Parking information integration method, system and electronic equipment

Abstract

The embodiment of the present application provides a parking information integration method, system and electronic device, the method comprising: collecting parking scene information of various data types; pre-processing the collected parking scene information of various data types in a unified format to obtain target parking data; extracting and fusing entity relationships of the target parking data through a knowledge graph model to determine each triple; and constructing a corresponding parking knowledge graph by merging knowledge of each triple. The purpose is to integrate parking information in urban traffic to improve the convenience and satisfaction of users in parking scenarios.

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, and in particular to a parking information integration method, system and electronic equipment. Background Art

[0002] At present, most smart car parking systems mainly focus on realizing basic parking functions, such as finding empty spaces, automatic or assisted parking, etc. These smart car parking systems often ignore the comprehensive collection and recording of key information in the parking environment. The opacity and incompleteness of these key information seriously affect the convenience and satisfaction of users. Summary of the invention

[0003] In view of this, the embodiments of the present application provide a parking information integration method, system and electronic device, which are intended to integrate parking information in urban traffic to improve the convenience and satisfaction of users in parking scenarios.

[0004] A first aspect of an embodiment of the present application provides a parking information integration method, the method comprising:

[0005] Collect parking scene information of various data types;

[0006] Preprocessing the collected parking scene information of various data types in a unified format to obtain target parking data;

[0007] Extracting and fusing entity relationships of the target parking data through a knowledge graph model to determine each triple;

[0008] By merging the knowledge of the triples, a corresponding parking knowledge graph is constructed.

[0009] Optionally, before extracting and fusing entity relationships of the target parking data through a knowledge graph model to determine each triple, the method further includes:

[0010] Inputting the target parking data into a pre-trained model for feature extraction to obtain target features;

[0011] The entity relationship extraction and fusion of the target parking data are performed by the knowledge graph model to determine each triple, including:

[0012] The target features are subjected to entity relationship extraction and fusion through the knowledge graph model to determine each triple.

[0013] Optionally, entity relationships of the target features are extracted and fused through a knowledge graph model to determine each triple, including:

[0014] Inputting the target feature into the knowledge graph model;

[0015] The target features are weighted by the weighted multi-head attention mechanism in the knowledge graph model to obtain weighted features;

[0016] The entity-relationship fusion sub-model in the knowledge graph model is used to extract entities and relationships from the weighted features to obtain each entity and the relationship between each entity;

[0017] The entity relationship fusion sub-model in the knowledge graph model is used to perform knowledge fusion on the entities and the relationships between the entities to obtain the triples.

[0018] Optionally, the entity relationship fusion sub-model in the knowledge graph model is used to perform knowledge fusion on the entities and the relationships between the entities to obtain the triples, including:

[0019] Determining the matching scores between the extracted entities and the inter-entity relations through the scoring function in the entity relation fusion sub-model;

[0020] The entities and inter-entity relationships whose matching scores meet the set conditions are determined as target entities and inter-target entity relationships to construct corresponding triples.

[0021] Optionally, when the data types include image data type, point cloud data type, text data type, digital data type, and video data type, preprocessing the collected parking scene information of the various data types in a unified format to obtain target parking data includes:

[0022] Determine the data type of the collected parking scene information;

[0023] Determining, according to the data type of the parking scene information, a target processing algorithm corresponding to the data type of the parking scene information;

[0024] The parking scene information is pre-processed in a unified format by the target processing algorithm to obtain target parking data.

[0025] Optionally, the method further includes:

[0026] When new target parking data is obtained, entity relationship extraction and fusion are performed on the new target parking data through the knowledge graph model to determine triples to be updated;

[0027] The parking knowledge graph is updated using the triples to be updated.

[0028] Optionally, updating the parking knowledge graph using the triples to be updated includes:

[0029] Determining whether there is a target triplet corresponding to the triplet to be updated in the parking knowledge graph;

[0030] In the case where the target triplet exists, comparing the triplet to be updated with the target triplet to obtain a corresponding comparison result;

[0031] According to the comparison result, replacing the target triplet with the triplet to be updated;

[0032] In the case that the target triple does not exist, the triple to be updated is updated into the parking knowledge graph through knowledge merging.

[0033] Optionally, the method further includes:

[0034] Receiving a parking information acquisition request from a terminal;

[0035] In response to the parking information acquisition request, acquiring target parking information corresponding to the parking information acquisition request in the parking knowledge graph;

[0036] The target parking information is sent to the terminal for display.

[0037] A second aspect of an embodiment of the present application provides a parking information integration system, the system comprising:

[0038] A data collection module, used to collect parking scene information of various data types;

[0039] A preprocessing module, used for preprocessing the collected parking scene information of various data types in a unified format to obtain target parking data;

[0040] An extraction and fusion module, used to extract and fuse entity relationships of the target parking data through a knowledge graph model to determine each triple;

[0041] The knowledge merging module is used to construct a corresponding parking knowledge graph by merging the knowledge of each triple.

[0042] A third aspect of an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, a parking information integration method as described in the first aspect of an embodiment of the present application is implemented.

[0043] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, a parking information integration method as described in the first aspect of an embodiment of the present application is implemented.

[0044] The parking information integration method provided by this application has the following advantages:

[0045] The parking information integration method provided in the embodiment of the present application first collects parking scene information of various data types; performs preprocessing of the collected parking scene information of various data types in a unified format to obtain target parking data; extracts and fuses entity relationships of the target parking data through a knowledge graph model to determine each triple; and constructs a corresponding parking knowledge graph by performing knowledge merging on each triple. Therefore, the present application integrates parking scene information of various data types through a knowledge graph to construct a more comprehensive parking knowledge graph in urban transportation, which is used to improve the convenience and satisfaction of users in parking scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor.

[0047] Figure 1 A flowchart of a parking information integration method shown in one embodiment of the present application;

[0048] Figure 2 An exemplary schematic diagram of a parking knowledge graph obtained by integrating a parking information integration method according to an embodiment of the present application;

[0049] Figure 3 A structural diagram of a knowledge graph model in a parking information integration method according to an embodiment of the present application;

[0050] Figure 4 A flow chart of a target processing algorithm corresponding to image data types in a parking information integration method according to an embodiment of the present application;

[0051] Figure 5 A flow chart of a target processing algorithm corresponding to a point cloud data type in a parking information integration method shown in one embodiment of the present application;

[0052] Figure 6 A flow chart of a target processing algorithm corresponding to text data types in a parking information integration method according to an embodiment of the present application;

[0053] Figure 7 A flow chart of a target processing algorithm corresponding to a digital data type in a parking information integration method according to an embodiment of the present application;

[0054] Figure 8 A flow chart of a target processing algorithm corresponding to a video data type in a parking information integration method according to an embodiment of the present application;

[0055] Fig. 9 A schematic diagram of a parking information integration method according to an embodiment of the present application;

[0056] Fig.10 A schematic diagram of a parking information integration system according to an embodiment of the present application. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0058] refer to Figure 1 , Figure 1 The following is a flow chart of a parking information integration method according to an embodiment of the present application. Figure 1 As shown, the parking information integration method of this embodiment includes:

[0059] Step S1: Collect parking scene information of various data types.

