data may include internet maps such as Baidu and Gaode, population heat maps, and data from the Third National Land Survey. The aforementioned target community can be any community that requires the delineation of its living area boundaries. The planning base map is an image containing features such as the community's residential areas, public service facility areas, and road areas, for example, a community architectural planning base map. This planning base map is a drawing designed to actual dimensions, and its size can be enlarged according to the scale on the map. The residential area can refer to the area used for residential purposes on the planning base map, the public service facility area can refer to the area used for placing public service facilities on the planning base map, and the road area can refer to all roads on the planning base map.

[0039] Center point representation is the representation of closed areas such as residential areas and public service facility areas in the form of center points. That is, the center point of a certain area in the planning base map is extracted and used as the representation of that area. Path representation is the representation of road areas in the form of linear paths. That is, the linear path of a certain road in the planning base map is extracted and used as the representation of that road.

[0040] The center point of the residential area is represented by the centroid of the residential area graphic, and the corresponding image coordinates are obtained. The center point of the public service facility area is represented by the centroid of the public service facility area graphic, or by the specific location of the exit of the public service facility area, and the corresponding image coordinates are obtained. Lines are extracted from the road areas in the planning base map to obtain a binary road image. In this binary road image, the pixel value of the pixel unit representing the road area is 1, and the pixel value of the pixel unit in other areas is 0, thus forming a linear path.

[0041] The base map of the target community can essentially be an electronic planning and design map (GIS data format). The following content needs to be marked on the map and shown in a table. The table is only an example and does not limit the actual use, as follows:

[0042]

[0043] Step S102: Using a GIS traffic network analysis model and a heuristic search algorithm, combining center point representation and path representation, determine the optimal predicted path between each residential area and each public service facility area.

[0044] In this application, a GIS traffic network analysis model is used to predict the optimal path between residential areas and public service facility areas on the planning base map. A heuristic search algorithm can be used as the prediction method in this process, as follows:

[0045] Based on the center point representation of residential areas and public service facility areas, the image coordinates of the corresponding center points of residential areas and public service facility areas in the image can be determined. Based on the path representation of road areas, the image coordinates of the paths in the road areas in the image can be determined.

[0046] For any residential area's center point and any public service facility area's center point, a heuristic search algorithm can be used to predict the path connecting the two center points via actual roads. This path is the optimal predicted path between the corresponding residential area and the corresponding public service facility area.

[0047] Optionally, a heuristic search algorithm, combining centroid representation and path representation, can be used to determine the optimal predicted path between each residential area and each public service facility area, including:

[0048] The theoretical path is formed by connecting the center point of each residential area with the center point of each public service facility area;

[0049] Starting from one end of any theoretical path and ending at the other, search within a preset range to identify the pixel unit that meets the preset conditions as the search point. The preset range is a fan-shaped area with an included angle of no more than 180°. The vertex of the fan-shaped area is the starting point, the center line of symmetry of the fan-shaped area is the corresponding theoretical path, and the radius of the fan-shaped area is the target radius.

[0050] Using the search point as the starting point, search again within the corresponding preset range until the search point coincides with the endpoint, and determine the optimal predicted path of the corresponding theoretical path by connecting all search points.

[0051] Using the aforementioned binary road image, combined with the image coordinates of the two center points, the positions of the two center points in the image are determined, and the two center points are directly connected to form the theoretical path, that is, the straight-line distance between the two center points.

[0052] like Figure 2The image shown is a binary image of a road, where the black curved lines represent the road. Taking the route from residential area C to the park as an example, the center point of residential area C is used as the starting point, and the coordinates of the park's main entrance are used as the ending point. A fan-shaped search is performed starting from the starting point. The fan-shaped search range is determined by the ray direction defined by the theoretical path. The angle between the two boundaries of the fan-shaped search range and the ray direction is set to 90°, and the fan radius is set to 2 pixels. For pixels with a value of 1 within the fan-shaped search range, it is calculated whether the pixel meets a preset condition. The preset condition can be that the vertical distance between the pixel and the theoretical path is minimized, or that the sum of the vertical distance between the pixel and the theoretical path and the straight-line distance to the endpoint is minimized.

[0053] Pixels that meet preset conditions are used as search points, and the search continues from these points until they coincide with the destination. The resulting predicted path is the optimal predicted path. Figure 3 The diagram shown is a flowchart of the search process described above.

