Building function identification method and device based on interest points in building

Abstract

The embodiment of the invention provides a building function identification method and device based on an interest point in a building, and the method comprises the steps: extracting the interest point from a target building, and enabling the interest point to be a place which affects the function type of the target building; constructing graph structure data of the interest points; extracting a level feature, a type feature and a vertical spatial position feature of the point of interest, wherein the vertical spatial position feature is used for indicating a floor where the point of interest is located; extracting spatial form characteristics of the target building; fusing the level features, the spatial form features, the type features and the vertical spatial position features to obtain a point-of-interest feature matrix; and based on a preset weighted graph isomorphic network model, taking the graph structure data and the interest point feature matrix as input, and outputting the function type of the target building. The technical scheme provided by the invention is used for solving the problems in the prior art.

Description

Technical Field

[0001] This document relates to the field of geographic information technology, and particularly to a method and device for identifying building functions based on points of interest (POIs) within a building. Background Art

[0002] With the continuous advancement of urbanization and the construction of smart cities, researchers have begun to focus on the study of building semantic features, such as building types, social attributes of buildings, etc.

[0003] Currently, the research on buildings using graph neural networks generally includes two methods: (1) regarding a single building as a graph structure and the vertices of the building as graph nodes; (2) regarding a group of buildings as a graph structure and individual buildings as graph nodes.

[0004] However, there is an imbalance problem in the training samples of the above methods, which will make the recognition accuracy of the functions of various types of buildings relatively unbalanced. For example, the proportion of residential buildings is much larger than that of other types of buildings. During the recognition process, only the recognition accuracy of residential buildings is relatively high, while the recognition accuracy of other types of buildings is relatively low, or even unable to reach the preset value. Summary of the Invention

[0005] In view of the above solutions, the present application aims to propose a method and device for identifying building functions based on points of interest within a building to solve the problem of unbalanced recognition accuracy of the functions of various types of buildings.

[0006] In a first aspect, one or more embodiments of the present application provide a method for identifying building functions based on points of interest within a building, including:

[0007] Extracting points of interest from a target building, where the points of interest are places that affect the functional type of the target building;

[0008] Constructing graph structure data of the points of interest;

[0009] Extracting the level feature, type feature, and vertical spatial position feature of the points of interest, where the vertical spatial position feature is used to indicate the floor where the point of interest is located;

[0010] Extracting the spatial form feature of the target building;

[0011] Fusing the level feature, the spatial form feature, the type feature, and the vertical spatial position feature to obtain a point-of-interest feature matrix; and

[0012] Based on a preset weighted graph isomorphism network model, using the graph structure data and the point-of-interest feature matrix as inputs, outputting the functional type of the target building.

[0013] Further, extract points of interest from within the target building, including:

[0014] Obtain the name text of each location within the target building;

[0015] Based on a preset model, perform semantic analysis on the name text to obtain a text feature vector; and

[0016] Based on a deep pyramid convolutional network model, classify the text feature vector to obtain the type of the point of interest.

[0017] Further, before obtaining the name text of each location within the target building, the method further includes:

[0018] Set a buffer on the outer side of the target building according to preset buffer parameters; and

[0019] Regard the buffer as the interior of the target building.

[0020] Further, the method further includes:

[0021] Preset multiple selectable function types; and

[0022] Based on a preset weighted graph isomorphism network model, using the graph structure data and the point of interest feature matrix as inputs, output the function type of the target building, including:

[0023] Based on the weighted graph isomorphism network model, determine the functional characteristics of each point of interest according to the graph structure data and the point of interest feature matrix;

[0024] Based on the graph structure data, aggregate the functional characteristics of each point of interest to obtain the graph structure functional characteristics of the target building, and classify according to the functional characteristics to obtain the scores corresponding to each selectable function type;

[0025] Determine that the selectable function type with the maximum score is the function type of the target building.

[0026] Further, determining the functional characteristics of each point of interest includes:

[0027] Use a weighted aggregator to determine the functional characteristics of each point of interest, and the weighted aggregator is obtained by weighting a sum aggregator and a max aggregator.