[0060] In this embodiment, in the process of constructing the parking knowledge graph, in order to ensure the accuracy and completeness of the parking knowledge graph, this application adopts multiple data sources and multiple data types to collect parking scene information. Parking scene information includes but is not limited to parking lot data, parking space data, parking lot internal facilities and rules data, and parking lot surrounding facilities data. Parking lot data includes but is not limited to name, address, capacity, opening hours, contact information, and price; parking space data includes but is not limited to parking space number, location of parking space in the parking lot, and parking space type (vertical parking space, horizontal parking space, charging parking space); parking lot internal facilities and rules data include but are not limited to internal facilities (restrooms, charging piles, automatic payment machines), sign information, and driving rules; parking lot surrounding facilities data include but are not limited to shops, restaurants, hotels, and gas stations.

[0061] In this embodiment, the multiple data sources at least include data collection of parking scene information from public databases, specifically by accessing public databases of local government transportation or urban planning departments, which usually contain basic parking lot information, such as name, address, capacity and other parking scene information; and data collection of parking scene information from online map services, which can be used to obtain information such as the geographical location, opening hours and user evaluation of parking lots; and establishing cooperation with parking lots to directly obtain internal data of parking lots to collect parking scene information. On the one hand, a cooperative relationship is established with parking lot management system providers to directly obtain parking scene information. On the other hand, if the parking lot management system provides an API interface, data can be automatically obtained programmatically; and when the user turns on smart parking / memory parking, the parking scene information is collected through the sensors of the intelligent driving system and uploaded to the cloud after decryption. The controlled parking scene information includes: speed bumps, drivable areas, parking areas, lane lines (solid lines, dotted lines, stop lines, intersections, etc.), traffic signs (traffic lights, speed limit signs, signs, etc.), prohibited areas (fire passages, sidewalks, etc.), and parking lot information feedback from users, including the opening hours, charges, and facility conditions of the parking lot.

[0062] Step S2: pre-processing the collected parking scene information of various data types in a unified format to obtain target parking data.

[0063] In this embodiment, in order to ensure the accuracy and completeness of the constructed parking knowledge graph, the present application collects parking scene information through multiple data sources. Different data sources or the same data source provide different data types. Therefore, the collected parking scene information involves various data types, which makes it impossible to directly use these collected parking scene information to construct the parking knowledge graph. Therefore, after collecting a large amount of parking scene information of various data types through multiple data sources, the present application performs a unified format preprocessing on the collected parking scene information of various data types to convert the parking scene information of various data types into target parking data of a unified format. Among them, the unified data format is preferably JSON, CSV, XML data format, so as to facilitate the subsequent construction of a comprehensive parking knowledge graph in urban traffic scenarios based on these target parking data.

[0064] Step S3: extracting and fusing entity relationships of the target parking data through a knowledge graph model to determine each triple.

[0065] In this embodiment, the present application predefines various entities and relationships between entities required for constructing a parking knowledge graph, and precreates an initial deep learning model. Then, the initial deep learning model is trained based on the predefined various entities and relationships between entities, so that the trained initial deep learning model can have the ability to identify which entities need to be extracted and which relationships between entities need to be extracted, and has the ability to fuse the extracted entities and the relationships between entities to obtain various triples. After the initial deep learning model is trained, the qualified initial deep learning model is the knowledge graph model. The determined target parking data is input into the knowledge graph model, and the knowledge graph model will extract entities and relationships between entities from the target parking data, and fuse the extracted entities and relationships between entities to form various triples.

[0066] For example, for a piece of target parking data "Parking lot A has 500 parking spaces", the target parking data is processed by the knowledge graph model, and the knowledge graph model will extract the entities: parking lot A, parking spaces, 500; and the relationships between entities: ownership, quantity; then the knowledge graph model fuses the extracted entities and the relationships between entities to obtain the following triples (parking lot A, ownership, parking spaces), (parking spaces, quantity, 500).

[0067] In this embodiment, the entities predefined in this application include but are not limited to parking lots, parking lot price values, parking spaces, parking space quantity values, gas stations, charging piles, hotels, and shops. It should be understood that the number of entities in the parking scene is very large, and only some entities are exemplarily described here. The relationships between entities predefined in this application include but are not limited to the ownership relationship of parking lots, the ownership relationship of parking lots with internal facilities in parking lots, the quantitative relationship of the number of parking spaces in parking lots, and the relationship of the surrounding facilities around parking lots. It should be understood that the number of relationships between entities and entities in the parking scene is very large, and only some relationships are exemplarily described here. Among them, each entity has a unique identifier (such as ID or URI), which is at least used for the subsequent change of the data of the entity node in the parking knowledge graph, when the data of the entity node needs to be replaced, the entity node can be located in the parking knowledge graph based on the unique identifier, and the current data of the entity node is replaced with the changed data of the entity node. Among them, the relationship between entities has clear semantics and directionality.

[0068] Step S4: Construct a corresponding parking knowledge graph by merging the knowledge of each triple.

[0069] In this embodiment, after the target parking data is processed by the knowledge graph model to obtain a large number of triples, all triples are merged to obtain the corresponding parking knowledge graph, such as Figure 2 As shown, Figure 2 A schematic diagram of a simple parking knowledge graph is shown. When actually constructing a parking knowledge graph, more parking scene information will be involved, and the constructed parking knowledge graph will be more complex and large. Figure 2 A simple parking knowledge graph is only shown as an example to facilitate understanding of the solution.

[0070] The parking information integration method provided in the embodiment of the present application first collects parking scene information of various data types; performs preprocessing of the collected parking scene information of various data types in a unified format to obtain target parking data; extracts and fuses entity relationships of the target parking data through a knowledge graph model to determine each triple; and constructs a corresponding parking knowledge graph by performing knowledge merging on each triple. Therefore, the present application integrates parking scene information of various data types through a knowledge graph to construct a more comprehensive parking knowledge graph in urban transportation, which is used to improve the convenience and satisfaction of users in parking scenarios.

[0071] In combination with the above embodiments, in one implementation, the present application embodiment further provides a parking information integration method. In the parking information integration method, before step S3, the method further includes step S03:

[0072] Step S03: Input the target parking data into a pre-trained model for feature extraction to obtain target features.

