Road berth dynamic planning method and device for intelligent transportation

Abstract

The present application discloses a method and device for dynamic planning of road parking spaces for smart transportation. The server-side method includes: responding to a parking space planning request sent by a client, planning the shortest driving path according to the destination of the user and the current location of the user carried in the parking space planning request; obtaining parking space videos of sub-sections at a preset distance from the destination in the shortest driving path from a high-position video monitoring system; identifying multiple idle road parking spaces on the sub-sections according to the parking space videos of the sub-sections; identifying the optimal target parking space from multiple idle road parking spaces; starting the navigation service to generate navigation information feedback to the client with the location of the target parking space as the end point, and continuing to execute the step of obtaining parking space videos of sub-sections at a preset distance from the destination in the shortest driving path to perform dynamic planning of road parking spaces. Therefore, by adopting the embodiment of the present application, the user can quickly and accurately park the vehicle at a parking space near the destination.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation technology, and in particular to a method and device for dynamic planning of road berths for intelligent transportation. Background Art

[0002] In modern urban environments, with the surge in the number of motor vehicles, urban traffic congestion and parking difficulties are becoming increasingly serious. Especially in commercial areas, office areas and residential areas, the contradiction between the supply and demand of parking spaces is particularly prominent. Users often need to spend a lot of time driving at low speeds on the road when looking for parking spaces, which not only increases fuel consumption and emissions, but also aggravates traffic pressure. Therefore, the development of a parking space system that can effectively manage and optimize urban parking resources is of great significance to alleviating urban traffic pressure and improving parking efficiency.

[0003] In the related art, the parking system is, for example, an existing parking search system based on video surveillance, which monitors the occupancy status of parking spaces through video. Users can query available parking spaces through corresponding mobile applications and navigate to them by themselves. Due to the special nature of road parking spaces being open, the parking spaces may be occupied by other vehicles during the time when users arrive at the reserved parking spaces, and users cannot park their vehicles quickly and accurately at parking spaces near their destinations. Summary of the invention

[0004] The embodiments of the present application provide a method and device for dynamic planning of road berths for intelligent transportation. In order to have a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. This summary is not a general review, nor is it intended to identify key / important components or describe the scope of protection of these embodiments. Its only purpose is to present some concepts in a simple form as a preface to the detailed description that follows.

[0005] In a first aspect, an embodiment of the present application provides a method for dynamic planning of road berths for smart transportation, which is applied to a server, and the method includes:

[0006] In response to the berth planning request sent by the client, the shortest driving route is planned according to the destination of the user and the current location of the user carried in the berth planning request;

[0007] Obtaining parking videos of sub-sections at a preset distance from the destination in the shortest driving path from a high-position video surveillance system;

[0008] According to the parking space video of the sub-road section, a plurality of idle road parking spaces on the sub-road section are identified; wherein the plurality of idle road parking spaces are obtained by analyzing the key frame vector and the description text vector of each key frame according to the pre-trained idle road parking space determination model, the key frame vector of each key frame is generated based on a plurality of key frames containing clear parking space images extracted from the parking space video, the description text vector is generated based on the parking space description information of each key frame, and the pre-trained idle road parking space determination model is a mathematical model for determining whether a road parking space on the sub-road section is idle;

[0009] Identify the best target parking space from multiple idle road parking spaces; wherein the best target parking space is the idle road parking space with the highest matching score with the user, and the idle road parking space with the highest matching score with the user is obtained by performing data analysis and processing on the parking space identity information of each idle road parking space according to the parking demand information of the user, and the parking space identity information of each idle road parking space is the size, type, fee information and GPS coordinates of each idle road parking space collected;

[0010] Taking the location of the target parking space as the end point, the navigation service is started to generate navigation information and feedback it to the client, and the step of obtaining the parking space video of the sub-section at a preset distance from the destination in the shortest driving path is continued to perform dynamic planning of road parking spaces until the user reaches the target parking space.

[0011] Optionally, based on the parking space video of the sub-road section, a plurality of vacant road parking spaces on the sub-road section are identified, including:

[0012] Extracting multiple key frames containing clear images of the parking spaces on the sub-road section from the parking space video on the sub-road section;

[0013] Obtain the text description information corresponding to each key frame, and obtain the berth description information of each key frame;

[0014] Perform motion correction, time calibration and spatial normalization on each key frame to obtain the key frame vector of each key frame;

[0015] Use Word2Vec network or BERT network to vectorize the berth description information of each key frame to obtain the description text vector of each key frame;

[0016] The key frame vector of each key frame and its description text vector are input into a pre-trained idle road parking space determination model, and a plurality of idle road parking spaces corresponding to the sub-road section are output.

[0017] Optionally, the pre-trained idle road parking space determination model includes a spatial feature extraction module, a text feature extraction module, a spatial feature and text feature fusion module, and a parking space prediction module;

[0018] The key frame vector and its description text vector of each key frame are input into the pre-trained idle road parking space determination model, and multiple idle road parking spaces corresponding to the sub-section are output, including:

[0019] The spatial feature extraction module performs spatial feature extraction processing on the key frame vector of each key frame to obtain the spatial features corresponding to the key frame vector of each key frame;

[0020] The text feature extraction module performs text feature extraction processing on the description text vector of each key frame to obtain the text features corresponding to the description text vector of each key frame;

[0021] The spatial feature and text feature fusion module performs feature fusion processing on the spatial feature corresponding to the key frame vector of each key frame and the text feature corresponding to the description text vector of each key frame to obtain the frame-text fusion feature of each key frame;

[0022] The frame-text fusion feature of each key frame is input into the parking space prediction module, and multiple vacant road parking spaces corresponding to the sub-section are output; wherein, the expression of feature fusion processing is:

[0023]

[0024] in, For the The key frame vector of the key frame corresponds to the Spatial features With The description text vector of the key frame corresponds to the Text features The frame-text fusion features obtained by fusion.

[0025] Optionally, a spatial feature extraction process is performed on the key frame vector of each key frame to obtain a spatial feature corresponding to the key frame vector of each key frame, including:

[0026] Extracting data features of a key frame vector of each key frame;

[0027] Extract the spatial features of data features in the time dimension;

[0028] The spatial vector corresponding to each spatial point in the spatial feature is used as a network node, and the spatial connection information in the data feature is used as the edge of the network;

[0029] Based on network nodes and network edges, a node graph is constructed;

[0030] Perform linear transformation on the attributes of each node in the node graph to map the information of different nodes into the same vector space and obtain the basic vector representation of each key frame;

[0031] Feature aggregation is performed on the basis vector representation of each key frame to obtain the spatial features corresponding to the key frame vector of each key frame.

[0032] Optionally, the pre-trained free road parking space determination model is generated according to the following steps, including:

[0033] Obtain historical berth videos within a preset period on any road section, and extract multiple historical key frames containing clear images of berths in the historical berth videos;

[0034] For each historical key frame, generate text description information including the location, size, color and whether there is a vehicle parked at the berth, and obtain historical berth description information of each historical key frame;

[0035] Generate a model training set based on each historical key frame and its historical berth description information;

[0036] Create and aggregate a spatial feature extraction module, a text feature extraction module, a spatial feature and text feature fusion module, and a berth prediction module to obtain an idle road berth determination model;

[0037] Input the model training set into the idle road parking space determination model and output the model loss value;

[0038] When the model loss value reaches the minimum, the pre-trained vacant road parking space determination model is obtained; or when the model loss value does not reach the minimum, the model loss value is back-propagated to update the model parameters of the vacant road parking space determination model, and the step of inputting the model training set into the vacant road parking space determination model is continued until the model loss value reaches the minimum.