[0028] Further, the weighted aggregator is specifically:

[0029]

[0030] where MLP is a multi-layer perceptron that can fit any function, h i(k) is the feature of node i after k iterations, h i (k -1) is the feature of node i after k-1 iterations, h u (k-1) is the feature of the u-th neighbor node of node i after k-1 iterations, ∈ is a learnable parameter, α and β are the corresponding weight values of the sum aggregator and the max aggregator respectively, and N(i) is the set of neighbor nodes of node i.

[0031] Furthermore, aggregating the functional features of each of the said points of interest includes:

[0032] Aggregating the functional features of each of the said points of interest based on a graph pooling function.

[0033] Furthermore, specifically aggregating the functional features of each of the said points of interest is:

[0034]

[0035] h G is the aggregated feature, Readout is the graph pooling function, h i (k) is the feature of node i after k iterations, k is the number of iterations, K is the total number of iterations, and G is the selectable function type.

[0036] In a second aspect, one or more embodiments of the present application provide a building function recognition device based on points of interest in a building, including:

[0037] An extraction module, configured to extract points of interest from a target building, where the points of interest correspond to places that affect the function type of the target building; and extract the spatial form characteristics of the target building;

[0038] A data processing module, configured to construct graph structure data of the points of interest; extract the level feature, type feature, and vertical spatial position feature of the points of interest, where the vertical spatial position feature is used to indicate the floor where the point of interest is located;

[0039] A feature fusion module, configured to fuse the level feature, the spatial form feature, the type feature, and the vertical spatial position feature to obtain a point of interest feature matrix; and

[0040] A classification module, configured to output the function type of the target building based on a preset weighted graph isomorphism network model, with the graph structure data and the point of interest feature matrix as inputs.

[0041] In a third aspect, an embodiment of the present application provides a storage medium for storing computer-executable instructions, characterized in that the computer-executable instructions, when executed, implement the steps of the method for identifying building functions based on points of interest in a building according to any one of the first aspect.

[0042] Compared with the prior art, the present application can at least achieve the following technical effects:

[0043] Regarding the places distributed on the building as points of interest, and then extracting their type features and vertical spatial position features for the points of interest. It can be seen that the above method increases the spatial and semantic features of the points of interest on the building, thereby increasing the accuracy of subsequent identification of building types. In addition, since naturally distributed buildings are inherently sample-imbalanced, the present application, based on the method of feature aggregation, constructs the graph features of the building by using the features of each point of interest and the spatial features of the building to increase the number of features used for determining the functional type of the building, thereby improving the accuracy of identifying the building type, and effectively avoiding the problem that the model trained by using the building as a node to construct a graph cannot be applied to the actual situation. The method adopted by the present application makes the identification effects of various building functions more balanced while conforming to the natural distribution law of buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in one or more embodiments of the present application or in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 It is a flowchart of a method for identifying building functions based on points of interest in a building provided by one or more embodiments of the present application;

[0046] Figure 2 It is a schematic structural diagram of a device for identifying building functions based on points of interest in a building provided by one or more embodiments of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of the present application, the following will clearly and completely describe the technical solutions in one or more embodiments of the present application with reference to the drawings in one or more embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on one or more embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this document.

[0048] Naturally distributed buildings usually have the problem of sample imbalance. For example, residential buildings are significantly more than commercial buildings. This will lead to a much higher accuracy in identifying residential buildings than commercial buildings when identifying buildings subsequently. Although it is possible to adjust the ratio of the two to be equal in the training samples to solve the problem of the large difference in accuracy between the two. However, this approach makes the ratio of the two far from the actual situation, ultimately resulting in the trained model not meeting the actual needs. That is, the existing technology cannot balance the actual needs and the model design needs.

[0049] In view of the above scenarios and problems, the embodiments of the present application provide a method for identifying building functions based on points of interest in a building, as Figure 1 shown, including the following steps:

[0050] Step 1: Extract points of interest from the target building.