[0054] Optionally, starting from the starting point, a search is performed within a preset range to determine pixel units that meet preset conditions as search points, including:

[0055] Starting from the starting point, search within a preset range to determine all pixel units within the preset range and on the path representation;

[0056] Calculate the vertical distance between each pixel unit and its corresponding theoretical line, as well as the straight-line distance to the endpoint.

[0057] Sum the vertical distance and the linear distance corresponding to each pixel unit, and determine the pixel unit with the smallest sum as the search point.

[0058] Wherein, the vertical distance between the pixel unit and the theoretical route is d_1, the straight-line distance between the pixel unit and the endpoint is d_2, and the pixel unit corresponding to min(d_1+d_2) within the fan-shaped search range is selected as the search point.

[0059] Step S103: Based on the access frequency of each public service facility area, assign values ​​to the pixel units of the optimal predicted path connected to the corresponding public service facility area on the planning base map, traverse each public service facility area, and obtain the predicted heat map of the path on the planning base map.

[0060] In this application, the frequency of visits to public service facility areas is used as the basis for modifying path pixel units, thereby characterizing the path's popularity and determining its thermal image. The frequency of visits can be collected using mobile signaling OD data.

[0061] like Figure 4As shown, based on the access frequency S, the pixel units of the path connecting the four public service facility areas are modified. The path pixel units after being given initial values ​​are Gaussian blurred to obtain the predicted path heat map. The size of the Gaussian kernel can be set by the implementer, but it must be consistent with the size when generating annotation data for subsequent key point detection. In this application, the Gaussian kernel size is set to 5*5. All predicted path heat maps are superimposed point by point and the average value is taken to obtain the predicted heat map.

[0062] Optionally, before assigning values ​​to the pixel units of the optimal predicted path connecting each public service facility area on the planning base map based on the access frequency of each public service facility area, the following steps are also included:

[0063] Obtain the number of visitors per unit time for each public service facility area, and determine the total number of visitors for all public service facility areas within the unit time.

[0064] The ratio of the number of visitors per unit to the total number of visitors for each public service facility area is determined as the visit frequency of the corresponding public service facility area.

[0065] The frequency of visits to public service facility areas is represented by the ratio of the average daily visits to a single facility to the average daily visits to all facilities. This frequency of visits can be used to characterize the popularity of paths connected to public service facility areas.

[0066] Optionally, based on the access frequency of each public service facility area, the pixel units of the optimal predicted path connecting to the corresponding public service facility area are assigned values ​​on the planning base map. This process is repeated for each public service facility area to obtain a predicted heatmap of the paths on the planning base map, including:

[0067] For a public service facility area, the pixel units of the optimal preset path connected to the public service facility area are assigned values ​​on the planning base map by multiplying the access frequency of the public service facility area by the maximum pixel value. The pixel units of other paths in the planning base map are assigned values ​​of 0, thus obtaining the path heatmap corresponding to the public service facility area.

[0068] Traverse each public service facility area to obtain the corresponding path heatmap. Sum the pixel values ​​of each pixel unit in all path heatmaps and calculate the average value. Determine the average value as the predicted pixel value of the corresponding pixel unit and assign it to the corresponding pixel unit to obtain the predicted heatmap of the path in the planning base map.

[0069] The calculation method for assigning values ​​is 255 * access frequency. Pixel units for other paths are assigned a value of 0, thus obtaining a path heatmap for a public service facility area. By overlaying all the path heatmaps and averaging them, a predicted heatmap can be formed. Through the above process, the heatmap of a path can be predicted relatively accurately.

[0070] Step S104: Obtain the actual thermal image of each path in the road area, compare the predicted thermal image with the actual thermal image, and determine the area where the comparison result is greater than the difference threshold as the difference area.

[0071] In this application, the actual thermal image of each path in the road area can be retrieved from the corresponding database and can be accessed by the terminal device of this application. In one embodiment, the terminal device can acquire the corresponding personnel trajectory information in the road area and generate the actual thermal image using the personnel trajectory information. Figure 5 The image shown is a schematic diagram of the actual thermal image generated based on information such as the movement trajectory of people.

[0072] The pixel value of each pixel unit in the predicted thermal image is compared with the corresponding pixel unit in the actual thermal image to obtain the difference. The part with a large difference is the difference region.