[0039] Optionally, based on each historical key frame and its historical berth description information, a model training set is generated, including:

[0040] Perform motion correction, time calibration and spatial normalization on each historical key frame to obtain a historical key frame vector for each historical key frame;

[0041] Use the Word2Vec network or the BERT network to vectorize the historical berth description information of each historical key frame to obtain the historical description text vector of each historical key frame;

[0042] For each historical key frame vector and its historical description text vector, the input real label used to characterize whether the road berth in each historical key frame is vacant is received for data calibration to obtain a model training set; wherein,

[0043] The model loss function of the pre-trained idle road parking space determination model is:

[0044]

[0045] in, is the loss value, is the true label of the model training data, is the predicted label predicted by the idle road parking space determination model, It is obtained by model processing based on a model training data.

[0046] Optionally, identifying the best target parking space from a plurality of free road parking spaces, including:

[0047] Collect the size, type, fee information and GPS coordinates of each free road berth to obtain the berth identity information of each free road berth;

[0048] According to the user's identity, obtaining corresponding historical parking data from a pre-established mapping relationship between the identity and the historical parking data;

[0049] Analyze users’ parking demand information based on historical parking data;

[0050] According to the user's parking demand information, the parking space identity information of each idle road parking space is analyzed and processed to obtain a matching score between each idle road parking space and the user;

[0051] The vacant road parking space with the highest matching score with the user is determined as the optimal target parking space.

[0052] Optionally, the user's parking demand information includes the user's vehicle size information, the user's parking space preference information, the user's payment preference information, and the user's destination GPS;

[0053] According to the user's parking demand information, the parking space identity information of each idle road parking space is analyzed and processed to obtain the matching score between each idle road parking space and the user, including:

[0054] According to the size of each free road parking space and the vehicle size information, the first matching degree between the parking space size and the user's vehicle is calculated; wherein,

[0055] First matching degree The calculation formula is:

[0056]

[0057] in, is an adjustment factor used to ensure that the berth size is at least a certain proportion of the vehicle size; where is a function that returns the minimum of two values;

[0058] Calculating a second matching degree between the berth type and the user's berth preference according to the type of each idle road berth and the user's berth preference information;

[0059] Second matching degree The calculation formula is:

[0060]

[0061] Calculating a third matching degree between parking space fees and user payment according to fee information of each idle road parking space and user payment preference information;

[0062] The third matching The calculation formula is:

[0063]

[0064] Calculating a fourth matching degree between the geographic location of each free road parking space and the user's destination based on the GPS coordinates of each free road parking space and the user's destination GPS;

[0065] Fourth matching degree The calculation formula is:

[0066]

[0067] in, is the attenuation factor, which is used to adjust the effect of distance on matching degree. is the exponent of the constant e, which is 2.71828;

[0068] The first matching degree, the second matching degree, the third matching degree and the fourth matching degree are weighted and summed to obtain a matching score between each free road parking space and the user; wherein,

[0069]

[0070] Among them, S is the matching score between each idle road parking space and the user, , , , are weight values ​​and the sum is 1, is the first matching degree between the berth size and the user's vehicle, is the second matching degree between the berth type and the user's berth preference, The third matching degree for berth fee and user payment, It is the fourth matching degree between the berth geographical location and the user's destination.

[0071] Optionally, based on historical parking data, analyze the user's parking demand information, including:

[0072] Extract the user’s vehicle size information from historical parking data;

[0073] Based on historical parking data, analyze the parking space type selected by the user and the user's preference for specific parking space features to obtain the user's parking space preference information;

[0074] Based on historical parking data, analyze the user's parking fee payment records and payment amounts to obtain the user's payment preference information;

[0075] Get the GPS of your destination;

[0076] The user's vehicle size information, the user's parking space preference information, the user's payment preference information, and the user's destination GPS are used as the user's parking demand information.

[0077] In a second aspect, an embodiment of the present application provides a road berth dynamic planning device for smart transportation, the device comprising:

[0078] The shortest driving route planning module is used to respond to the berth planning request sent by the client, and plan the shortest driving route according to the destination of the user and the current location of the user carried in the berth planning request;

[0079] The parking space video acquisition module is used to acquire the parking space video of the sub-section at a preset distance from the destination in the shortest driving path from the high-position video monitoring system;

[0080] an idle road parking space identification module, for identifying a plurality of idle road parking spaces on a sub-road section according to a parking space video of the sub-road section; wherein the plurality of idle road parking spaces are obtained by analyzing a key frame vector and a description text vector of each key frame according to a pre-trained idle road parking space determination model, the key frame vector of each key frame is generated based on a plurality of key frames containing clear parking space images extracted from the parking space video, the description text vector is generated based on parking space description information of each key frame, and the pre-trained idle road parking space determination model is a mathematical model for determining whether a road parking space on a sub-road section is idle;

[0081] The optimal target parking space identification module is used to identify the optimal target parking space from multiple idle road parking spaces; wherein the optimal target parking space is the idle road parking space with the highest matching score with the user, and the idle road parking space with the highest matching score with the user is obtained by performing data analysis and processing on the parking space identity information of each idle road parking space according to the parking demand information of the user, and the parking space identity information of each idle road parking space is the size, type, fee information and GPS coordinates of each idle road parking space collected;

[0082] The navigation service startup module is used to start the navigation service with the location of the target parking space as the end point, generate navigation information and feedback it to the client, and continue to execute the step of obtaining the parking space video of the sub-section at a preset distance from the destination in the shortest driving path to perform dynamic planning of road parking spaces until the user reaches the target parking space.

[0083] The technical solution provided by the embodiments of the present application may have the following beneficial effects:

[0084] In the embodiments of the present application, on the one hand, the present application extracts key frames from a continuous video stream, and uses a deep learning model to analyze the key frame vectors and their description text vectors, so as to efficiently and accurately determine the vacancy status of the parking space. At the same time, it can continuously track the user's driving status and the real-time changes of the parking space, and dynamically update the optimal parking space recommendation until the user successfully parks the vehicle. This method not only greatly improves the speed and accuracy of the parking space search, but also effectively alleviates urban traffic congestion by reducing the time wasted in the search process. On the other hand, the present application adopts a flexible personalized service mechanism. The system's recommendation of vacant parking spaces not only takes into account the physical properties of the parking space, such as size and type, but also takes into account the geographical distance from the user's destination and the user's willingness to pay. Through this personalized data analysis and processing, the system can recommend the most matching parking space to the user, thereby providing a more intimate and convenient parking experience, allowing the user to quickly and accurately park the vehicle at a parking space near the destination.

[0085] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0087] Figure 1 This is a method flow chart of a method for dynamic planning of road berths for smart transportation provided by an embodiment of the present application;

[0088] Figure 2 is a schematic diagram of a destination input on a client interface provided in an embodiment of the present application;

[0089] Figure 3 It is a schematic diagram of a data table of historical berth description information of each historical key frame provided by the present application;

[0090] Figure 4 It is a schematic diagram of a berth identity information data table of each vacant road berth provided by the present application;

[0091] Figure 5It is a schematic diagram of a user's parking demand information data table provided by the present application;

[0092] Figure 6 It is a schematic diagram of an application scenario provided by this application;

[0093] Figure 7 It is a structural schematic diagram of a road berth dynamic planning device for smart transportation provided by the present application;

[0094] Figure 8 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0095] The following description and the accompanying drawings sufficiently illustrate specific embodiments of the present application to enable those skilled in the art to practice them.

