CN116187602B - Parking space occupation prediction method for parking lot - Google Patents
Parking space occupation prediction method for parking lot Download PDFInfo
- Publication number
- CN116187602B CN116187602B CN202310490841.4A CN202310490841A CN116187602B CN 116187602 B CN116187602 B CN 116187602B CN 202310490841 A CN202310490841 A CN 202310490841A CN 116187602 B CN116187602 B CN 116187602B
- Authority
- CN
- China
- Prior art keywords
- parking
- data
- parking lot
- space occupation
- parking space
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 24
- 230000007246 mechanism Effects 0.000 claims abstract description 6
- 238000000547 structure data Methods 0.000 claims abstract description 5
- 239000013598 vector Substances 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 230000004927 fusion Effects 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G06Q50/40—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/586—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of parking space
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention belongs to the technical field of traffic, and relates to a parking space occupation prediction method for a parking lot, which comprises the steps of firstly collecting parking data of the parking lot; training a parking space occupation prediction model according to the collected parking data; finally, inputting the current parking space data into a trained parking space occupation prediction model to obtain a parking space occupation prediction interval, returning the result to a user through a corresponding APP, modeling the parking space structure data, and capturing potential relations between different areas through a GCN and an attention mechanism; and the parking space occupation prediction is carried out through the improved TFT, the relation between different types of inputs is captured, the final prediction interval is output for a user to make a decision, the improved TFT can accept and distinguish various heterogeneous inputs, meanwhile, the inputs are comprehensively utilized, long-distance dependence can be captured, and a vehicle owner can judge the prediction confidence according to the prediction interval, so that whether the vehicle owner goes to a target parking lot or not is determined.
Description
Technical Field
The invention belongs to the technical field of traffic, and relates to a parking lot parking space occupation prediction method, in particular to a parking lot parking space occupation prediction method based on Temporal Fusion Transformer (a transform-based deep neural network, called TFT for short) and GCN (graph convolution neural network).
Background
With the development of economy and the increase of urban population, vehicle data in cities are rapidly increased, and with the rapid increase of demand for parking spaces, people hope to find free parking spaces accurately and rapidly. However, the current parking data of the current parking lot is not opened to the public, and the parking space occupation information has timeliness, so that even if an idle parking space exists currently, the idle parking space still exists after the vehicle arrives, and the time is wasted; in addition, because the parking space occupation information is not easy to obtain and has uncertainty, partial idle parking spaces cannot be well utilized.
For the above problems, parking space occupation prediction is a mainstream solution. For a parking lot, the prediction data of the parking space occupation condition of each area of the parking lot in a certain time period (such as 15 minutes) are provided for a user in real time, the user can orderly arrange the travel and determine the parking position, so that the time for the user to find the parking space is reduced, and the utilization rate of the parking space of the parking lot is improved. A good parking space occupation prediction method can provide great convenience for vehicle owners and parking lots. Parking space occupation algorithms generally predict the occupation situation of a parking lot within a certain period of time by modeling information such as historical parking data, current parking data (current occupation situation, surrounding vehicle situation of the parking lot) and other auxiliary information (such as current date).
The existing parking space occupation prediction methods mainly comprise two types, namely a statistical-based method, the method generally adopts a regression idea, a parking space occupation curve is fitted by modeling historical data, so that the prediction of future data is realized, but the prediction step length of the method is limited, long-time dependency cannot be captured, and meanwhile, when the number of variables increases, the prediction precision of the methods is poor; the other type is a deep learning-based method, which generally uses a deep neural network as a main network (such as an LSTM) and uses historical parking data as input to predict the parking space occupation situation in a certain step length, but cannot model the data well, and directly splice the historical data as input, so that the obtained prediction model cannot capture the deep dependency relationship among various types of data, and meanwhile, the effect is further improved when facing long-distance dependence.
Disclosure of Invention
In order to solve the problems, the invention provides a novel parking space occupation prediction method of a parking lot, and the GCN is used for modeling the dependency relationship between each floor and each region of the parking lot, so that deep connection among different types of data is captured, and prediction of each floor and each region is facilitated; temporal Fusion Transformer is adopted to accept different types of input information, meanwhile, the output is a range value, and in practical application, one range value shows the current parking space condition to a user, so that the user can conveniently schedule.