[0073] Optionally, obtaining the actual thermal image of each path in the road area includes:

[0074] By using mobile phone signaling OD data, real-time images containing all paths in the road area can be obtained at any time, including pedestrian data.

[0075] Input the real-time image into the trained keypoint detection network, and output a heat map of the keypoints of the real-time image;

[0076] The key point heat images of two consecutive time points are weighted and averaged to determine the corresponding personnel trajectory image. All key point heat images within the target time period are traversed to obtain all personnel trajectory images.

[0077] The average of all the trajectory images of people is summed up, and the result of the summation and averaging is determined to be the actual heat map of each path in the road area.

[0078] Among them, pedestrian data at any time can be obtained through mobile phone signaling OD data, which can then be used to generate subsequent pedestrian trajectory images, thereby obtaining actual thermal images.

[0079] For any pedestrian's mobile phone signaling OD data, it can be mapped onto an image containing all paths in the road area, thereby obtaining a real-time image of the pedestrian's location within the target community. Keypoint extraction is performed on the location points in the real-time image, transforming the real-time image into a keypoint thermal image representing thermal information.

[0080] The keypoint detection network constructed above has an encoder-decoder structure. The network input is a real-time image representing the location of pedestrians, and the network output is a multi-channel keypoint thermal image. The training method of this keypoint detection network is as follows: the location points mapped by the mobile phone signaling OD data of pedestrians in multiple road areas at different times in different communities are used as keypoints, and Gaussian blur is performed with the keypoints as the center. The image after Gaussian blur is used as the labeled image, and the loss function is the mean square error loss function.

[0081] By overlaying all the key thermal images representing thermal information, we can obtain a thermal image representing the thermal contribution of all personnel to the road area, which is the actual thermal image containing each path within the road area.

[0082] In one embodiment, cameras are used to cover the road area of ​​the target community. The sampling frequency and sampling time of the cameras are consistent. The pose of each camera is fixed and known. The images collected by each camera at the same time are projected onto the ground coordinate system, and a real-time panoramic image of the road area at that time is obtained by image stitching.

[0083] A keypoint detection network was constructed with an encoder-decoder structure. The network input was a panoramic image of the road area, and the network output was a multi-channel keypoint heat map. The training method of this keypoint detection network was as follows: multiple panoramic images of the road area from different communities at different times were used as the training dataset. The center points of the two feet of the people were used as keypoints, and Gaussian blur was performed with the keypoints as the center. The Gaussian kernel size used for Gaussian blur was 5*5. The image after Gaussian blur was used as the labeled image, and the mean squared error loss function was used.

[0084] The key point heatmaps corresponding to panoramic images of the road area taken at consecutive times within the target community are superimposed. The superposition formula is p = αp' + (1-α)p”, where p is the superimposed pixel value, p' is the pixel value of the current frame, p” is the pixel value of the previous frame, and α is a weighting coefficient set to 0.05. Using this superposition method, a person trajectory map can be approximately obtained. By superimposing all the person trajectory maps and taking the average, the actual heatmap image can be obtained. It should be noted that when a key point reaches the center point of any residential area or the center point of a public service facility area, the superposition of that trajectory is stopped to avoid the trajectory pixel value gradually decreasing over time, which would affect subsequent comparisons.

[0085] The above process involves analyzing panoramic images of the community road area to obtain the movement trajectories of people in the road area.

[0086] In one implementation, the implementer can acquire actual thermal images for multiple time periods and select one of the more similar actual thermal images as the actual thermal image for comparison, so as to avoid special cases that may exist in a single time period, thereby improving the accuracy of comparative analysis.

[0087] Optionally, the predicted thermal image is compared with the actual thermal image, and the regions where the comparison result is greater than the difference threshold are identified as difference regions, including:

[0088] Perform a difference operation between a pixel unit at any position in a preset thermal image and a pixel unit at the corresponding position in an actual thermal image, and determine the pixel units whose difference operation result is greater than the difference threshold. The difference operation is to subtract the pixel values ​​of the two pixel units and take the absolute value.

[0089] The number of connected pixels in the same connected region is counted. If the number of connected pixels is greater than the threshold, the corresponding connected region is determined to be a difference region.