[0073] In this embodiment, in order to enhance the effect of the knowledge graph model in extracting relevant entities and relationships between entities in the target parking data, the present application first extracts features of the target parking data through a pre-trained model to obtain target features corresponding to the target parking data.

[0074] In the present application, when the parking information integration method further includes step S03, step S3 may include: extracting and fusing entity relationships of the target features through a knowledge graph model to determine each triple.

[0075] In this embodiment, after the target features of the target parking data are determined in step S03, the target features are input into the knowledge graph model for entity extraction, and the relationships between entities are extracted, and the extracted entities and relationships are fused to obtain corresponding triples.

[0076] In combination with the above embodiments, in one implementation, the present application embodiment further provides a parking information integration method. In the parking information integration method, step S3 may include steps S31 to S34:

[0077] Step S31: input the target features into the knowledge graph model.

[0078] In this embodiment, the target features obtained in step S3 are input into the knowledge graph model to extract and fuse entities and relationships.

[0079] Step S32: weighting the target features through the weighted multi-head attention mechanism in the knowledge graph model to obtain weighted features.

[0080] In this embodiment, if Figure 3 As shown, Figure 3 The schematic diagram of the structure of the knowledge graph model in this application is shown. Figure 3 As shown, in the parking scenario, some entities (such as the name of the parking lot and the specific location of the parking space) are more important than other entities (such as surrounding facilities and business hours) in the context of the parking task. Therefore, the knowledge graph model constructed in this application includes a weighted multi-head attention mechanism to automatically assign different weights to each head, so that the knowledge graph model can better focus on important entities. The attention weights of different heads in the weighted multi-head attention mechanism are dynamically adjusted. Among them, the expression of the weighted multi-head attention mechanism is:

[0081]

[0082] Among them, H represents the number of heads of the weighted multi-head attention mechanism. More heads mean that the knowledge graph model can capture richer contextual information. During specific training, this parameter determines the ability of the knowledge graph model to process target features in parallel.

[0083] w h It represents the dynamic weight of each head in the weighted multi-head attention mechanism, which is obtained through network learning and training, and is used to reflect the importance of different heads in the current context. By adjusting the dynamic weights of each head, the knowledge graph model can more flexibly adapt to different input and task requirements. In the parking scenario, each entity will be automatically assigned a dynamic weight, so that the knowledge graph model pays more attention to the entities that users care about and the relationship between entities, such as the location and parking spaces of the parking lot.

[0084] head h (Q, K, V) represents the attention output calculated by each head. Each head has its own independent set of parameters and can learn different attention modes.

[0085] In this embodiment, the target feature data is first weighted by the weighted multi-head attention mechanism in the knowledge graph model, and the weighted multi-head attention mechanism outputs the weighted feature corresponding to the target feature after weighted processing.

[0086] In this embodiment, different entities and relationships have different importance in the parking scenario. For example, the name and location of the parking lot may be more important than the operating hours of the parking space. Therefore, this application constructs a corresponding weighted cross entropy loss function for the knowledge graph model, and assigns different weights to different entity and relationship categories during the training of the knowledge graph model. The expression of the weighted cross entropy loss function is as follows:

[0087]

[0088] Where N represents the number of word fragments in the target parking data; y i is the true label, indicating the entity category to which the word fragment actually belongs; Predict the probability of a parking entity category to which a word fragment belongs for the knowledge graph model; Represents the weight of the parking entity category. According to actual needs, this application assigns higher weights to important entity categories so that the knowledge graph model pays more attention to these important entity categories during the training process.

[0089] Step S33: extract entities and relationships from the weighted features through the entity-relationship fusion sub-model in the knowledge graph model to obtain each entity and the relationship between each entity.

[0090] In this embodiment, traditional entity extraction and relationship extraction between entities are usually two independent steps, while this application is aimed at parking scenarios, in which entities (such as parking lot name, address) and relationships between entities (such as the relationship between parking lot and parking spaces) are equally important and related. Therefore, this application uses the idea of ​​embedded representation in natural language processing to extract entities and relationships between entities at the same time. Specifically: the knowledge graph model constructed by the present invention includes an entity relationship fusion sub-model, which considers entities and relationships between entities at the same time, wherein the expression of the entity relationship fusion sub-model is:

[0091] P Entity-Relationship (e i ,e j ,r)=f(e i )·g(e j )·h(r)

[0092] Among them, e i , e jis the embedding vector representation of the entity (such as parking lot, parking space); r is the entity e i and entity e j The relationship between parking lots and parking spaces is represented by an embedded vector.

[0093] f(·), g(·), h(·) are learnable functions used to calculate the matching scores between entities and the relationships between entities. During the training of the knowledge graph model, by optimizing these learnable functions, the knowledge graph model can better understand the entities and the relationships between entities, thereby accurately extracting the entities and the relationships between entities and fusing them to obtain each triple.

[0094] In this embodiment, the obtained weighted features are used to extract entities through the word embedding model (such as Word2Vec, BERT, etc.) in the entity relationship fusion sub-model in the knowledge graph model, and then the extracted entities are represented by embedded vectors. The embedded vector representations of these entities can record the semantic information and contextual relationships of the entities. At the same time, the obtained weighted features are used to extract the relationship between entities through the word embedding model (such as Word2Vec, BERT, etc.) in the entity relationship fusion sub-model in the knowledge graph model, and then the extracted relationship between entities is represented by embedded vectors. The embedded vector representations between these entities can record the semantic information and contextual relationships between entities.

[0095] Step S34: Perform knowledge fusion on the entities and the relationships between the entities through the entity relationship fusion sub-model in the knowledge graph model to obtain each triple.

[0096] In this embodiment, the entity relationship fusion sub-model in the knowledge graph model is used to further perform knowledge fusion on the extracted entities and the relationships between entities to obtain various triples.

[0097] In combination with the above embodiments, in one implementation, the present application embodiment further provides a parking information integration method. In the parking information integration method, step S34 may include steps S341 to S342:

[0098] Step S341: Determine the matching score between the extracted entities and the inter-entity relationships through the scoring function in the entity relationship fusion sub-model.

[0099] In this embodiment, the scoring function in the entity-relationship fusion sub-model is used to calculate the matching scores of the extracted entities and the relationships between entities. Specifically, a large number of entities represented by embedded vectors and a large number of relationships between entities represented by embedded vectors are extracted through the knowledge graph model, and then two uncombined entities and one relationship are taken out from all the extracted entities and all the relationships each time, and the matching scores between the two entities and the one relationship are calculated through the scoring function, that is, the above-mentioned learnable functions f(·), g(·), h(·), thereby obtaining a large number of matching scores. During the iterative training process of the knowledge graph model, the knowledge graph model will continuously adjust the parameters of the above three scoring functions according to a large number of training samples to optimize the above three scoring functions so that the calculated matching scores between entities and relationships are more accurate.