In order to achieve the above purpose, the specific process of the invention for realizing the parking space occupation prediction is as follows:
s1, collecting parking data of a parking lot;
s2, training a parking space occupation prediction model according to the parking data collected in the step S1;
s3, inputting the current parking space data into a trained parking space occupation prediction model to obtain a parking space occupation prediction interval, and returning the result to the user through the corresponding APP.
As a further technical scheme of the present invention, the parking data in step S1 includes a parking lot structure, historical parking data, current parking data, road conditions around the parking lot and auxiliary information; wherein the parking lot structure comprises a parking lot, an adjacent relation among parking buildings and an adjacent relation among parking areas; the historical parking data contains the number of parks in each area at a certain moment in the history; the current parking data is the parking quantity of each area in the current time period; the historical parking data and the current parking data form parking data of a parking lot; the road vehicles around the parking lot are data of vehicles nearby the parking lot; the auxiliary information includes date information and holiday information.
As a further technical scheme of the invention, the specific process of the step S2 is as follows:
step S21, pair S1 collectPreprocessing parking data of a parking lot, vehicle data nearby the parking lot, date and holiday data at each moment to obtain each timeTime->Feature vector +.>,/>As the dimension of the feature vector,;
step S22, embedding the parking lot structure data through the GCN to obtain each momentParking lot structure embedded feature->;
Step S23, each time is setParking lot structure embedded feature->Along with the feature vector->And training a parking space occupation prediction model as input data. As a further technical scheme of the invention, the specific process of the step S22 is as follows:
taking one floor of the parking building as a basic unit, wherein the outdoor parking lot is treated as one floor and is shared by the parking buildingsLayer, time->The parking area structure of each layer can be modeled as a graph structure +.>,/>Wherein->Represents a set of parking areas, let->Layer commonness->Each region is defined as +.>Is a vertex of (1)Wherein->Represents the parking lot->All areas of the layer;>representation->Positional relationship between individual parking areas at the moment, wherein ∈ ->Is area->The weights between the two are initialized according to the following rules: if->And->Adjacent, then->On the contrary, the->;
By passing throughCalculating the relation weight between each parking area according to the parking quantity of each area at the moment, and setting +.>Time parking area->Is +.>Definitions->Difference vector from other region->In the formula +.>Represents->Number of layer areas>Using disparity vector->Derived from self-attention mechanismsWeight between other nodes->Update by the weight +.>Middle->Weights between regionsWherein->Is->Weights in between;
map after updatingInput into GCN to get->Time->Embedded features of layers->Calculating attention weights between different layers using a self-attention mechanism>Obtain->Parking lot structure embedded feature->。
As a further technical scheme of the invention, the specific process of the step S23 is as follows:
will each time instantFeature vector +.>Embedding features with corresponding parking structures>Integration to obtain input feature vector->;
The parking space occupation prediction model adopts a deep neural network based on a transducer as a main network, and inputs characteristic vectorsInputting a parking space occupation prediction model, and obtaining a feature map ++using 1×1 convolution on the features passing through the transform decoder>Wherein->Representing the total number of parking floors>And->Height and width of the feature map; for->And (3) flattening the characteristics of the channel, obtaining a final prediction interval through a linear layer, and finally iterating 100 times by using the MSE loss function as a loss function training model.
Compared with the prior art, the invention has the following advantages:
(1) Modeling the parking lot structure data, and capturing potential links between different areas through the GCN and an attention mechanism;
(2) The parking space occupation prediction is carried out through the improved TFT, the relation among different types of inputs is captured, the final prediction interval is output for a user to make a decision, the improved TFT can accept and distinguish various heterogeneous inputs, meanwhile, the inputs are comprehensively utilized, long-distance dependence can be captured, and a vehicle owner can judge the prediction confidence according to the prediction interval, so that whether the vehicle owner goes to a target parking lot or not is determined.
Drawings
FIG. 1 is a block diagram of a workflow for implementing parking space occupancy prediction in accordance with the present invention.
Fig. 2 is a flow chart of the parking lot structure embedding according to the present invention.
Detailed Description
The invention is further illustrated by the following examples in conjunction with the accompanying drawings.