[0051] In the embodiments of the present application, there are usually many places in a building. For example, there are many shops in a building, and these shops are collectively referred to as places. A point of interest is a place that affects the functional type of the target building. For example, when a building includes educational or cultural places, then the building is likely to be of the science, education, and culture type. That is, educational or cultural places will affect the functional type of the building. By converting the places in the building into points of interest, it lays a foundation for increasing the number of building features subsequently.

[0052] Step 2: Construct graph structure data of the points of interest.

[0053] In the embodiments of the present application, based on the distances between points of interest, a connection relationship between points of interest is generated through the Delaunay triangulation method to construct graph structure data.

[0054] Step 3: Extract the level feature, type feature, and vertical spatial position feature of the points of interest.

[0055] In the embodiments of the present application, the type features of the points of interest include: commercial, science, education, and culture, office, medical, and residential. Among them, the commercial category includes: food, shopping, life services, entertainment, sports and fitness, automobiles, etc.; science, education, and culture include: cultural venues, educational schools, etc.; office includes: companies, institutions, banks and finance, etc.; medical includes: general hospitals, specialized hospitals, emergency centers, health centers, community hospitals, and disease treatment, etc.; residential includes: residential communities, villas, dormitories, etc. It should be noted that the specific content of the above type features is only an optimization solution of the present application and cannot be regarded as a limitation of the type features.

[0056] The vertical spatial location feature is used to indicate the floor where the point of interest is located. In the actual scenario, the building function type is not only related to the type of point of interest and the combination mode of points of interest. It is also related to the floor information of the points of interest. For example, the commercial places distributed on residential buildings are mainly ground-floor stores, that is, most of them are on the first floor and generally do not contain floor information (there are also some commercial tenants on the second floor of some residential buildings), while for buildings with a commercial function type, there will be commercial places on each floor.

[0057] When extracting the vertical spatial location feature, if there is no floor number information, it is 0, and if there is floor number information, the floor feature is the number of floors.

[0058] The method for identifying the vertical spatial location feature mainly includes three parts;

[0059] ① Extract address element feature words: Analyze the combination mode of floor elements in the point of interest address, and extract floor information feature words. Among them, "layer", "building", and "F" are the three main feature words of floor elements in the address.

[0060] ② Establish extraction rules: Floor elements are mainly formed by the combination of numbers and feature words. The following table shows 5 common combination modes. Generally, they can be divided into the combination mode with "layer / building" as the feature word, the combination mode with "F" as the feature word, and the combination mode where both types of feature words appear at the same time. Among them, the combination mode with "layer / building" as the feature word includes combination 1 and 2. In combination 2, a "-", "negative", or "underground" is added in front of combination 1 to indicate that the entity is distributed in the underground space; the combination rules with "F" as the feature word include combination 4 and 5. In combination 4, the feature word is in front and the number is behind, and in combination 5, it is the opposite. Both of these methods are relatively common; combination 3 is the case where both types of feature words appear at the same time, and the general combination mode is "F + number + layer / building".

[0061] ③ Extract floor elements: Extract the floor information from the address according to the combination rules to obtain the floor number where the entity is located.

[0062] To distinguish the contributions of different types of points of interest (POIs) to the functions of buildings, the concept of level weights is proposed. Specifically, POIs are classified into primary functional POIs, auxiliary functional POIs, and dual-use POIs to grade the POIs (as shown in the following table). Primary functional POIs include 9 categories: housing estates, educational institutions, cultural venues, general hospitals, specialized hospitals, emergency centers, health centers, community hospitals, and disease prevention and control. These POIs can usually undertake the main functions of buildings; Auxiliary functional POIs include 7 categories: entertainment and leisure, food, cars, life services, shopping, hotels, and sports and fitness. These POIs usually appear as the ground-floor shops of buildings or the peripheral supporting entities of primary functional entities and are usually not the main functions of buildings (excluding the case of commercial complexes, which are usually composed of many auxiliary functional entities); Dual-use POIs include 3 categories: commercial offices, government agencies, and social organizations. They may either be the main entities undertaking the functions of buildings or one of the multiple entities distributed on buildings.