[0100] For example, assuming that 4 entities A1, A2, A3, A4 and 2 relationships a1 and a2 between entities are extracted, the entities A1, A2 and relationship a1 are taken out for the first time to calculate the matching score, and the matching score x1 is obtained; the entities A1, A2 and relationship a2 are taken out for the second time to calculate the matching score, and the matching score x2 is obtained; the entities A1, A3 and relationship a1 are taken out for the third time to calculate the matching score, and the matching score x3 is obtained; the entities A1, A3 and relationship a2 are taken out for the fourth time to calculate the matching score, and the matching score x4 is obtained; the entities A1, A4 and relationship a1 are taken out for the fifth time to calculate the matching score, and the matching score x5 is obtained; the entities A1, A4 and relationship a2 are taken out for the sixth time to calculate the matching score. The matching score is calculated for the seventh time, and the matching score x6 is obtained; the matching score x7 is obtained by taking out entities A2, A3 and relationship a1 for the seventh time; the matching score x8 is obtained by taking out entities A2, A3 and relationship a2 for the eighth time; the matching score x9 is obtained by taking out entities A2, A4 and relationship a1 for the ninth time; the matching score x10 is obtained by taking out entities A2, A4 and relationship a2 for the tenth time; the matching score x11 is obtained by taking out entities A3, A4 and relationship a1 for the eleventh time; the matching score x12 is obtained by taking out entities A3, A4 and relationship a2 for the twelfth time, and 12 matching scores are obtained.

[0101] Step S342: The entities and inter-entity relationships whose matching scores satisfy the set conditions are determined as target entities and inter-target entity relationships to construct corresponding triples.

[0102] In this embodiment, after each matching score is calculated and obtained by step S341, each matching score is compared with a set threshold, and two entities and one relationship corresponding to the matching scores whose values ​​exceed the set threshold are determined as the target entity and the relationship between the target entities, and then the two entities and the one relationship are used to construct the corresponding triple. The set threshold can be set according to the actual application scenario and is not specifically limited here.

[0103] For example, continuing with the above example, it is determined that the matching scores x2 and x11 exceed the set threshold, so entities A1, A2 and relationship a2 are used to construct a corresponding triple, and entities A3, A4 and relationship a1 are used to construct a corresponding triple.

[0104] For example, suppose there is a data segment to be processed "Parking lot A" and "Parking lot B" can be extracted through the word embedding model in the knowledge graph model, and the entity "Parking lot A" and the entity "Parking lot B" can be represented as an embedded vector, and the relationship "belongs to" can be extracted and represented as an embedded vector. Then the scoring function in the knowledge graph model will calculate the matching score of the two entities and the one relationship, and the matching score between the embedding vector of "Parking lot A", the embedding vector of "Parking lot B" and the embedding vector of "belongs to" can be calculated. Based on the value of the matching score, it can be determined that the three meet the conditions and can form a corresponding triple.

[0105] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a parking information integration method. In the parking information integration method, when the data type includes an image data type, a point cloud data type, a text data type, a digital data type, and a video data type, step S2 may include steps S21 to S23:

[0106] Step S21: Determine the data type of the collected parking scene information.

[0107] In this embodiment, the data type of the parking scene information collected by the present application includes but is not limited to an image data type, a point cloud data type, a text data type, a digital data type, and a video data type. First, the data type of the collected parking scene information is determined so that the parking scene information can be subsequently processed in a unified format using a corresponding preprocessing algorithm.

[0108] Step S22: determining a target processing algorithm corresponding to the data type of the parking scene information according to the data type of the parking scene information.

[0109] In this embodiment, according to the data type of the parking scene information, a target processing algorithm corresponding to the data type is determined. Figure 4 As shown, the target processing algorithm corresponding to the image data type is specifically: inputting the collected original image data belonging to the parking scene information into a deep learning framework (such as TensorFlow, PyTorch); then standardizing the original image data, and the standardization processing at least includes: cropping, rotation, scaling, image enhancement, contrast adjustment, brightness adjustment, and noise removal; then inputting the standardized original image data into a deep learning target detection or semantic segmentation model (such as U-Net, DeepLab, YOLO, etc.) for feature extraction, and finally outputting structured data in the format of JSON, CSV, or XML.

[0110] In this embodiment, if Figure 5 As shown in the figure, the target processing algorithm corresponding to the point cloud data type is as follows: the collected original point cloud data belonging to the parking scene information is loaded into the point cloud processing library (such as PCL, Open3D) for basic coordinate transformation and normalization processing; then the original point cloud data that has undergone basic coordinate change and normalization processing is subjected to standard point cloud coordinate transformation and point cloud normalization processing; then the point cloud data that has undergone standard point cloud coordinate transformation and point cloud normalization processing is input into the deep learning model (such as PointNet, PointCNN, etc.) for feature extraction, and the extracted features include at least the density, shape, and spatial relationship of the point cloud; the final output format is structured data in JSON, CSV, and XML.

[0111] In this embodiment, if Figure 6 As shown in the figure, the target processing algorithm corresponding to the text data type is as follows: the collected original text data belonging to the parking scene information is cleaned to remove irrelevant characters, punctuation marks, stop words, etc. in the text; then, the text embedding technology based on deep learning (Word2Vec, BERT, etc.) is used to segment the text and convert the text data into a vector representation of a fixed dimension, and the text entities are identified through the entity recognition model (BiLSTM-CRF, Transformer, etc.), such as parking space type, parking lot name, etc., and the final output format is structured data in JSON, CSV, and XML.

[0112] In this embodiment, if Figure 7As shown, the target processing algorithm corresponding to the digital data type is specifically: digitally clean the collected original digital data belonging to the parking scene information and remove outliers to ensure the accuracy and consistency of the digital data; perform feature conversion and normalization on the digitally cleaned digital data through a deep learning network (neural network, decision tree, etc.) to eliminate dimensional differences, and finally output structured data in the format of JSON, CSV, and XML.

[0113] In this embodiment, if Figure 8 As shown, the target processing algorithm corresponding to the video data type is specifically as follows: loading and decoding the collected original video data belonging to the parking scene information; using the video processing library (OpenCV, ffmpeg) in the deep learning framework (such as TensorFlow, PyTorch) to load the original video data and decode it into a series of image frame sequences; selecting key frames in the image frame sequence by frame sampling to reduce the amount of calculation (such as determining key frames based on frame intervals and scene change detection); standardizing the collected key frame images, and the standardization processing at least includes: cropping, rotation, scaling, image enhancement, contrast adjustment, brightness adjustment, and noise removal; then inputting the standardized original image data into the deep learning target detection or semantic segmentation model (such as U-Net, DeepLab, YOLO, etc.) for feature extraction; using RNN, LSTM, and GRU to perform time series modeling on the image frame sequence, capturing the time series dependency between frames, obtaining data in Tensor format, and converting the data into structured data in the format of JSON, CSV, and XML.