[0096] It should be clear that the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0097] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are only examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims.

[0098] In the description of the present application, it should be understood that the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances. In addition, in the description of the present application, unless otherwise specified, "multiple" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship.

[0099] At present, the berth system is, for example, an existing berth search system based on video surveillance, which monitors the occupancy status of berths through videos. Users can query the available berths through corresponding mobile applications and navigate to them by themselves.

[0100] The inventors realized that due to the special nature of road parking spaces being open, the parking spaces may be occupied by other vehicles during the time it takes for users to reach the reserved parking spaces, making it impossible for users to quickly and accurately park their vehicles in parking spaces near their destinations.

[0101] In order to solve the above problems, the present application provides a method and device for dynamic planning of road parking spaces for smart transportation to solve the problems existing in the above related technical problems. In the embodiment of the present application, on the one hand, the present application extracts key frames from continuous video streams, uses deep learning models to analyze key frame vectors and their description text vectors, thereby efficiently and accurately determining the idle state of parking spaces, and can continuously track the user's driving status and real-time changes of parking spaces, and dynamically update the optimal parking space recommendation until the user successfully parks the vehicle. This method not only greatly improves the speed and accuracy of parking space search, but also effectively alleviates urban traffic congestion by reducing the time waste in the search process; on the other hand, the present application adopts a flexible personalized service mechanism, and the system's recommendation of idle parking spaces not only considers the physical properties of the parking spaces, such as size and type, but also considers the geographical distance from the user's destination and the user's willingness to pay. Through this personalized data analysis and processing, the system can recommend the most matching parking spaces for users, thereby providing a more intimate and convenient parking experience, and enabling users to quickly and accurately park their vehicles at parking spaces near their destinations. The following exemplary embodiments are used for detailed description.

[0102] The following will be combined with the attached Figure 1 -Attached Figure 6 , a road berth dynamic planning method for smart transportation provided by an embodiment of the present application is introduced in detail. The method can be implemented by a computer program and can be run on a road berth dynamic planning device for smart transportation based on the von Neumann system. The computer program can be integrated into an application or run as an independent tool application.

[0103] See also Figure 1 , which is a flowchart of a method for dynamic planning of road berths for smart transportation, and is applied to a server. Figure 1 As shown, the method of the embodiment of the present application may include the following steps:

[0104] S101, in response to a berth planning request sent by a client, planning the shortest driving route according to the destination of the user and the current location of the user carried in the berth planning request;

[0105] The client is the device or software used by the user, such as a smartphone application, through which the user initiates a parking space planning request. A parking space planning request is a request initiated by the user from the client to ask the system to plan a parking space search and navigation for the user. The destination is the final location that the user wants to reach. The current location is the user's current geographical location. The shortest driving path refers to the route with the shortest distance or the shortest estimated driving time from the current location to the destination, calculated by the map service API.

[0106] In some embodiments, when a user uses the "Smart Parking" application to plan a parking space for a certain commercial center, the user opens the "Smart Parking" application and enters the destination (e.g., XXX shopping center) on the interface. The client interface inputs Figure 2 After the input is completed, the application automatically obtains the user's current GPS location, and when the user submits a berth planning request, the destination and current location information are sent to the server. After receiving the berth planning request, the server first verifies the destination and current location information contained in the request, and uses the built-in map service API to calculate the shortest driving path from the user's current location to the destination.

[0107] It should be noted that the built-in map service API can be a service provided by any map service provider, based on which the shortest driving route can be automatically planned.

[0108] S102, obtaining a parking space video of a sub-section of the shortest driving path at a preset distance from the destination from a high-position video monitoring system;

[0109] Among them, the high-position video surveillance system refers to a surveillance camera system installed at a higher position (such as street lights, building tops, etc.), which is used to monitor the status of road parking spaces below in real time. Sub-segment refers to a part of the shortest driving path, which refers to a section of the path from the user's current location to the destination. The preset distance refers to a specific distance range set at the user's destination, and a suitable parking space is found within this range. Parking space video refers to the real-time video stream of road parking spaces captured by the high-position video surveillance system, which can be used to analyze the occupancy of parking spaces.

[0110] In some embodiments of the present application, a preset distance from a destination in the shortest driving path is determined, and a road section with the preset distance is determined as a sub-road section, such as a road section within 100 meters of a specific destination as a sub-road section. A real-time parking video of the sub-road section is obtained from a high-position video monitoring system.

[0111] S103, identifying a plurality of vacant road parking spaces on the sub-road segment according to the parking space video of the sub-road segment; wherein the plurality of vacant road parking spaces are obtained by analyzing a key frame vector and a description text vector of each key frame according to a pre-trained vacant road parking space determination model, the key frame vector of each key frame is generated based on a plurality of key frames containing clear parking space images extracted from the parking space video, the description text vector is generated based on parking space description information of each key frame, and the pre-trained vacant road parking space determination model is a mathematical model for determining whether a road parking space on the sub-road segment is vacant;

[0112] In some embodiments of the present application, the specific process of identifying multiple vacant road parking spaces on a sub-road section based on the parking space video of the sub-road section includes: extracting multiple key frames containing clear images of parking spaces on the sub-road section from the parking space video on the sub-road section; obtaining text description information corresponding to each key frame to obtain parking space description information of each key frame; performing motion correction, time calibration and spatial normalization processing on each key frame to obtain a key frame vector for each key frame; using a Word2Vec network or a BERT network to vectorize the parking space description information of each key frame to obtain a description text vector for each key frame; inputting the key frame vector and its description text vector of each key frame into a pre-trained vacant road parking space determination model, and outputting multiple vacant road parking spaces corresponding to the sub-road section.

[0113] Among them, the parking space video on the sub-segment refers to the continuous video stream obtained from the high-position video surveillance system for a specific sub-segment, which contains real-time images of road parking spaces. Keyframes refer to frames containing important information, and refer to frames containing parking space status information. Text description information refers to metadata or annotations associated with the video content, which may include the location, type, and restrictions of the parking space. Motion correction is used to eliminate motion blur caused by camera movement or vehicle entry / exit of parking spaces. Time calibration is used to ensure that the time tags of video frames are accurate and synchronized with other data sources. Spatial normalization is used to adjust the resolution or size of video frames to meet the requirements of model input. Keyframe vector is a numerical representation of preprocessed keyframe data. Word2Vec is a model for predicting word sequences that can convert words into vectors. BERT is a pre-trained deep learning model for understanding the context of natural language. Description text vector is a numerical representation of text information obtained after processing using models such as Word2Vec or BERT.

[0114] In an embodiment of the present application, by extracting key frames from the parking space video of the sub-section and combining it with text description information, the method can accurately capture the real-time status of the parking space. By using advanced natural language processing techniques, such as Word2Vec or BERT, the text description is converted into a vector form, which not only improves the efficiency of information processing, but also enhances the model's understanding of the parking space status. By performing motion correction, time calibration and spatial normalization on the key frames, the quality and consistency of the data are ensured, providing accurate input for the idle road parking space determination model. Ultimately, the model can effectively identify idle parking spaces and provide users with real-time updated parking space information, thereby significantly improving parking efficiency, reducing the time to find parking spaces, and providing great convenience for users.