[0090] The process involves subtracting the predicted thermal image from the actual thermal image pixel by pixel and taking the absolute value. A difference threshold is set, and pixels in the difference image larger than the threshold are extracted. Then, the connected components of the difference image are analyzed. If the connected component of a pixel is smaller than a certain number of nodes, it is considered a noise point and is ignored. If the connected component is greater than or equal to the threshold, it is considered a difference region, and its centroid is extracted as the difference point. In this application, the difference threshold is set to 50, and the number threshold is set to 20. Implementers can adjust these thresholds according to actual conditions. This process reduces noise interference, thus more accurately identifying difference regions.

[0091] Step S105: Input the location of each residential area and each public service facility area in the planning base map, as well as the location of the difference area, into the trained boundary delineation model to generate the planning base map after delineating the living circle boundary.

[0092] In this application, the training set of the trained boundary delineation model consists of the locations of known disparate areas, residential areas, and public service facility areas, and includes a base map of planning boundaries for different types of living circles. The cross-entropy loss function is used during training. By embedding disparate areas within the community as high-dimensional information into the subsequent boundary delineation model, the generalization and accuracy of the community living circle delineation method are improved.

[0093] Construct a boundary delineation model, such as Figure 6The diagram shows the structure of the boundary delineation model. The network corresponding to this boundary delineation model contains two inputs, which are fed into the encoder. The extracted features are concatenated and then fed into the encoder-decoder structure. The decoder outputs the boundary delineation map.

[0094] Specifically, in the boundary delineation model, the difference regions are transformed into difference information images. The size of the difference information images is consistent with the size of the planning base map. The difference information images contain facility distribution information. A straight line search is performed from the difference regions in all directions. The public service facility area corresponding to the center point of the first public service facility area found in each direction is taken as the facility distribution information. All pixel units outside the facility distribution information are set to 0.

[0095] The planning base map is fed into the first encoder, and the difference information is fed into the second encoder. The first encoder and the second encoder have the same structure, only the convolution kernel parameters are different, so as to facilitate subsequent feature concatenation. The concatenated features are fed into the third encoder and decoder, and the planning base map after the boundary is defined is output.

[0096] This application directly extracts features from the planning base map of the target community, including residential areas, public service facility areas, and road areas. Then, a heuristic search algorithm is used to determine the optimal predicted path between each residential area and each public service facility area. Combining the access frequency of public service facility areas, the pixel units corresponding to the optimal predicted path connecting these public service facility areas are assigned values ​​to obtain the predicted heat map of the path in the planning base map. This predicted heat map is compared with the actual heat map to find discrepancy areas. These discrepancy areas, residential areas, and public service facility areas are used as inputs to a trained boundary delineation model, which then outputs a planning base map with delineated boundaries. This achieves the delineation of the living circle boundary of the target community. By predicting the optimal path and heat map of the path between residential areas and public service facility areas and comparing it with the actual path heat map, discrepancy areas are obtained. These discrepancy areas are used as high-dimensional information to embed into the boundary delineation model for training, resulting in a more realistic boundary delineation model. Thus, using the trained boundary delineation model can obtain more accurate living circle boundary delineation results. In addition, the above process does not require the collection of sensitive information such as the location of community residents, resulting in a high degree of anonymization.

[0097] Corresponding to the community living circle delineation method in the above embodiments, Figure 7 This diagram illustrates the structural block diagram of a community living circle delineation device based on multi-source big data, as provided in Embodiment 2 of this application. This device is applied to a terminal device, which connects to a corresponding database via the internet to obtain data such as the corresponding planning base map, facility access traffic, and real-time path heatmaps, thereby achieving big data collection. For ease of explanation, only the parts relevant to the embodiments of this application are shown.

[0098] See Figure 7 The community living area delineation device includes:

[0099] The representation data extraction module 71 is used to obtain the planning base map of the target community, extract the center point representation of each residential area and each public service facility area in the planning base map, and extract the path representation of the road area in the planning base map;

[0100] The optimal path prediction module 72 is used to determine the optimal predicted path between each residential area and each public service facility area using a heuristic search algorithm that combines centroid representation and path representation.

[0101] The path heat prediction module 73 is used to assign values ​​to the pixel units of the optimal predicted path connected to the corresponding public service facility area on the planning base map according to the access frequency of each public service facility area, and traverse each public service facility area to obtain the predicted heat image of the path on the planning base map.