[0114] Step S23: pre-processing the parking scene information in a unified format using the target processing algorithm to obtain target parking data.

[0115] In this embodiment, a corresponding target processing algorithm is selected according to the data type of the parking scene information to format the parking scene information to obtain corresponding target parking data, so that parking scene information of various different data types will be converted into target parking data of a unified format.

[0116] In combination with the above embodiments, in one implementation, the present application embodiment further provides a parking information integration method. In the parking information integration method, the method further includes steps S5 to S6:

[0117] Step S5: When new target parking data is obtained, entity relationship extraction and fusion are performed on the new target parking data through the knowledge graph model to determine the triples to be updated.

[0118] In this embodiment, the application continuously collects parking scene information, and continuously expands and updates the constructed parking knowledge graph based on the newly collected parking scene information. Specifically: continuously collect new parking scene information, and pre-process the parking scene information of multiple data types from multiple data sources in a unified format through the same implementation method as above to obtain the corresponding target parking data, and then use the knowledge graph model to extract and fuse the entities and the relationships between entities of the target parking data in the same implementation method as above to obtain each triple to be updated.

[0119] Step S6: updating the parking knowledge graph using the triples to be updated.

[0120] In this embodiment, after each triple to be updated is determined in step S5, each triple to be updated is constructed into the current parking knowledge graph through knowledge merging to update the parking knowledge graph.

[0121] In combination with the above embodiments, in one implementation, the present application embodiment further provides a parking information integration method. In the parking information integration method, step S6 may include steps S61 to S64:

[0122] Step S61: Determine whether there is a target triplet corresponding to the triplet to be updated in the parking knowledge graph.

[0123] In this embodiment, after each triple to be updated is determined in step S5, each triple to be updated is compared with all triples used to construct the current parking knowledge graph, and for each triple to be updated, it is determined whether there is a corresponding target triple in all triples used to construct the current parking knowledge graph. Specifically, a triple to be updated is taken out, and unique identifiers of two entities in the unit group to be updated are extracted, and then the unique identifiers of the two extracted entities are compared with all triples used to construct the current parking knowledge graph to determine whether there is a triple whose two unique identifiers are respectively the same as the two unique identifiers of the two extracted entities, that is, one unique identifier in the triple is the same as one of the two extracted unique identifiers, and the other unique identifier in the triple is the same as the other unique identifier of the two extracted unique identifiers. If such a triple exists, the triple is determined as the target triple corresponding to the triple to be updated. In the same implementation manner, it can be determined whether other triples to be updated have corresponding target triples in all triples used to construct the current parking knowledge graph.

[0124] Step S62: When the target triplet exists, the triplet to be updated is compared with the target triplet to obtain a corresponding comparison result.

[0125] In this embodiment, for any triple to be updated, if there is a target triple corresponding to the any triple to be updated among all triples used to construct the current parking knowledge graph, it indicates that the triple to be updated and the corresponding target triple represent the same triple relationship, the difference is that the specific parameters of the entity in the triple to be updated may have changed compared with the target triple corresponding to itself. For example, for the triple to be updated (parking space, quantity, 500), the parking space in the triple has a unique identifier and the value 500 in the triple has a unique identifier; the target triple corresponding to the triple to be updated is (parking space, quantity, 520), wherein the parking space in the target triple has the same unique identifier as the parking space in the above triple to be updated, and they point to parking spaces in the same parking lot, wherein the value 520 in the target triple has the same unique identifier as the value 500 in the above triple to be updated, and they point to the number of parking spaces in the same parking lot; therefore, the triple to be updated and the corresponding target triple represent the same triple relationship, the difference is that the value 500 of the entity in the triple to be updated has changed compared to the value 520 of the entity in the target triple corresponding to itself.

[0126] In this embodiment, each triple to be updated that has a corresponding target triple is compared with the corresponding target triple to determine whether specific parameters of entities in the two corresponding triples have changed, thereby obtaining a corresponding comparison result.

[0127] Step S63: According to the comparison result, the target triplet is replaced with the triplet to be updated.

[0128] In this embodiment, the comparison result obtained by step S62 records the result obtained by comparing each triple to be updated with the corresponding target triple. For any triple to be updated among the triples to be updated included in the corresponding result, if the specific parameters of the entity in the target triple corresponding to the triple to be updated are changed, the validity check is performed on the triple to be updated to determine whether the change of the specific parameters of the entity in the triple to be updated is correct. If the validity check of the triple to be updated passes, it is determined that the change of the specific parameters of the entity in the triple to be updated is correct. At this time, the target triple corresponding to the triple to be updated is replaced by the triple to be updated, and then the triple to be updated is constructed into the parking knowledge graph through knowledge merging to update the entity node of the target triple corresponding to the triple to be updated in the parking knowledge graph. If the validity check fails, the triple to be updated is deleted. Among them, an optional implementation method of the validity check is: engineering personnel conduct an on-site investigation and confirmation of the information in any one of the triples to be updated to determine whether the validity check of the any one of the triples to be updated is passed; another optional implementation method of the validity check is: a preset number of parking scene information corresponding to the any one of the triples to be updated is collected, and based on the preset number of collected parking scene information, a corresponding preset number of triples to be updated that represent the same triple relationship as the any one of the triples to be updated are determined, and then it is determined whether the specific parameters of the entities in the preset number of triples to be updated that are currently determined are all consistent. If they are all consistent, it is determined that the validity check of the any one of the triples to be updated is passed.

[0129] Step S64: When the target triple does not exist, the triple to be updated is updated into the parking knowledge graph through knowledge merging.

[0130] In this embodiment, for any triple to be updated, if there is no target triple corresponding to the triple to be updated among all the triples used to construct the current parking knowledge graph, it indicates that the triple to be updated is not in the currently constructed parking knowledge graph. At this time, the validity check is performed on the triple to be updated in the same manner as above. If the validity check passes, it is determined that the specific parameters of the entity in the triple to be updated are correct. At this time, the triple to be updated is constructed into the parking knowledge graph through knowledge merging to expand the constructed parking knowledge graph. If the validity check fails, the triple to be updated is deleted.