[0115] Among them, the pre-trained idle road parking space determination model includes a spatial feature extraction module, a text feature extraction module, a spatial feature and text feature fusion module and a parking space prediction module.

[0116] In some embodiments of the present application, the key frame vector of each key frame and its description text vector are input into a pre-trained vacant road parking space determination model, and the specific process of outputting multiple vacant road parking spaces corresponding to the sub-section includes: a spatial feature extraction module performs spatial feature extraction processing on the key frame vector of each key frame to obtain spatial features corresponding to the key frame vector of each key frame; a text feature extraction module performs text feature extraction processing on the description text vector of each key frame to obtain text features corresponding to the description text vector of each key frame; a spatial feature and text feature fusion module performs feature fusion processing on the spatial features corresponding to the key frame vector of each key frame and the text features corresponding to the description text vector of each key frame to obtain the frame-text fusion features of each key frame; and the frame-text fusion features of each key frame are input into the parking space prediction module to output multiple vacant road parking spaces corresponding to the sub-section.

[0117] Among them, the expression of feature fusion processing is:

[0118]

[0119] in, For the The key frame vector of the key frame corresponds to the Spatial features With The description text vector of the key frame corresponds to the Text features The frame-text fusion features obtained by fusion.

[0120] In some embodiments of the present application, a spatial feature extraction process is performed on the key frame vector of each key frame to obtain the specific process of the spatial features corresponding to the key frame vector of each key frame, including: extracting the data features of the key frame vector of each key frame; extracting the spatial features of the data features in the time dimension; using the spatial vector corresponding to each spatial point in the spatial features as a network node, and using the spatial connection information in the data features as the edge of the network; constructing a node graph based on the network nodes and the edges of the network; performing a linear transformation on the attributes of each node in the node graph to map the information of different nodes to the same vector space to obtain the basic vector representation of each key frame; and performing feature aggregation on the basic vector representation of each key frame to obtain the spatial features corresponding to the key frame vector of each key frame.

[0121] In an embodiment of the present application, a pre-trained idle road parking space determination model is generated according to the following steps, including: obtaining a historical parking space video within a preset period on any road section, and extracting a plurality of historical key frames containing clear parking space images in the historical parking space video; for each historical key frame, generating text description information including the location, size, color and whether there is a vehicle parked status information of the parking space, and obtaining the historical parking space description information of each historical key frame; generating a model training set based on each historical key frame and its historical parking space description information; creating and aggregating a spatial feature extraction module, a text feature extraction module, a spatial feature and text feature fusion module and a parking space prediction module to obtain an idle road parking space determination model; inputting the model training set into the idle road parking space determination model and outputting a model loss value; when the model loss value reaches a minimum, obtaining the pre-trained idle road parking space determination model; or when the model loss value does not reach a minimum, back-propagating the model loss value to update the model parameters of the idle road parking space determination model, and continuing to perform the step of inputting the model training set into the idle road parking space determination model until the model loss value reaches a minimum.

[0122] Among them, the preset period refers to collecting data within a specific time range, such as a certain period of time in a day or a certain day of a week. Historical berth video refers to video data related to berths recorded in a certain period of time in the past. Historical key frames are representative single image frames extracted from historical berth videos, which usually contain important information about berths. Historical berth description information is a text description of the specific features of the berth in each historical key frame (such as location, size, color, and vehicle parking status). The model training set is a data set used to train machine learning models, including image data of historical key frames and corresponding historical berth description information. The model loss value is used to measure the difference between model predictions and actual data in machine learning.

[0123] Specifically, the data table of historical berth description information of each historical key frame is as follows: Figure 3As shown, the keyframe identification is a unique identifier for each historical keyframe, such as A, B, and C. The parking space location describes the geographical location or area where the parking space is located, such as the city center square, the entrance to the hospital parking lot, or the west side of the commercial street. The parking space size describes the size of the vehicle that the parking space can accommodate, such as small, medium, or large. The parking space color provides color information for the parking space, which may be related to the city's parking regulations, such as blue, green, or white. The vehicle parking status indicates whether the parking space is occupied by a vehicle when the keyframe is captured.

[0124] In an embodiment of the present application, by acquiring historical parking videos within a specific time period and extracting key frames from them, the system is able to collect rich data reflecting the usage of parking spaces. Generating a detailed text description including parking space features and vehicle parking status not only provides additional contextual information for each key frame, but also enhances the model's understanding of the parking space status. Combining spatial feature extraction and text feature extraction, as well as further feature fusion, the system is able to construct a free road parking space determination model that comprehensively considers multi-source information. Through continuous training and parameter optimization, the model can predict the parking space status more and more accurately, and ultimately achieve high-precision parking space vacancy determination.

[0125] In some embodiments of the present application, based on each historical key frame and its historical berth description information, the specific process of generating a model training set includes: performing motion correction, time calibration and spatial normalization processing on each historical key frame to obtain a historical key frame vector for each historical key frame; using a Word2Vec network or a BERT network to vectorize the historical berth description information of each historical key frame to obtain a historical description text vector for each historical key frame; for each historical key frame vector and its historical description text vector, receiving an input real label for characterizing whether a road berth in each historical key frame is vacant for data calibration to obtain a model training set; wherein the model loss function of the pre-trained vacant road berth determination model is:

[0126]

[0127] in, is the loss value, is the true label of the model training data, is the predicted label predicted by the idle road parking space determination model, It is obtained by model processing based on a model training data.

[0128] S104, identifying an optimal target parking space from a plurality of idle road parking spaces; wherein the optimal target parking space is an idle road parking space with a highest matching score with the user, and the idle road parking space with a highest matching score with the user is obtained by performing data analysis and processing on parking space identity information of each idle road parking space according to the parking demand information of the user, and the parking space identity information of each idle road parking space is the size, type, fee information and GPS coordinates of each idle road parking space collected;

[0129] In some embodiments of the present application, the specific process of identifying the optimal target parking space from multiple idle road parking spaces includes: collecting the size, type, fee information and GPS coordinates of each idle road parking space to obtain the parking space identity information of each idle road parking space; according to the user's identity identifier, obtaining the corresponding historical parking data from a pre-established mapping relationship between the identity identifier and the historical parking data; based on the historical parking data, analyzing the user's parking demand information; according to the user's parking demand information, performing data analysis and processing on the parking space identity information of each idle road parking space to obtain a matching score between each idle road parking space and the user; and determining the idle road parking space with the highest matching score with the user as the optimal target parking space.

[0130] Among them, the parking space identity information includes the size, type, fee information and GPS coordinates of the parking space, which are used to uniquely identify and describe the characteristics of each parking space. The identity identifier is the unique identifier of the user, which may refer to the user's account, mobile phone number or other information used to identify the user. Historical parking data is a record of the user's past parking behavior, including parking location, parking duration, selected parking space type and other information. Parking demand information is based on the user's historical parking data, and the analysis results show the user's preferences and needs for parking spaces, such as frequently visited locations, preferred parking space types, etc. The matching score is an evaluation of each idle road parking space based on the user's parking demand information, and the degree of matching with the user's needs is obtained. The optimal target parking space is the idle road parking space with the highest matching score with the user, which is considered to be the parking space that best meets the user's current needs.

[0131] Specifically, the berth identity information of each idle road berth is, for example, Figure 4 As shown, it includes the size, type, fee information and GPS coordinates of the parking space. The user's parking demand information data table is as follows: Figure 5 As shown, it includes vehicle size information, user berth preference information, user payment preference information, and user destination GPS.