[0063] To obtain the level feature representation of each POI, the One-hot encoding form is also used to represent the POI level features. For POI i, its type feature vector is represented as:

[0064]

[0065] where NL represents the number of POI levels. Taking residential POIs as an example, its level feature vector is represented as:

[0066]

[0067] Step 4: Extract the spatial form features of the target building.

[0068] In the embodiments of this application, the spatial form features include: the number of building sides, elongation rate, compactness, perimeter, area, total number of floors, direction, and the shortest distance between the building and the main road, etc.

[0069] Among them, the compactness calculation formula is:

[0070] IPQ = 4πS / P 2

[0071] S represents the building area, and P represents the building;

[0072] The elongation rate calculation formula is:

[0073] E = L l / L s

[0074] L l represents the long side length of the minimum circumscribed rectangle of the building polygon, and L sRepresents the short side length of the minimum bounding rectangle of the building polygon.

[0075] The calculation formula for the building direction is:

[0076]

[0077] (x1, x2), (x2, y2) and (x3, y3) are vertex coordinates, and L1 and L2 are the side lengths of the minimum bounding rectangle.

[0078] Step 5: Integrate the level feature, spatial form feature, type feature, and vertical spatial position feature to obtain the point-of-interest feature matrix.

[0079] In the embodiment of the present application, since there are multiple points of interest, the type feature and the vertical spatial position feature are multi-dimensional features. And the spatial form feature belongs to the building, so the spatial form feature belongs to a one-dimensional feature. When the three are integrated, the spatial form feature needs to be converted into the corresponding multi-dimensional feature. For example, if the spatial form feature is (S1, S2,..., Sn) and there are 3 points of interest, then the spatial form feature is converted to:

[0080] {S1, S2,..., Sn;

[0081] S1, S2,..., Sn;

[0082] S1, S2,..., Sn.}

[0083] Step 6: Based on the preset weighted graph isomorphism network model, use the graph structure data and the point-of-interest feature matrix as inputs to output the functional type of the target building.

[0084] In the embodiment of the present application, the weighted graph isomorphism network model includes a model input layer, a node embedding layer, a graph embedding layer, and a graph classification layer.

[0085] It can be seen that the present application adds the spatial form feature, the type feature, the vertical spatial position feature, and the graph structure data, so the recognition accuracy of each functional type can be improved.

[0086] In the embodiment of the present application, the specific process of extracting the point of interest is as follows:

[0087] A1. Obtain the name text of each place in the target building.

[0088] In the embodiment of the present application, the name text includes the name, type, and address information of the place.

[0089] A2. Based on the preset model, perform semantic analysis on the name text to obtain the text feature vector.

[0090] In the embodiments of the present application, before semantic analysis, preprocessing such as character normalization, stop word removal, and text tokenization needs to be performed on the name text, and then a BERT model is used to generate the name text feature vector.

[0091] A3. Based on the deep pyramid convolutional network model, classify the text feature vector to obtain the type of the point of interest.

[0092] In the embodiments of the present application, the types of points of interest are preset in advance, and then the trained deep pyramid convolutional network model is used to map the name text to the corresponding classification.

[0093] In the embodiments of the present application, in order to ensure that no point of interest is missed, a buffer is set outside the target building according to the preset buffer parameters; the buffer is regarded as the inside of the target building. For example, a 5-meter buffer is set for the building, and the points of interest falling within the building buffer area are screened.

[0094] In the embodiments of the present application, the specific process of outputting the function type of the target building is as follows:

[0095] B1. Preset multiple selectable function types in advance.

[0096] In the embodiments of the present application, the selectable function types include: commercial, science and education culture, residential, office, medical, and mixed.

[0097] B2. Based on the weighted graph isomorphism network model, determine the function features of each point of interest according to the graph structure data and the point of interest feature matrix.

[0098] In the embodiments of the present application, a weighted aggregator is used to determine the function features of each point of interest, and the weighted aggregator is obtained by weighting the sum aggregator and the max aggregator.