[0131] In this embodiment, in order to compare the triples to be updated and the corresponding target triples in this application, this application will record and save all triples involved in the latest parking knowledge graph in the database. When the triples to be updated are updated to the parking knowledge graph, the records of the triples to be updated are correspondingly stored in the database; and when the triples to be updated need to replace the entity nodes corresponding to the target triples in the parking knowledge graph corresponding to the triples to be updated, the records of the triples to be updated are also stored in the database, and the target triples corresponding to the updated triples recorded in the database are deleted, thereby ensuring that the database always records all triples involved in the latest parking knowledge graph.

[0132] In this embodiment, the parking knowledge graph may include parking lot entities and parking lot entities having entity-to-entity relationships, and the relationships include at least geographical proximity relationships, traffic connectivity relationships, shared resource relationships, and functional complementarity relationships. Among them, the geographical proximity relationship is used to describe the geographical proximity or proximity of two parking lots, so that drivers can choose a suitable parking place according to real-time needs, such as the triple (parking lot A, adjacent, parking lot B); the traffic connectivity relationship is used to describe that one can easily reach another parking lot or public transportation station from one parking lot, such as the triple (parking lot A, connection, bus line X), the triple (parking lot A, connection, subway station Y), and the triple (parking lot A, direct access, parking lot B); the shared resource relationship is used to describe the resource sharing or user diversion mechanism between the two parking lots during peak hours or other specific circumstances, such as (parking lot A, diversion service, parking lot B); the functional complementarity relationship is used to describe the functional complementarity of one-stop shopping and entertainment services between the two parking lots, such as the triple (parking lot A, supporting facilities, supermarket) and the triple (parking lot B, supporting facilities, cinema), and then through the triple (supermarket, partner, cinema), it is deduced that there is a functional complementarity between parking lots A and B with one-stop shopping and entertainment services.

[0133] In combination with the above embodiments, in one implementation, the present application embodiment further provides a parking information integration method. In the parking information integration method, the method further includes steps S7 to S9:

[0134] Step S7: receiving a parking information acquisition request from a terminal.

[0135] In this embodiment, the server side having the parking knowledge graph receives the parking information acquisition request sent by the terminal device.

[0136] Step S8: In response to the parking information acquisition request, obtain the target parking information corresponding to the parking information acquisition request in the parking knowledge graph, wherein the parking knowledge graph is the parking knowledge graph in the parking information integration method described in the first aspect of the present application.

[0137] In this embodiment, the server side responds to the parking information acquisition request sent by the terminal device and obtains the target parking information corresponding to the parking information acquisition request in the parking knowledge graph. The parking information acquisition request includes but is not limited to a query request, a push request and a recommendation request. Specifically: based on the current location of the vehicle and the destination to which the vehicle plans to navigate, a corresponding parking information acquisition request is sent through the terminal device, wherein the terminal device may be a vehicle terminal and / or a mobile terminal bound to the vehicle. Based on the received parking information acquisition request, the server side obtains each entity node within a preset length range around the current location of the vehicle from the parking knowledge graph and sends it to the terminal device to help the user select the parking lot closest to the destination or starting point. The entity nodes sent to the terminal device may include the specific value entity nodes of the charges of each parking lot, so as to help the user select the parking lot based on cost-effectiveness, such as the parking knowledge graph includes the entity node parking lot A and the entity node 3 yuan per hour constructed based on the triple (parking lot A, charge, 3 yuan per hour). Each entity node sent to the terminal device may include a specific value entity node of the business hours of each parking lot, so as to help users select parking lots based on the opening hours of the parking lot, such as the parking knowledge graph includes the entity node parking lot A and the entity node 9:00 to 23:00 constructed based on the triple (parking lot A, business hours, 9:00 to 23:00). Each entity node sent to the terminal device may include the internal facility entity node of each parking lot, so as to help users select a suitable parking lot based on their own needs, such as the parking knowledge graph includes the entity node parking lot A and the entity node car wash service constructed based on the triple (parking lot A, with, car wash service). Each entity node sent to the terminal device may include the current status entity node of each parking lot, so as to help users select available parking lots, such as the parking knowledge graph includes the entity node parking lot A and the entity node maintenance constructed based on the triple (parking lot A, status, maintenance).

[0138] Step S9: Send the target parking information to the terminal for display.

[0139] In this embodiment, the determined target parking information is sent to the corresponding terminal device for display. At the same time, the present application establishes a user evaluation mechanism in each terminal device, and makes intelligent recommendations for each terminal based on the user's ratings and comments on each parking lot and the user's historical preferences.

[0140] In this embodiment, in order to synchronize the relevant target parking information to the relevant application programs of the terminal, the present application first develops an API interface and implements a data synchronization mechanism, and integrates the API into the relevant application programs of the terminal.

[0141] Specifically, first develop the API and design an API that is simple, easy to use and conforms to HTTP standards, such as RESTful API, and define clear resource paths (endpoints), such as / parking / locations for obtaining parking lot location information, / parking / spaces for obtaining parking space information, etc. Define the data model returned by the API, use JSON as the data exchange format, and ensure that the data format is unified and easy to parse. Implement a user authentication mechanism to ensure that only authorized users can access the data. When implementing it, you can use authentication protocols such as OAuth2.0. Next, design error codes and error messages so that the APP can handle abnormal situations. Finally, when implementing it, provide detailed API documentation, including interface descriptions, request parameters, return results, etc. Provide a test interface to facilitate developers to verify API functions during the development process.

[0142] Specifically, data synchronization is then performed. During the specific implementation, real-time synchronization or periodic synchronization is selected according to the needs. Specifically, when the demand requires real-time data update, real-time data synchronization is achieved using technologies such as WebSocket or long polling. When real-time synchronization is not necessary, a periodic synchronization strategy is set, such as synchronizing data once an hour or a day. Regardless of the method, this application sets an incremental update mechanism to only synchronize data that has changed since the last synchronization, reducing the amount of data transmission and synchronization time. Formulate a conflict resolution strategy to deal with data conflicts that may arise during the synchronization process.

[0143] Specifically, the application is then integrated, the API is integrated into the application, and the API calling logic is implemented in the application, including user authentication, data request and parsing, etc. UI / UX design is performed to design an intuitive and easy-to-use user interface to facilitate users to view and manage parking information on various terminal devices, and provide search, filtering, sorting and other functions to help users quickly find the required information. Design a local storage mechanism to cache part or all of the data on the device to improve the APP response speed. In specific implementation, local storage can be implemented using technologies such as SQLite, Core Data (iOS) or Realm. In addition, notification and reminder functions are implemented as required, such as notifying the user when the status of the parking space in the parking lot where the user wants to park changes.

[0144] Specifically, the API is deployed on a stable and reliable server to ensure the availability and stability of the service. During the specific implementation, tools such as ELK Stack (Elasticsearch, Logstash, Kibana) are used to collect, analyze and visualize logs, so as to capture potential problems in a timely manner and handle them, thereby continuously improving service performance and user experience.