[0132] In an embodiment of the present application, by collecting detailed identity information of each free road parking space, including size, type, cost and location, the system is able to create a detailed profile for each parking space. Combined with the user's identity identification, the system can quickly extract the user's historical parking behavior from big data analysis, and then deeply understand the user's parking preferences. Through intelligent analysis of these historical data, the system can accurately infer the user's parking demand information. Using this information, the system performs personalized data analysis on each free parking space, calculates the matching score with the user's needs, and intelligently recommends the best parking space. This method not only improves the utilization rate of parking spaces, but also significantly improves the user's parking experience. By reducing the time to find parking spaces, it reduces traffic congestion, and realizes efficient allocation of resources and personalized services for users.

[0133] In some embodiments of the present application, the specific process of performing data analysis and processing on the parking space identity information of each idle road parking space according to the parking demand information of the user to obtain the matching score between each idle road parking space and the user includes: calculating the first matching degree between the parking space size and the user's vehicle according to the size of each idle road parking space and the vehicle size information; wherein,

[0134] First matching degree The calculation formula is:

[0135]

[0136] in, is an adjustment factor used to ensure that the berth size is at least a certain proportion of the vehicle size; where is a function that returns the minimum of two values;

[0137] Calculating a second matching degree between the berth type and the user's berth preference according to the type of each idle road berth and the user's berth preference information;

[0138] Second matching degree The calculation formula is:

[0139]

[0140] Calculating a third matching degree between parking space fees and user payment according to fee information of each idle road parking space and user payment preference information;

[0141] The third matching The calculation formula is:

[0142]

[0143] Calculating a fourth matching degree between the geographic location of each free road parking space and the user's destination based on the GPS coordinates of each free road parking space and the user's destination GPS;

[0144] Fourth matching degree The calculation formula is:

[0145]

[0146] in, is the attenuation factor, which is used to adjust the effect of distance on matching degree. is the exponent of the constant e, which is 2.71828;

[0147] The first matching degree, the second matching degree, the third matching degree and the fourth matching degree are weighted and summed to obtain a matching score between each free road parking space and the user; wherein,

[0148]

[0149] Among them, S is the matching score between each idle road parking space and the user, , , , are weight values ​​and the sum is 1, is the first matching degree between the berth size and the user's vehicle, is the second matching degree between the berth type and the user's berth preference, The third matching degree for berth fee and user payment, It is the fourth matching degree between the berth geographical location and the user's destination.

[0150] In some embodiments of the present application, the process of analyzing the user's parking demand information based on historical parking data includes: extracting the user's vehicle size information from the historical parking data; analyzing the berth type selected by the user and the user's preference for specific berth features based on the historical parking data to obtain the user's berth preference information; analyzing the user's parking fee payment records and payment amounts based on the historical parking data to obtain the user's payment preference information; obtaining the GPS of the destination; and using the user's vehicle size information, the user's berth preference information, the user's payment preference information, and the user's destination GPS as the user's parking demand information.

[0151] S105, taking the location of the target parking space as the end point, starting the navigation service to generate navigation information and feeding it back to the client, and continuing to execute the step of obtaining parking space videos of sub-sections at a preset distance from the destination in the shortest driving path to perform dynamic road parking planning until the user reaches the target parking space.

[0152] Among them, the target parking space refers to the optimal parking space recommended by the system for the user, which is determined based on the user's parking needs and real-time parking space information. Navigation service refers to a service that can provide route guidance from the current location to the target location. Navigation information refers to the guidance information generated by the navigation service, including route, direction, distance, etc., to help users reach their destination. Road parking space dynamic planning refers to the process of dynamically planning parking spaces for users based on real-time parking space information.

[0153] In some embodiments, after determining that the location of the target parking space is the destination, the server sets its location as the navigation destination. The system starts the navigation service through the built-in map service or the third-party map API. The navigation service generates detailed navigation information, including turn instructions, distance, and estimated arrival time. The system sends the navigation information to the user's "Smart Parking" application, and the user can start the vehicle's navigation system or mobile phone application to follow the navigation. While the user is heading to the target parking space, the system continues to monitor the sub-section parking space video within a preset distance of the destination. If the system finds that the target parking space is occupied or a more suitable parking space appears, it will re-plan and update the navigation information to ensure that the user can reach a new available target parking space. The user drives according to the real-time updated navigation information until he reaches the final target parking space.

[0154] For example Figure 6 As shown, Figure 6 This is a schematic diagram of an application scenario provided by the present application. A user sends a parking planning request to a server through a client. The server responds to the parking planning request sent by the client, and plans the shortest driving route according to the destination of the user and the current location of the user carried in the parking planning request; obtains a parking video of a sub-section of the shortest driving route at a preset distance from the destination from a high-position video surveillance system; identifies multiple idle road parking spaces on the sub-section based on the parking video of the sub-section; identifies the optimal target parking space from the multiple idle road parking spaces; starts a navigation service with the location of the target parking space as the end point to generate navigation information and feedback it to the client, and continues to execute the step of obtaining a parking video of a sub-section of the shortest driving route at a preset distance from the destination to perform dynamic road parking planning.

[0155] In the embodiments of the present application, on the one hand, the present application extracts key frames from a continuous video stream, and uses a deep learning model to analyze the key frame vectors and their description text vectors, so as to efficiently and accurately determine the vacancy status of the parking space. At the same time, it can continuously track the user's driving status and the real-time changes of the parking space, and dynamically update the optimal parking space recommendation until the user successfully parks the vehicle. This method not only greatly improves the speed and accuracy of the parking space search, but also effectively alleviates urban traffic congestion by reducing the time wasted in the search process. On the other hand, the present application adopts a flexible personalized service mechanism. The system's recommendation of vacant parking spaces not only takes into account the physical properties of the parking space, such as size and type, but also takes into account the geographical distance from the user's destination and the user's willingness to pay. Through this personalized data analysis and processing, the system can recommend the most matching parking space to the user, thereby providing a more intimate and convenient parking experience, allowing the user to quickly and accurately park the vehicle at a parking space near the destination.

[0156] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0157] See also Figure 7 , which shows a schematic diagram of the structure of a road berth dynamic planning device for smart transportation provided by an exemplary embodiment of the present application. The road berth dynamic planning device for smart transportation can be implemented as all or part of an electronic device through software, hardware or a combination of both. The device 1 includes a shortest driving path planning module 10, a berth video acquisition module 20, an idle road berth identification module 30, an optimal target berth identification module 40, and a navigation service startup module 50.

[0158] The shortest driving route planning module 10 is used to respond to the berth planning request sent by the client, and plan the shortest driving route according to the destination of the user and the current location of the user carried in the berth planning request;

[0159] The parking space video acquisition module 20 is used to acquire the parking space video of the sub-section at a preset distance from the destination in the shortest driving path from the high-position video monitoring system;

[0160] The idle road parking space identification module 30 is used to identify multiple idle road parking spaces on the sub-road section according to the parking space video of the sub-road section; wherein the multiple idle road parking spaces are obtained by analyzing the key frame vector and the description text vector of each key frame according to the pre-trained idle road parking space determination model, the key frame vector of each key frame is generated based on multiple key frames containing clear parking space images extracted from the parking space video, the description text vector is generated based on the parking space description information of each key frame, and the pre-trained idle road parking space determination model is a mathematical model for determining whether the road parking space on the sub-road section is idle;

[0161] The optimal target parking space identification module 40 is used to identify the optimal target parking space from a plurality of idle road parking spaces; wherein the optimal target parking space is the idle road parking space with the highest matching score with the user, and the idle road parking space with the highest matching score with the user is obtained by performing data analysis and processing on the parking space identity information of each idle road parking space according to the parking demand information of the user, and the parking space identity information of each idle road parking space is the size, type, fee information and GPS coordinates of each idle road parking space collected;

[0162] The navigation service start module 50 is used to start the navigation service with the location of the target parking space as the end point, generate navigation information and feed it back to the client, and continue to execute the step of obtaining the parking space video of the sub-section at a preset distance from the destination in the shortest driving path to perform dynamic road parking planning until the user reaches the target parking space.