[0099] Specifically, for the function features of the point of interest, to obtain the feature representation of the point of interest - building entity graph, first, the feature representation of the point of interest needs to be generated. The features of each point of interest are not only affected by its own factors but also related to the features of adjacent objects. For example, the types of shops on a commercial building are usually highly related to the types of neighboring shops, and catering shops are usually distributed adjacently. Therefore, to calculate the feature representation of the point of interest, the features of the current point of interest and its neighboring points of interest need to be considered simultaneously. The graph isomorphism model follows the neighborhood aggregation scheme. According to this idea, the feature vector of the current point of interest object can be obtained by aggregating the attribute features of neighboring point of interest nodes and merging the aggregated neighboring features with the current node features.

[0100] Currently, the commonly used aggregation functions include: Sum, Mean, and Max. Among them, the mean aggregator (Mean) can capture the feature distribution of entities, but cannot identify the number of occurrences of a certain element, and is suitable for tasks that only care about whether an element exists or not, rather than how many there are; the max aggregator (Max) is suitable for capturing representative elements, but not for tasks that require distinguishing exact structures or distributions. The sum aggregator (Sum) can identify the number of occurrences of elements and is suitable for tasks that require distinguishing exact distributions.

[0101] For the building function classification task in this application, the number of distributed entities on the building is also one of the important features for distinguishing building functions. For example, the number of point-of-interest data distributed on commercial and mixed-function buildings is significantly more than that on residential and educational buildings. Therefore, the building function classification task is a task that requires identifying the number of occurrences of elements, and the Sum aggregator is suitable. In addition, for the nodes that the Mean aggregator can distinguish, the Sum aggregator can also distinguish them; for the nodes that the Mean aggregator cannot distinguish, the Sum aggregator can also distinguish them. However, for the nodes that the Max aggregator can distinguish, the Sum aggregator may not necessarily be able to distinguish them. Therefore, considering the above characteristics, a graph isomorphism network with a weighted aggregator that comprehensively considers the Sum and Max methods is proposed to identify the building graph structure, and its node update is expressed as:

[0102]

[0103] where MLP is a multi-layer perceptron that can fit any function, h i (k) is the feature of node i after k iterations, h i (k -1) is the feature of node i after k - 1 iterations, h u (k-1) is the feature of the u-th neighbor node of node i after the (k - 1)-th iteration, ∈ is a learnable parameter, α and β are the corresponding weight values of the sum aggregator and the max aggregator respectively, and N(i) is the set of neighbor nodes of node i.

[0104] B3. Based on the graph-structured data, aggregate the functional features of each point of interest to obtain the functional features of the target building graph structure, and classify according to the functional features to obtain the scores corresponding to each of the selectable functional types.

[0105] In the embodiments of the present application, the functional features of each of the above-mentioned points of interest are aggregated based on a graph pooling function. Specifically, since the building function type is not affected by the number of points of interest on the building. For example, Building 1 has 1 point of interest of the residential type and 1 of the commercial type, and Building 2 has 1 residential type and 3 commercial types. Although the number of points of interest distributed on the two buildings is different, they both belong to residential buildings. Therefore, in order to avoid the influence of the number of points of interest on the recognition result of the function type, the Mean method is used instead of the Sum method to integrate the point-of-interest objects distributed on the building. After aggregating the node features obtained in each iteration into graph features through the Mean method, the graph features of all iterations are summed to obtain the final building graph features. The Mean method is specifically as follows:

[0106]

[0107] hG is the aggregated feature, Readout is the graph pooling function, h i (k) is the feature of node i after k iterations, k is the number of iterations, K is the total number of iterations, and G is the selectable function type.

[0108] B4. Determine the selectable function type with the highest score as the function type of the target building.

[0109] In the embodiments of the present application, after aggregating the functional features of each point of interest, the functional features of the building graph structure are obtained, and the functional features of the building graph structure are input into a fully connected layer to calculate the scores of each category

[0110] y = {yi|i ∈ {residential, medical, education, office, commercial, mixed}}

[0111] Then, the SoftMax function normalizes the scores of each category obtained from the fully connected layer to obtain the scores of each category after normalization, and the building function category corresponding to the selectable function type with the highest score.