[0145] In this embodiment, an optional implementation method for displaying the target parking information obtained based on the parking knowledge graph in the terminal application is: setting a title bar at the top to display the title of the current page, such as "Parking Information", and providing a return button to facilitate the user to return to the previous page. A parking lot information area is set in the page to display the parking lot name, parking lot location and map location selected by the user. The parking lot name is displayed in large and bold fonts, such as "XX Shopping Center Parking Lot". The parking lot location displays the detailed address of the parking lot, using a smaller font, but can be clicked to expand and view the full address. For map location, a small map is embedded to display the specific location of the parking lot and mark the parking space location of the vehicle. The user can click on the map to navigate. A vehicle information area is set in the page to display the license plate number and the parking space number where the vehicle is parked. The license plate number can be highlighted using an icon or background color, and the specific parking space number of the vehicle is displayed for the parking space number, such as "B2 floor, parking space 123". Set up the parking lot charging and payment information area on the page to display the parking time and fee, payment status and payment button. For parking time and fee, display the current parking time and fee, use large fonts and eye-catching colors, such as "Parking fee: ￥20.00", and display the payment status, such as "paid" or "unpaid", and use different colors or icons to distinguish. For the payment button, after clicking the button, you can jump to the payment page to make payment. Set up the surrounding facilities area on the page to display the surrounding facilities list and the charging pile information inside the parking lot. For the charging pile, if there is a charging pile, display the icon and text of "charging pile available" or "charging pile unavailable", and you can add more details, such as the number of charging piles and charging rate. For the surrounding facilities list, display the surrounding facilities in the form of cards, such as gas stations, restaurants, etc. Each card contains the facility name, distance and a brief description. Provide navigation buttons for the facilities in each card to facilitate users to go there. Including query and recommendation functions. Set up a bottom operation bar on the page, including a refresh button and a more options button. The refresh button will allow users to manually refresh the parking information to ensure the accuracy of the data. The more options button will allow users to click to expand more functions, such as contacting the parking lot, viewing historical parking records, etc.

[0146] In this embodiment, if Fig. 9As shown, the parking information integration method provided by the present application includes three parts, including a physical layer, a model layer and a business layer. The physical layer is used to store the target parking data obtained by pre-processing the parking scene information in a unified format, the model layer is used to construct a parking knowledge graph by processing the target parking data, and the business layer corresponds to the application program, directly interacts with the user, and provides the information and services required in the parking scene.

[0147] In this embodiment, a parking information integration method provided by this application is to effectively integrate, intelligently analyze, and share parking scene information in the cloud by introducing a parking knowledge graph, so as to provide strong support for the construction of an urban traffic big data analysis platform, thereby effectively alleviating the problem of parking difficulties and promoting the improvement of the smart transportation system. At the same time, it can more accurately understand parking data and provide more scientific and efficient decision-making support for urban traffic management. The construction method of the parking knowledge graph proposed in this application can collect parking scene information comprehensively and through multiple channels, and can deeply mine the relationship between parking data in the process of establishing the parking knowledge graph, thereby maximizing the value of a single data. The relevant data in the parking knowledge graph constructed by this application can be synchronized to the user terminal to provide support for user parking and other personalized needs, and can effectively improve the user experience. The parking knowledge graph constructed by this application can integrate and structure various information related to parking, including the location of parking stations, the number of parking spaces, parking fees, parking duration, etc. This data representation method makes information easier to query, understand and analyze, and improves the efficiency and accuracy of information utilization. Through the parking knowledge graph, users can obtain the required parking information more quickly, thus saving time and energy. At the same time, the parking knowledge graph can support intelligent parking plan recommendations and decision support. Through the entity and relationship information in the parking knowledge graph, the user's parking needs and preferences can be analyzed, and combined with real-time parking information, suitable parking stations and parking spaces can be recommended to users. This personalized recommendation service can improve user experience and optimize the allocation and utilization of parking resources. At the same time, the parking knowledge graph can promote the intercommunication and sharing of parking information. By storing the parking knowledge graph on the cloud server, relevant units can access and share the parking knowledge graph data through the cloud interface. This sharing mechanism enables all parties to obtain the latest parking information in real time, realize information intercommunication, and improve the utilization efficiency of parking resources. At the same time, the parking knowledge graph can support cross-domain and cross-platform data integration and application, providing more comprehensive and in-depth data support for urban traffic management and planning. In addition, the parking knowledge graph has the ability to be dynamically updated. As parking information changes, such as the emergence of new parking stations and changes in the number of parking spaces, the parking knowledge graph can be updated in real time to maintain the accuracy and timeliness of the information. This dynamic update feature enables the parking knowledge graph to better adapt to the changing needs of actual parking scenarios.

[0148] Based on the same inventive concept, the second aspect of the embodiment of the present application provides a parking information integration system, such as Fig.10 As shown, the system 1000 includes:

[0149] The data collection module 1001 is used to collect parking scene information of various data types.

[0150] The preprocessing module 1002 is used to preprocess the collected parking scene information of various data types in a unified format to obtain target parking data.

[0151] The extraction and fusion module 1003 is used to extract and fuse entity relationships of the target parking data through a knowledge graph model to determine each triple.

[0152] The knowledge merging module 1004 is used to construct a corresponding parking knowledge graph by merging the knowledge of the triples.

[0153] Optionally, the system further includes:

[0154] A feature extraction module, used for inputting the target parking data into a pre-trained model for feature extraction to obtain target features;

[0155] The extraction and fusion module 1003 includes: an extraction and fusion submodule, which is used to extract and fuse the entity relationships of the target features through a knowledge graph model to determine each triple.

[0156] Optionally, the extraction fusion submodule includes:

[0157] An input module, used to input the target feature into the knowledge graph model;

[0158] A weighted processing module, used to perform weighted processing on the target features through a weighted multi-head attention mechanism in the knowledge graph model to obtain weighted features;

[0159] An extraction module, used to extract entities and relationships from the weighted features through the entity-relationship fusion sub-model in the knowledge graph model to obtain each entity and the relationship between each entity;

[0160] A fusion module is used to perform knowledge fusion on the entities and the relationships between the entities through the entity relationship fusion sub-model in the knowledge graph model to obtain each triple.

[0161] Optionally, the fusion module includes:

[0162] A scoring module, used to determine the matching scores between the extracted entities and the inter-entity relations through the scoring function in the entity-relation fusion sub-model;

[0163] The fusion submodule is used to determine the entities and inter-entity relationships whose matching scores meet the set conditions as target entities and inter-target entity relationships to construct corresponding triples.