[0163] It should be noted that the road berth dynamic planning device for smart transportation provided in the above embodiment only uses the division of the above functional modules as an example when executing the road berth dynamic planning method for smart transportation. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the road berth dynamic planning device for smart transportation provided in the above embodiment and the road berth dynamic planning method for smart transportation belong to the same concept. The implementation process is detailed in the method embodiment, which will not be repeated here.

[0164] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0165] In the embodiments of the present application, on the one hand, the present application extracts key frames from a continuous video stream, and uses a deep learning model to analyze the key frame vectors and their description text vectors, so as to efficiently and accurately determine the vacancy status of the parking space. At the same time, it can continuously track the user's driving status and the real-time changes of the parking space, and dynamically update the optimal parking space recommendation until the user successfully parks the vehicle. This method not only greatly improves the speed and accuracy of the parking space search, but also effectively alleviates urban traffic congestion by reducing the time wasted in the search process. On the other hand, the present application adopts a flexible personalized service mechanism. The system's recommendation of vacant parking spaces not only takes into account the physical properties of the parking space, such as size and type, but also takes into account the geographical distance from the user's destination and the user's willingness to pay. Through this personalized data analysis and processing, the system can recommend the most matching parking space to the user, thereby providing a more intimate and convenient parking experience, allowing the user to quickly and accurately park the vehicle at a parking space near the destination.

[0166] The present application also provides a computer-readable medium having program instructions stored thereon, which, when executed by a processor, implement the road berth dynamic planning method for smart transportation provided by the above-mentioned various method embodiments.

[0167] The present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the road berth dynamic planning method for intelligent transportation of each of the above-mentioned method embodiments.

[0168] See also Figure 8 , is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 8 As shown, the electronic device 1000 may include: at least one processor 1001 , at least one network interface 1004 , a user interface 1003 , a memory 1005 , and at least one communication bus 1002 .

[0169] The communication bus 1002 is used to realize the connection and communication between these components.

[0170] The user interface 1003 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 1003 may also include a standard wired interface and a wireless interface.

[0171] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0172] Among them, the processor 1001 may include one or more processing cores. The processor 1001 uses various interfaces and lines to connect various parts of the entire electronic device 1000, and executes various functions and processes data of the electronic device 1000 by running or executing instructions, programs, code sets or instruction sets stored in the memory 1005, and calling data stored in the memory 1005. Optionally, the processor 1001 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 1001 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 1001, and it can be implemented separately through a chip.

[0173] Among them, the memory 1005 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 1005 includes a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 1005 may also be optionally at least one storage system located away from the aforementioned processor 1001. As Figure 8 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a road berth dynamic planning application for smart transportation.

[0174] exist Figure 8In the electronic device 1000 shown, the user interface 1003 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 1001 can be used to call the road berth dynamic planning application for smart transportation stored in the memory 1005, and specifically perform the following operations:

[0175] In response to the berth planning request sent by the client, the shortest driving route is planned according to the destination of the user and the current location of the user carried in the berth planning request;

[0176] Obtaining parking videos of sub-sections at a preset distance from the destination in the shortest driving path from a high-position video surveillance system;

[0177] According to the parking space video of the sub-road section, a plurality of idle road parking spaces on the sub-road section are identified; wherein the plurality of idle road parking spaces are obtained by analyzing the key frame vector and the description text vector of each key frame according to the pre-trained idle road parking space determination model, the key frame vector of each key frame is generated based on a plurality of key frames containing clear parking space images extracted from the parking space video, the description text vector is generated based on the parking space description information of each key frame, and the pre-trained idle road parking space determination model is a mathematical model for determining whether a road parking space on the sub-road section is idle;

[0178] Identify the best target parking space from multiple idle road parking spaces; wherein the best target parking space is the idle road parking space with the highest matching score with the user, and the idle road parking space with the highest matching score with the user is obtained by performing data analysis and processing on the parking space identity information of each idle road parking space according to the parking demand information of the user, and the parking space identity information of each idle road parking space is the size, type, fee information and GPS coordinates of each idle road parking space collected;

[0179] Taking the location of the target parking space as the end point, the navigation service is started to generate navigation information and feedback it to the client, and the step of obtaining the parking space video of the sub-section at a preset distance from the destination in the shortest driving path is continued to perform dynamic planning of road parking spaces until the user reaches the target parking space.

[0180] In one embodiment, when the processor 1001 identifies a plurality of free road parking spaces on a sub-road segment according to the parking space video of the sub-road segment, the processor 1001 specifically performs the following operations:

[0181] Extracting multiple key frames containing clear images of the parking spaces on the sub-road section from the parking space video on the sub-road section;

[0182] Obtain the text description information corresponding to each key frame, and obtain the berth description information of each key frame;

[0183] Perform motion correction, time calibration and spatial normalization on each key frame to obtain the key frame vector of each key frame;

[0184] Use Word2Vec network or BERT network to vectorize the berth description information of each key frame to obtain the description text vector of each key frame;

[0185] The key frame vector of each key frame and its description text vector are input into a pre-trained idle road parking space determination model, and a plurality of idle road parking spaces corresponding to the sub-road section are output.

[0186] In one embodiment, when the processor 1001 inputs the key frame vector and the description text vector of each key frame into the pre-trained free road parking space determination model and outputs a plurality of free road parking spaces corresponding to the sub-road section, the processor 1001 specifically performs the following operations:

[0187] The spatial feature extraction module performs spatial feature extraction processing on the key frame vector of each key frame to obtain the spatial features corresponding to the key frame vector of each key frame;

[0188] The text feature extraction module performs text feature extraction processing on the description text vector of each key frame to obtain the text features corresponding to the description text vector of each key frame;

[0189] The spatial feature and text feature fusion module performs feature fusion processing on the spatial feature corresponding to the key frame vector of each key frame and the text feature corresponding to the description text vector of each key frame to obtain the frame-text fusion feature of each key frame;

[0190] The frame-text fusion feature of each key frame is input into the parking space prediction module, and multiple vacant road parking spaces corresponding to the sub-section are output; wherein, the expression of feature fusion processing is:

[0191]

[0192] in, For the The key frame vector of the key frame corresponds to the Spatial features With The description text vector of the key frame corresponds to the Text features The frame-text fusion features obtained by fusion.

[0193] In one embodiment, when the processor 1001 performs spatial feature extraction processing on the key frame vector of each key frame to obtain the spatial feature corresponding to the key frame vector of each key frame, the processor 1001 specifically performs the following operations:

[0194] Extracting data features of a key frame vector of each key frame;

[0195] Extract the spatial features of data features in the time dimension;

[0196] The spatial vector corresponding to each spatial point in the spatial feature is used as a network node, and the spatial connection information in the data feature is used as the edge of the network;

[0197] Based on network nodes and network edges, a node graph is constructed;

[0198] Perform linear transformation on the attributes of each node in the node graph to map the information of different nodes into the same vector space and obtain the basic vector representation of each key frame;

[0199] Feature aggregation is performed on the basis vector representation of each key frame to obtain the spatial features corresponding to the key frame vector of each key frame.