[0102] The difference region determination module 74 is used to acquire the actual thermal image of each path in the road area, compare the predicted thermal image with the actual thermal image, and determine the area where the comparison result is greater than the difference threshold as the difference region.

[0103] The living circle boundary delineation module 75 is used to input the location of each residential area and each public service facility area in the planning base map, as well as the location of the differential area, into the trained boundary delineation model, and output the planning base map after the boundary is delineated. The training set of the trained boundary delineation model consists of the known locations of differential areas, residential areas and public service facility areas, and the planning base map with the living circle boundary marked.

[0104] Optionally, the above-mentioned optimal path prediction module 72 includes:

[0105] Theoretical path forming unit is used to connect the center point of each residential area with the center point of each public service facility area to form a theoretical path;

[0106] The search point determination unit is used to search within a preset range starting from one end of any theoretical path as the starting point and the other end as the ending point, and to determine the pixel unit that meets the preset conditions as the search point. The preset range is a fan-shaped area with an included angle of no more than 180°. The vertex of the fan-shaped area is the starting point, the center line of symmetry of the fan-shaped area is the corresponding theoretical path, and the radius of the fan-shaped area is the target radius.

[0107] The optimal path prediction unit is used to search again within the corresponding preset range using the search point as the starting point until the search point coincides with the endpoint, and to determine the optimal predicted path of the line connecting all search points as the corresponding theoretical path.

[0108] Optionally, the search point determination unit described above is specifically used for:

[0109] Starting from the starting point, search within a preset range to determine all pixel units within the preset range and on the path representation;

[0110] Calculate the vertical distance between each pixel unit and its corresponding theoretical line, as well as the straight-line distance to the endpoint.

[0111] Sum the vertical distance and the linear distance corresponding to each pixel unit, and determine the pixel unit with the smallest sum as the search point.

[0112] Optionally, the community living circle delineation device also includes:

[0113] The visitor count determination module is used to obtain the number of visitors per unit time for each public service facility area based on the frequency of visits to each public service facility area, before assigning values ​​to the pixel units of the optimal predicted path connected to the corresponding public service facility area on the planning base map, and to determine the total number of visitors for all public service facility areas within a unit time.

[0114] The access frequency determination module is used to determine the ratio of the number of visitors per unit to the total number of visitors for each public service facility area as the access frequency of the corresponding public service facility area.

[0115] Optionally, the aforementioned path thermal prediction module 73 includes:

[0116] The path heat map determination unit is used to assign values ​​to the pixel units of the optimal preset path connected to the public service facility area on the planning base map by multiplying the access frequency of the public service facility area by the maximum pixel value. The pixel units of other paths on the planning base map are assigned a value of 0, thus obtaining the path heat map corresponding to the public service facility area.

[0117] The path heat map prediction unit is used to traverse each public service facility area to obtain the path heat map corresponding to each public service facility area. The pixel values ​​of each pixel unit in all path heat maps are added together and averaged. The average value is determined as the predicted pixel value of the corresponding pixel unit and assigned to the corresponding pixel unit to obtain the predicted heat map of the path in the planning base map.

[0118] Optionally, the aforementioned difference region determination module 74 includes:

[0119] The real-time image acquisition unit is used to acquire real-time images of all paths in the road area at any time through mobile phone signaling OD data. The real-time images include pedestrian data.

[0120] The thermal image output unit is used to input real-time images into the trained keypoint detection network and output the keypoint thermal image of the real-time image.

[0121] The trajectory image determination unit is used to perform weighted averaging of the key point heat images at two consecutive time points, and determine the result of the weighted averaging as the corresponding personnel trajectory image. It traverses all key point heat images within the target time period to obtain all personnel trajectory images.

[0122] The thermal image determination unit is used to add up and average all personnel trajectory images to determine the actual thermal image of each path in the road area.

[0123] Optionally, the aforementioned difference region determination module 74 includes:

[0124] The difference operation unit is used to perform a difference operation between a pixel unit at any position in a preset thermal image and a pixel unit at the corresponding position in the actual thermal image, and to determine the pixel unit whose difference operation result is greater than the difference threshold. The difference operation is to subtract the pixel values ​​of the two pixel units and take the absolute value.