[0112] To illustrate the feasibility of the technical solutions described in the above embodiments, the technical solutions provided in the present application are compared and analyzed with GCN (Graph Neural Network) and GraphSAGE (Graph Sample and Aggregated). Among them, GCN and GraphSAGE construct a building group graph structure with buildings as nodes). As shown in Table 1, the experimental results show that the method proposed in the present application has a more stable classification effect on various types of buildings compared with other methods, and can effectively alleviate the influence of the sample imbalance problem on the model. And the overall classification accuracy of the model reaches 82.27%, the Macro-F1 value is 74.57, and the Kappa coefficient is 0.716, which is significantly improved compared with other solutions.

[0113] Table 1 Performance Comparison among the Present Application, GCN, and GraphSAGE

[0114]

[0115] An embodiment of the present application provides a building function recognition device based on points of interest in a building, including:

[0116] An extraction module 201, configured to extract points of interest from a target building, where the points of interest correspond to places that affect the function type of the target building; and extract the spatial form features of the target building;

[0117] A data processing module 202, configured to construct graph structure data of the points of interest; and extract the level features, type features, and vertical spatial position features of the points of interest, where the vertical spatial position features are used to indicate the floor where the points of interest are located;

[0118] A feature fusion module 203, configured to fuse the level features, the spatial form features, the type features, and the vertical spatial position features to obtain a point of interest feature matrix; and

[0119] A classification module 204, configured to, based on a preset weighted graph isomorphism network model, use the graph structure data and the point of interest feature matrix as inputs and output the function type of the target building.

[0120] An embodiment of the present application provides a storage medium for storing computer-executable instructions, characterized in that the computer-executable instructions, when executed, implement the steps of the building function recognition method based on points of interest in a building according to any one of the embodiments.

[0121] It should be noted that the embodiment of the storage medium in the present application and the embodiment of the method for providing blockchain-based services in the present application are based on the same inventive concept. Therefore, the specific implementation of this embodiment can refer to the corresponding implementation of the method for providing blockchain-based services described above, and repeated parts will not be elaborated.

[0122] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve the desired results. Additionally, the processes depicted in the figures do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] In the 1930s, it was obvious to distinguish whether an improvement in a technology was an improvement in hardware (e.g., improvement in circuit structures such as diodes, transistors, switches, etc.) or an improvement in software (improvement in method flows). However, with the development of technology, many improvements in method flows today can be regarded as direct improvements in hardware circuit structures. Almost all designers obtain the corresponding hardware circuit structures by programming the improved method flows into the hardware circuits. Therefore, it cannot be said that an improvement in a method flow cannot be implemented with a hardware entity module. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logical function is determined by the user programming the device. The designer can program by himself / herself to "integrate" a digital system on a piece of PLD, without having to ask a chip manufacturer to design and fabricate a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly implemented using "logic compiler" software, which is similar to the software compiler used in program development and writing. The original code before compilation also has to be written in a specific programming language, which is called Hardware Description Language (HDL). And there is not only one kind of HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones currently are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also be clear that only by slightly logically programming the method flow with the above-mentioned several hardware description languages and programming it into the integrated circuit can the hardware circuit implementing the logical method flow be easily obtained.

[0124] The controller can be implemented in any suitable manner. For example, the controller can take the form of, for example, a microprocessor or a processor and a computer-readable medium storing computer-readable program code (such as software or firmware) executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of the controller include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in the form of pure computer-readable program code, it is entirely possible to logically program the method steps to enable the controller to be implemented in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, and embedded microcontrollers, etc. to achieve the same function. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component. Or even, the devices for implementing various functions can be regarded as either software modules for implementing the method or structures within the hardware component.

[0125] The systems, devices, modules, or units illustrated in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0126] For the convenience of description, when describing the above devices, they are described separately as various units according to their functions. Of course, when implementing the embodiments of the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0127] Those skilled in the art should understand that one or more embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, one or more embodiments of the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0128] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, as well as the combination of flows and / or blocks in the flowchart and / or block diagram. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0129] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0130] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0131] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0132] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0133] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0134] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0135] One or more embodiments of the present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of the present application may also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0136] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0137] The above are only examples of this document and are not intended to limit this document. For those skilled in the art, this document may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this document shall be included within the scope of the claims of this document.