[0200] In one embodiment, when the processor 1001 generates a pre-trained free road parking space determination model, the processor 1001 specifically performs the following operations:

[0201] Obtain historical berth videos within a preset period on any road section, and extract multiple historical key frames containing clear images of berths in the historical berth videos;

[0202] For each historical key frame, generate text description information including the location, size, color and whether there is a vehicle parked at the berth, and obtain historical berth description information of each historical key frame;

[0203] Generate a model training set based on each historical key frame and its historical berth description information;

[0204] Create and aggregate a spatial feature extraction module, a text feature extraction module, a spatial feature and text feature fusion module, and a berth prediction module to obtain an idle road berth determination model;

[0205] Input the model training set into the idle road parking space determination model and output the model loss value;

[0206] When the model loss value reaches the minimum, the pre-trained vacant road parking space determination model is obtained; or when the model loss value does not reach the minimum, the model loss value is back-propagated to update the model parameters of the vacant road parking space determination model, and the step of inputting the model training set into the vacant road parking space determination model is continued until the model loss value reaches the minimum.

[0207] In one embodiment, when the processor 1001 generates a model training set based on each historical key frame and its historical berth description information, the processor 1001 specifically performs the following operations:

[0208] Perform motion correction, time calibration and spatial normalization on each historical key frame to obtain a historical key frame vector for each historical key frame;

[0209] Use the Word2Vec network or the BERT network to vectorize the historical berth description information of each historical key frame to obtain the historical description text vector of each historical key frame;

[0210] For each historical key frame vector and its historical description text vector, the input real label used to characterize whether the road berth in each historical key frame is vacant is received for data calibration to obtain a model training set; wherein,

[0211] The model loss function of the pre-trained idle road parking space determination model is:

[0212]

[0213] in, is the loss value, is the true label of the model training data, is the predicted label predicted by the idle road parking space determination model, It is obtained by model processing based on a model training data.

[0214] In one embodiment, when the processor 1001 identifies the optimal target parking space from a plurality of free road parking spaces, the processor 1001 specifically performs the following operations:

[0215] Collect the size, type, fee information and GPS coordinates of each free road berth to obtain the berth identity information of each free road berth;

[0216] According to the user's identity, obtaining corresponding historical parking data from a pre-established mapping relationship between the identity and the historical parking data;

[0217] Analyze users’ parking demand information based on historical parking data;

[0218] According to the user's parking demand information, the parking space identity information of each idle road parking space is analyzed and processed to obtain a matching score between each idle road parking space and the user;

[0219] The vacant road parking space with the highest matching score with the user is determined as the optimal target parking space.

[0220] In one embodiment, when the processor 1001 analyzes the user's parking demand information based on historical parking data, the processor 1001 specifically performs the following operations:

[0221] Extract the user’s vehicle size information from historical parking data;

[0222] Based on historical parking data, analyze the parking space type selected by the user and the user's preference for specific parking space features to obtain the user's parking space preference information;

[0223] Based on historical parking data, analyze the user's parking fee payment records and payment amounts to obtain the user's payment preference information; obtain the GPS of the destination;

[0224] The user's vehicle size information, the user's parking space preference information, the user's payment preference information, and the user's destination GPS are used as the user's parking demand information.

[0225] In the embodiments of the present application, on the one hand, the present application extracts key frames from a continuous video stream, and uses a deep learning model to analyze the key frame vectors and their description text vectors, so as to efficiently and accurately determine the vacancy status of the parking space. At the same time, it can continuously track the user's driving status and the real-time changes of the parking space, and dynamically update the optimal parking space recommendation until the user successfully parks the vehicle. This method not only greatly improves the speed and accuracy of the parking space search, but also effectively alleviates urban traffic congestion by reducing the time wasted in the search process. On the other hand, the present application adopts a flexible personalized service mechanism. The system's recommendation of vacant parking spaces not only takes into account the physical properties of the parking space, such as size and type, but also takes into account the geographical distance from the user's destination and the user's willingness to pay. Through this personalized data analysis and processing, the system can recommend the most matching parking space to the user, thereby providing a more intimate and convenient parking experience, allowing the user to quickly and accurately park the vehicle at a parking space near the destination.

[0226] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program for dynamic planning of road berths for smart transportation can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium of the program for dynamic planning of road berths for smart transportation can be a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0227] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A road berth dynamic planning method for smart transportation, characterized in that: Applied to the server, the method includes: In response to a berth planning request sent by a client, planning the shortest driving route according to the destination of the user and the current location of the user carried in the berth planning request; Obtaining, from a high-position video surveillance system, a berth video of a sub-section of the shortest driving path at a preset distance from the destination; According to the parking space video of the sub-road section, a plurality of vacant road parking spaces on the sub-road section are identified; wherein the plurality of vacant road parking spaces are obtained by analyzing the key frame vector and the description text vector of each key frame according to a pre-trained vacant road parking space determination model, wherein the key frame vector of each key frame is generated based on a plurality of key frames containing clear parking space images extracted from the parking space video, and the description text vector is generated based on the parking space description information of each key frame, and the pre-trained vacant road parking space determination model is a mathematical model for determining whether the road parking spaces on the sub-road section are vacant; wherein, The step of identifying a plurality of idle road parking spaces on the sub-road section according to the parking space video of the sub-road section comprises: Extracting a plurality of key frames containing clear images of the berths on the sub-road section from the berth video on the sub-road section; Obtain the text description information corresponding to each key frame, and obtain the berth description information of each key frame; Performing motion correction, time calibration and spatial normalization processing on each key frame to obtain a key frame vector of each key frame; Using a Word2Vec network or a BERT network, vectorizing the berth description information of each key frame to obtain a description text vector of each key frame; The key frame vector of each key frame and its description text vector are input into a pre-trained free road parking space determination model, and a plurality of free road parking spaces corresponding to the sub-road section are output; wherein, The pre-trained idle road berth determination model includes a spatial feature extraction module, a text feature extraction module, a spatial feature and text feature fusion module and a berth prediction module; The step of inputting the key frame vector of each key frame and its description text vector into a pre-trained idle road parking space determination model, and outputting a plurality of idle road parking spaces corresponding to the sub-road section, comprises: The spatial feature extraction module performs spatial feature extraction processing on the key frame vector of each key frame to obtain the spatial feature corresponding to the key frame vector of each key frame; The text feature extraction module performs text feature extraction processing on the description text vector of each key frame to obtain text features corresponding to the description text vector of each key frame; The spatial feature and text feature fusion module performs feature fusion processing on the spatial feature corresponding to the key frame vector of each key frame and the text feature corresponding to the description text vector of each key frame to obtain the frame-text fusion feature of each key frame; Inputting the frame-text fusion feature of each key frame into the parking space prediction module, and outputting a plurality of vacant road parking spaces corresponding to the sub-road section; Identify the best target parking space from the plurality of idle road parking spaces; wherein the best target parking space is the idle road parking space with the highest matching score with the user, and the idle road parking space with the highest matching score with the user is obtained by performing data analysis and processing on parking space identity information of each idle road parking space according to the parking demand information of the user, and the parking space identity information of each idle road parking space is the size, type, fee information and GPS coordinates of each idle road parking space collected; Taking the position of the target parking space as the end point, the navigation service is started to generate navigation information and fed back to the client, and the step of obtaining the parking space video of the sub-section at a preset distance from the destination in the shortest driving path is continued to perform dynamic road parking planning until the user reaches the target parking space.