[0125] The difference region determination unit is used to count the number of connected pixels in the same connected region. If the number of connected pixels is greater than the threshold, the corresponding connected region is determined to be a difference region.

[0126] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0127] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for delineating community living circles based on multi-source big data, characterized in that, The methods for delineating community living circles include: Based on multi-source big data, a planning base map of the target community is obtained, and the center point representation of each residential area and each public service facility area in the planning base map is extracted, as well as the path representation of the road area in the planning base map. Using a GIS traffic network analysis model and a heuristic search algorithm, combined with the central point representation and the path representation, the optimal predicted path between each residential area and each public service facility area is determined. Based on the access frequency of each public service facility area, the pixel units of the optimal predicted path connected to the corresponding public service facility area are assigned values ​​on the planning base map. By traversing each public service facility area, a predicted heat map of the path in the planning base map is obtained. Obtain the actual thermal image of each path in the road area, compare the predicted thermal image with the actual thermal image, and determine the area where the comparison result is greater than the difference threshold as the difference area; The location of each residential area and each public service facility area in the planning base map, as well as the location of the differential area, are input into the trained boundary delineation model to generate a planning base map after delineating the boundaries of the living circle. The training set of the trained boundary delineation model consists of the known locations of differential areas, residential areas, and public service facility areas, and the planning base map with different types of living circle boundaries is marked. Based on the access frequency of each public service facility area, the pixel units of the optimal predicted path connected to the corresponding public service facility area are assigned values ​​on the planning base map. This process is repeated for each public service facility area to obtain a predicted heatmap of the paths on the planning base map, including: For a public service facility area, the pixel units of the optimal preset path connected to the public service facility area are assigned values ​​on the planning base map by multiplying the access frequency of the public service facility area by the maximum pixel value. The pixel units of other paths in the planning base map are assigned values ​​of 0, thus obtaining the path heatmap corresponding to the public service facility area. Traverse each public service facility area to obtain the path heatmap corresponding to each public service facility area. Sum the pixel values ​​of each pixel unit in all path heatmaps and calculate the average value. Determine the average value as the predicted pixel value of the corresponding pixel unit and assign it to the corresponding pixel unit to obtain the predicted heatmap of the path in the planning base map.

2. The method for delineating community living circles according to claim 1, characterized in that, Using a GIS traffic network analysis model and a heuristic search algorithm, combining the centroid representation and the path representation, the optimal predicted path between each residential area and each public service facility area is determined, including: The theoretical path is formed by connecting the center point of each residential area with the center point of each public service facility area; Taking one end of any theoretical path as the starting point and the other end as the ending point, a search is conducted within a preset range starting from the starting point to determine the pixel unit that meets the preset conditions as the search point. The preset range is a fan-shaped area with an included angle of no more than 180°. The vertex of the fan-shaped area is the starting point, the center line of symmetry of the fan-shaped area is the corresponding theoretical path, and the radius of the fan-shaped area is the target radius. Using the search point as the starting point, search again within the corresponding preset range until the search point coincides with the endpoint, and determine the line connecting all search points as the optimal predicted path of the corresponding theoretical path.

3. The method for delineating community living circles according to claim 2, characterized in that, Starting from the aforementioned starting point, a search is conducted within a preset range to determine pixel units that meet preset conditions as search points, including: Starting from the starting point, search within a preset range to determine all pixel units within the preset range and on the path representation; Calculate the vertical distance between each pixel unit and its corresponding theoretical path, as well as the straight-line distance between each pixel unit and the endpoint. Sum the vertical distance of each pixel unit with the straight-line distance of the path, and determine the pixel unit with the smallest sum as the search point.

4. The method for delineating community living circles according to claim 1, characterized in that, Before assigning values ​​to the pixel units on the planning base map for the optimal predicted path connecting each public service facility area to the corresponding public service facility area based on the access frequency of each area, the process also includes: Obtain the number of visitors per unit time for each public service facility area, and determine the total number of visitors for all public service facility areas within the unit time. The ratio of the number of visitors per unit to the total number of visitors for each public service facility area is determined as the visit frequency of the corresponding public service facility area.