Claims

1. A method for identifying building functions based on points of interest in a building, characterized in that: include: Extracting points of interest from within a target building, wherein the points of interest are places that affect the functional type of the target building; Constructing graph structure data of the points of interest; Extracting a level feature, a type feature, and a vertical spatial position feature of the point of interest, wherein the vertical spatial position feature is used to indicate the floor where the point of interest is located; Extracting spatial morphological features of the target building; The level feature, the spatial morphology feature, the type feature and the vertical spatial position feature are integrated to obtain an interest point feature matrix; as well as Based on a preset weighted graph isomorphic network model, the graph structure data and the interest point feature matrix are used as input to output the functional type of the target building.

2. The method according to claim 1, characterized in that Extract points of interest from the target building, including: Obtain the name text of each place in the target building; Based on a preset model, semantic analysis is performed on the name text to obtain a text feature vector; and Based on the deep pyramid convolutional network model, the text feature vector is classified to obtain the type of interest point.

3. The method according to claim 2, characterized in that Before obtaining the name text of each place in the target building, the method further includes: According to preset buffer zone parameters, a buffer zone is set outside the target building; and The buffer zone is considered to be inside the target building.

4. The method according to claim 1, characterized in that: The method further comprises: Preset multiple selectable function types; and Based on a preset weighted graph isomorphic network model, the graph structure data and the interest point feature matrix are used as input to output the function type of the target building, including: Based on the weighted graph isomorphic network model, determining the functional characteristics of each of the points of interest according to the graph structure data and the point of interest feature matrix; Based on the graph structure data, the functional features of each of the points of interest are aggregated to obtain functional features corresponding to the building graph structure, and scores corresponding to each of the selectable functional types are obtained according to the functional feature classification; The selectable function type with the largest score is determined as the function type of the target building.

5. The method according to claim 4, characterized in that Determining the functional characteristics of each of the points of interest, including: A weighted aggregator is used to determine the functional characteristics of each of the interest points, wherein the weighted aggregator is obtained by weighting a sum aggregator and a maximum aggregator.

6. The method according to claim 5, characterized in that The weighted aggregator is specifically: Among them, MLP is a multi-layer perceptron, which is used to fit any function, h i (k) is the feature of node i after k iterations, h i (k-1) is the feature of node i after k-1 iterations, h u (k-1) is the feature of the u-th neighbor node of node i after the k-1th iteration, ∈ is a learnable parameter, α, β are the corresponding weight values ​​of the sum aggregator and the maximum aggregator, respectively, and N(i) is the set of neighbor nodes of node i.

7. The method according to claim 4, characterized in that Aggregate the functional features of each of the points of interest, including: The functional features of each of the interest points are aggregated based on a graph pooling function.

8. The method according to claim 7, characterized in that The functional features of aggregating the above points of interest are specifically: h G is the aggregated feature, Readout is the graph pooling function, and h i (k) is the feature of node i after k iterations, k is the number of iterations, K is the total number of iterations, and G is the selectable function type.

9. A building function identification device based on points of interest in a building, characterized in that: include: An extraction module is used to extract points of interest from within a target building, wherein the points of interest correspond to places that affect the functional type of the target building; and extract spatial morphological features of the target building; A data processing module, used to construct the graph structure data of the interest point; extract the level feature, type feature and vertical spatial position feature of the interest point, wherein the vertical spatial position feature is used to indicate the floor where the interest point is located; A feature fusion module, used to fuse the level feature, the spatial morphology feature, the type feature and the vertical spatial position feature to obtain an interest point feature matrix; as well as The classification module is used to output the functional type of the target building based on a preset weighted graph isomorphic network model and taking the graph structure data and the interest point feature matrix as input.

10. A storage medium for storing computer executable instructions, characterized in that: When the computer executable instructions are executed, the steps of the method for identifying building functions based on points of interest in a building according to any one of claims 1 to 8 are implemented.

Images