2. The method according to claim 1, characterized in that The expression of fusion processing is: in, For the The key frame vector of the key frame corresponds to the Spatial features With The description text vector of the key frame corresponds to the Text features The frame-text fusion features obtained by fusion.

3. The method according to claim 1, characterized in that: The performing of spatial feature extraction processing on the key frame vector of each key frame to obtain the spatial feature corresponding to the key frame vector of each key frame includes: Extracting data features of a key frame vector of each key frame; Extracting spatial features of the data features in the time dimension; The spatial vector corresponding to each spatial point in the spatial feature is used as a network node, and the spatial connection information in the data feature is used as an edge of the network; Constructing a node graph based on the network nodes and the network edges; Performing a linear transformation on the attributes of each node in the node graph to map the information of different nodes into the same vector space, thereby obtaining a basic vector representation of each key frame; Feature aggregation is performed on the basic vector representation of each key frame to obtain the spatial features corresponding to the key frame vector of each key frame.

4. The method according to claim 1, characterized in that: The following steps are used to generate a pre-trained free road parking space determination model, including: Acquire a historical berth video within a preset period on any road section, and extract a plurality of historical key frames containing clear images of berths in the historical berth video; For each historical key frame, generate text description information including the location, size, color and whether there is a vehicle parked at the berth, and obtain historical berth description information of each historical key frame; Generate a model training set based on each historical key frame and its historical berth description information; Create and aggregate a spatial feature extraction module, a text feature extraction module, a spatial feature and text feature fusion module, and a berth prediction module to obtain an idle road berth determination model; Inputting the model training set into the idle road berth determination model, and outputting a model loss value; When the model loss value reaches a minimum, a pre-trained vacant road parking space determination model is obtained; or when the model loss value does not reach a minimum, the model loss value is back-propagated to update the model parameters of the vacant road parking space determination model, and the step of inputting the model training set into the vacant road parking space determination model is continued until the model loss value reaches a minimum.

5. The method according to claim 4, characterized in that The method of generating a model training set based on each historical key frame and its historical berth description information includes: Performing motion correction, time calibration and space normalization processing on each of the historical key frames to obtain a historical key frame vector of each of the historical key frames; Using a Word2Vec network or a BERT network, vectorizing the historical berth description information of each historical key frame to obtain a historical description text vector of each historical key frame; For each historical key frame vector and its historical description text vector, a real label inputted to characterize whether a road berth is vacant in each historical key frame is received for data calibration to obtain a model training set; wherein, The model loss function of the pre-trained idle road parking space determination model is: in, is the loss value, is the true label of the model training data, is the predicted label predicted by the idle road berth determination model, It is obtained by model processing based on a model training data.

6. The method according to claim 1, characterized in that The step of identifying the optimal target parking space from the plurality of idle road parking spaces comprises: Collect the size, type, fee information and GPS coordinates of each free road berth to obtain the berth identity information of each free road berth; According to the user's identity, obtaining corresponding historical parking data from a pre-established mapping relationship between the identity and the historical parking data; Analyzing the user's parking demand information based on the historical parking data; According to the parking demand information of the user, data analysis and processing are performed on the parking space identity information of each idle road parking space to obtain a matching score between each idle road parking space and the user; The vacant road parking space with the highest matching score with the user is determined as the optimal target parking space.

7. The method according to claim 6, characterized in that The user's parking demand information includes the user's vehicle size information, the user's parking space preference information, the user's payment preference information, and the user's destination GPS; The step of performing data analysis and processing on the parking space identity information of each idle road parking space according to the parking demand information of the user to obtain a matching score between each idle road parking space and the user includes: According to the size of each idle road parking space and the vehicle size information, a first matching degree between the parking space size and the user vehicle is calculated; wherein, The first matching degree The calculation formula is: in, is an adjustment factor used to ensure that the berth size is at least a certain proportion of the vehicle size; where is a function that returns the minimum of two values; Calculating a second matching degree between the berth type and the user's berth preference according to the type of each idle road berth and the user's berth preference information; The second matching degree The calculation formula is: Calculating a third matching degree between parking space fees and user payment according to the fee information of each idle road parking space and the user payment preference information; The third matching degree The calculation formula is: Calculating a fourth matching degree between the geographic location of the parking space and the user destination according to the GPS coordinates of each idle road parking space and the user destination GPS; The fourth matching degree The calculation formula is: in, is the attenuation factor, which is used to adjust the effect of distance on matching. is the exponent of the constant e, which is 2.71828; The first matching degree, the second matching degree, the third matching degree and the fourth matching degree are weighted and summed to obtain a matching score between each vacant road parking space and the user; wherein, Where S is the matching score between each free road parking space and the user, , , , are weight values ​​and the sum is 1, is the first matching degree between the berth size and the user's vehicle, is the second matching degree between the berth type and the user's berth preference, The third matching degree for berth fee and user payment, It is the fourth matching degree between the berth geographical location and the user's destination.

8. The method according to claim 6, characterized in that The analyzing the user's parking demand information based on the historical parking data includes: Extracting the vehicle size information of the user from the historical parking data; Analyzing the parking space type selected by the user and the user's preference for specific parking space features according to the historical parking data to obtain the user's parking space preference information; Analyzing the user's parking fee payment records and payment amounts based on the historical parking data to obtain the user's payment preference information; Get the GPS of the destination; The user's vehicle size information, the user's parking space preference information, the user's payment preference information, and the user's destination GPS are used as the user's parking demand information.

9. A road berth dynamic planning device for intelligent transportation implemented based on the method described in any one of claims 1 to 8, characterized in that: The device comprises: The shortest driving route planning module is used to respond to the berth planning request sent by the client, and plan the shortest driving route according to the destination of the user carried in the berth planning request and the current location of the user; A berth video acquisition module, used to acquire a berth video of a sub-section of the shortest driving path at a preset distance from the destination from a high-position video monitoring system; an idle road parking space identification module, used for identifying a plurality of idle road parking spaces on the sub-road section according to the parking space video of the sub-road section; wherein the plurality of idle road parking spaces are obtained by analyzing a key frame vector and a description text vector of each key frame according to a pre-trained idle road parking space determination model, wherein the key frame vector of each key frame is generated based on a plurality of key frames containing clear parking space images extracted from the parking space video, and the description text vector is generated based on parking space description information of each key frame, and the pre-trained idle road parking space determination model is a mathematical model for determining whether a road parking space on the sub-road section is idle; an optimal target parking space identification module, for identifying an optimal target parking space from the plurality of idle road parking spaces; wherein the optimal target parking space is an idle road parking space with the highest matching score with the user, the idle road parking space with the highest matching score with the user is obtained by performing data analysis and processing on parking space identity information of each idle road parking space according to the parking demand information of the user, the parking space identity information of each idle road parking space being the size, type, fee information and GPS coordinates of each idle road parking space collected; The navigation service startup module is used to start the navigation service with the location of the target parking space as the end point, generate navigation information and feed it back to the client, and continue to execute the step of obtaining the parking space video of the sub-section of the shortest driving path at a preset distance from the destination to perform dynamic road parking planning until the user reaches the target parking space.

Images