5. The method for delineating community living circles according to claim 1, characterized in that, Obtaining the actual thermal image of each path in the road area includes: Real-time images containing all paths in the road area at any given time are obtained using mobile phone signaling OD data, and the real-time images include pedestrian data; The real-time image is input into a trained keypoint detection network, which outputs a thermal image of the keypoints in the real-time image. The key point heat images of two consecutive time points are weighted and averaged to determine the corresponding personnel trajectory image. All key point heat images within the target time period are traversed to obtain all personnel trajectory images. The average of all the personnel trajectory images is summed to determine the actual heat map of each path in the road area.

6. The method for delineating community living circles according to any one of claims 1 to 5, characterized in that, The predicted thermal image is compared with the actual thermal image, and the regions where the comparison result is greater than the difference threshold are identified as difference regions, including: Perform a difference operation between a pixel unit at any position in the predicted thermal image and a pixel unit at the corresponding position in the actual thermal image, and determine the pixel units whose difference operation result is greater than the difference threshold. The difference operation is to subtract the pixel values ​​of the two pixel units and take the absolute value. The number of connected pixels in the same connected region is counted. If the number of connected pixels is greater than the threshold, the corresponding connected region is determined to be a difference region.

7. A community living circle delineation device based on multi-source big data, characterized in that, The community living circle delineation device includes: The representation data extraction module is used to obtain the planning base map of the target community based on multi-source big data, extract the center point representation of each residential area and each public service facility area in the planning base map, and extract the path representation of the road area in the planning base map. The optimal path prediction module is used to determine the optimal predicted path between each residential area and each public service facility area by using a GIS traffic network analysis model and a heuristic search algorithm, combined with the center point representation and the path representation. The path heat map prediction module is used to assign values ​​to the pixel units of the optimal predicted path connected to the corresponding public service facility area on the planning base map according to the access frequency of each public service facility area, and traverse each public service facility area to obtain a predicted heat map image of the path in the planning base map. The difference region determination module is used to acquire the actual thermal image of each path in the road area, compare the predicted thermal image with the actual thermal image, and determine the region where the comparison result is greater than the difference threshold as the difference region; The living circle boundary delineation module is used to input the location of each residential area and each public service facility area in the planning base map, as well as the location of the difference area, into the trained boundary delineation model to generate a planning base map after delineating the living circle boundary. The training set of the trained boundary delineation model consists of the known locations of the difference area, residential area, and public service facility area, and the planning base map with different types of living circle boundaries is marked. The path thermal prediction module includes: The path heat map determination unit is used to assign values ​​to the pixel units of the optimal preset path connected to the public service facility area on the planning base map by multiplying the access frequency of the public service facility area by the maximum pixel value. The pixel units of other paths in the planning base map are assigned a value of 0, thereby obtaining the path heat map corresponding to the public service facility area. The path heat map prediction unit is used to traverse each public service facility area, obtain the path heat map corresponding to each public service facility area, add up the pixel values ​​of each pixel unit in all path heat maps, calculate the average value, determine the average value as the predicted pixel value of the corresponding pixel unit, and assign the value to the corresponding pixel unit to obtain the predicted heat map of the path in the planning base map.

8. The community living circle delineation device according to claim 7, characterized in that, The optimal path prediction module includes: Theoretical path forming unit is used to connect the center point of each residential area with the center point of each public service facility area to form a theoretical path; The search point determination unit is used to search within a preset range starting from one end of any theoretical path as the starting point and the other end as the ending point, and to determine the pixel unit that meets the preset conditions as the search point. The preset range is a fan-shaped area, the included angle of the fan-shaped area is not greater than 180°, the vertex of the fan-shaped area is the starting point, the center line of symmetry of the fan-shaped area is the corresponding theoretical path, and the radius of the fan-shaped area is the target radius. The optimal path prediction unit is used to use the search point as the starting point to search again within the corresponding preset range until the search point coincides with the endpoint, and to determine the optimal predicted path of the line connecting all search points as the corresponding theoretical path.

9. The community living circle delineation device according to claim 7 or 8, characterized in that, The difference region determination module includes: The difference operation unit is used to perform a difference operation between a pixel unit at any position in the predicted thermal image and a pixel unit at the corresponding position in the actual thermal image, and to determine the pixel unit whose difference operation result is greater than the difference threshold. The difference operation is to subtract the pixel values ​​of the two pixel units and take the absolute value. The difference region determination unit is used to count the number of connected pixels in the same connected region. If the number of connected pixels is greater than the threshold, the corresponding connected region is determined to be a difference region.

Images