[0164] Optionally, the preprocessing module 1002 includes:

[0165] a data type determination module, for determining a data type of the collected parking scene information when the data type includes an image data type, a point cloud data type, a text data type, a digital data type, and a video data type;

[0166] a target processing algorithm determination module, configured to determine a target processing algorithm corresponding to the data type of the parking scene information according to the data type of the parking scene information;

[0167] The preprocessing submodule is used to preprocess the parking scene information in a unified format using the target processing algorithm to obtain target parking data.

[0168] Optionally, the system further includes:

[0169] A first extraction and fusion module is used to extract and fuse entity relationships of new target parking data through a knowledge graph model to determine triples to be updated when new target parking data is acquired;

[0170] The parking knowledge graph refinement module is used to update the parking knowledge graph using the triples to be updated.

[0171] Optionally, the parking knowledge graph has a finer module, including:

[0172] a target triplet determination module, used to determine whether there is a target triplet corresponding to the triplet to be updated in the parking knowledge graph;

[0173] A comparison module, used for comparing the triple to be updated with the target triple to obtain a corresponding comparison result when the target triple exists;

[0174] An updating and replacing module, configured to replace the target triplet with the triplet to be updated according to the comparison result;

[0175] An updating module, used to update the to-be-updated triple to the parking knowledge graph by knowledge merging when the target triple does not exist

[0176] Optionally, the system 1000 further includes:

[0177] The request receiving module is used to receive a parking information acquisition request from a terminal.

[0178] A parking information acquisition module is used to respond to the parking information acquisition request and obtain target parking information corresponding to the parking information acquisition request in the parking knowledge graph, wherein the parking knowledge graph is a parking knowledge graph in a parking information integration method described in the first aspect of an embodiment of the present application.

[0179] The parking information feedback module is used to send the target parking information to the terminal for display.

[0180] A third aspect of an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, a parking information integration method as described in the first aspect of an embodiment of the present application is implemented.

[0181] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, a parking information integration method as described in the first aspect of an embodiment of the present application is implemented.

[0182] As 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 system embodiment.

[0183] It should be noted that, for the method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0184] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0185] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the embodiments of the present application may adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the embodiments of the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0186] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0187] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0188] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0189] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0190] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal 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, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0191] The above is a detailed introduction to a parking information integration method, system and electronic device provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and its core idea of ​​the present application. At the same time, for a person skilled in the art, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as limiting the present application.

Claims

1. A parking information integration method, characterized in that: The method comprises: Collect parking scene information of various data types; Preprocessing the collected parking scene information of various data types in a unified format to obtain target parking data; Extracting and fusing entity relationships of the target parking data through a knowledge graph model to determine each triple; By merging the knowledge of each triple, a corresponding parking knowledge graph is constructed; Wherein, before extracting and fusing entity relationships of the target parking data through a knowledge graph model to determine each triple, the method further includes: Inputting the target parking data into a pre-trained model for feature extraction to obtain target features; The entity relationship extraction and fusion of the target parking data are performed by the knowledge graph model to determine each triple, including: Extract and fuse entity relationships of the target features through a knowledge graph model to determine each triple; The target features are subjected to entity relationship extraction and fusion through the knowledge graph model to determine each triple, including: Inputting the target feature into the knowledge graph model; The target features are weighted by the weighted multi-head attention mechanism in the knowledge graph model to obtain weighted features; The entity-relationship fusion sub-model in the knowledge graph model is used to extract entities and relationships from the weighted features to obtain each entity and the relationship between each entity; The entity relationship fusion sub-model in the knowledge graph model is used to perform knowledge fusion on the entities and the relationships between the entities to obtain the triples.

2. The parking information integration method according to claim 1, characterized in that: The entity relationship fusion sub-model in the knowledge graph model performs knowledge fusion on the entities and the relationships between the entities to obtain the triples, including: Determining the matching scores between the extracted entities and the inter-entity relations through the scoring function in the entity relation fusion sub-model; The entities and inter-entity relationships whose matching scores meet the set conditions are determined as target entities and inter-target entity relationships to construct corresponding triples.

3. The parking information integration method according to claim 1, characterized in that: When the data types include image data types, point cloud data types, text data types, digital data types, and video data types, preprocessing the collected parking scene information of the various data types in a unified format to obtain target parking data includes: Determine the data type of the collected parking scene information; Determining, according to the data type of the parking scene information, a target processing algorithm corresponding to the data type of the parking scene information; The parking scene information is pre-processed in a unified format by the target processing algorithm to obtain target parking data.

4. The parking information integration method according to claim 1, characterized in that: The method further comprises: When new target parking data is obtained, entity relationship extraction and fusion are performed on the new target parking data through the knowledge graph model to determine triples to be updated; The parking knowledge graph is updated using the triples to be updated.

5. The parking information integration method according to claim 4, characterized in that: Updating the parking knowledge graph by using the triples to be updated includes: Determining whether there is a target triplet corresponding to the triplet to be updated in the parking knowledge graph; In the case where the target triplet exists, comparing the triplet to be updated with the target triplet to obtain a corresponding comparison result; According to the comparison result, replacing the target triplet with the triplet to be updated; In the case that the target triple does not exist, the triple to be updated is updated into the parking knowledge graph through knowledge merging.

6. A parking information integration method according to any one of claims 1 to 5, characterized in that: The method further comprises: Receiving a parking information acquisition request from a terminal; In response to the parking information acquisition request, acquiring target parking information corresponding to the parking information acquisition request in the parking knowledge graph; The target parking information is sent to the terminal for display.

7. A parking information integration system, characterized in that: The system comprises: A data collection module, used to collect parking scene information of various data types; A preprocessing module, used for preprocessing the collected parking scene information of various data types in a unified format to obtain target parking data; An extraction and fusion module, used to extract and fuse entity relationships of the target parking data through a knowledge graph model to determine each triple; A knowledge merging module, used to construct a corresponding parking knowledge graph by merging the knowledge of each triple; Wherein, the system further comprises: A feature extraction module, used for inputting the target parking data into a pre-trained model for feature extraction to obtain target features; The extraction and fusion module includes: an extraction and fusion submodule, which is used to extract and fuse entity relationships of the target features through a knowledge graph model to determine each triple; Wherein, the extraction and fusion submodule includes: An input module, used to input the target feature into the knowledge graph model; A weighted processing module, used to perform weighted processing on the target features through a weighted multi-head attention mechanism in the knowledge graph model to obtain weighted features; An extraction module, used to extract entities and relationships from the weighted features through the entity-relationship fusion sub-model in the knowledge graph model to obtain each entity and the relationship between each entity; A fusion module is used to perform knowledge fusion on the entities and the relationships between the entities through the entity relationship fusion sub-model in the knowledge graph model to obtain each triple.

8. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, a parking information integration method as claimed in claims 1 to 6 is implemented.

Images