Examples:
as shown in fig. 1 and 2, the parking space occupation prediction in this embodiment specifically includes the following steps:
s1, collecting parking data of a parking lot, wherein the parking data comprise a parking lot structure, historical parking data, current parking data, road vehicle conditions around the parking lot and auxiliary information; wherein the parking lot structure comprises a parking lot, an adjacent relation among parking buildings and an adjacent relation among parking areas; the historical parking data contains the number of parks in each area at a certain moment in the history; the current parking data is the parking quantity of each area in the current time period (such as 15 minutes); the historical parking data and the current parking data form parking data of a parking lot; the road vehicles around the parking lot are data of vehicles nearby the parking lot; the auxiliary information comprises date information and holiday information;
s2, training a parking space occupation prediction model according to the parking data collected in the step S1; the specific process is as follows:
step S21, pair S1 collectParking data at each moment (including +/each moment>Parking data of the parking lot, the number of vehicles nearby the parking lot, and date and holiday) to obtain feature vector +.>,/>For the dimension of the feature vector, +.>;
Step S22, embedding the parking lot structure data through the GCN to obtain each momentParking lot structure embedded feature->;
Taking one floor of the parking building as a basic unit, wherein the outdoor parking lot is treated as one floor and is shared by the parking buildingsLayer, time->The parking area structure of each layer can be modeled as a graph structure +.>,/>Wherein->Represents a set of parking areas, let->Layer commonness->Each region is defined as +.>Is a vertex of (1)Wherein->Represents the parking lot->All areas of the layer;>representation->Positional relationship between individual parking areas at the moment, wherein ∈ ->Is area->The weights between the two are initialized according to the following rules: if->And->Adjacent, then->On the contrary, the->. In the next step, the method will go through +.>Time parking data update ∈>;
By passing throughCalculating the relation weight between each parking area according to the parking quantity of each area at the moment, and setting +.>Time parking area->Is +.>Definitions->Difference vector from other region->In the formula +.>Represents->Number of layer areas>Using disparity vector->Derived from self-attention mechanismsWeight between other nodes->Update by the weight +.>Middle->Weights between regionsWherein->Is->Weight of the two, inInitializing according to a certain rule in the last step;
map after updatingInput into GCN to get->Time->Embedded features of layers->Calculating attention weights between different layers using a self-attention mechanism>Obtain->Parking lot structure embedded feature->;
S23, using a depth neural network based on a transducer as a backbone network to input feature vectors in a parking space occupation prediction modelInputting a parking space occupation prediction model, and obtaining a feature map ++using 1×1 convolution on the features passing through the transform decoder>Wherein->Representing the total number of parking floors>And->Height and width of the feature map; for->The characteristics of each channel (representing the characteristics of each floor) of the channel are flattened, a final prediction interval is obtained through a linear layer, and finally an MSE loss function (mean square loss function) is used as a loss function training model to iterate 100 times, so that a trained parking space occupation prediction model is obtained;
s3, inputting the current parking space data into a trained parking space occupation prediction model to obtain a parking space occupation prediction interval, and returning the result to the user through the corresponding APP.
Algorithms, network structures, and computational processes not described in detail herein are all general techniques in the art.
It should be noted that the purpose of the disclosed embodiments is to aid further understanding of the present invention, but those skilled in the art will appreciate that: various alternatives and modifications are possible without departing from the spirit and scope of the invention and the appended claims, and therefore the invention should not be limited to the embodiments disclosed, but rather the scope of the invention is defined by the appended claims.
Claims (1)
1. A parking space occupation prediction method for a parking lot is characterized by comprising the following specific steps:
s1, collecting parking data of a parking lot; the parking data comprise a parking lot structure, historical parking data, current parking data, road vehicle conditions around a parking lot and auxiliary information; wherein the parking lot structure comprises a parking lot, an adjacent relation among parking buildings and an adjacent relation among parking areas; the historical parking data contains the number of parks in each area at a certain moment in the history; the current parking data is the parking quantity of each area in the current time period; the historical parking data and the current parking data form parking data of a parking lot; the road vehicles around the parking lot are data of vehicles nearby the parking lot; the auxiliary information comprises date information and holiday information;
s2, training a parking space occupation prediction model according to the parking data collected in the step S1, wherein the parking space occupation prediction model specifically comprises the following steps:
step S21, pair S1 collectPreprocessing parking data of a parking lot, vehicle data nearby the parking lot, date and holiday data at each moment to obtain +.>Feature vector +.>,/>For the dimension of the feature vector, +.>;
Step S22, embedding the parking lot structure data through the GCN to obtain each momentParking lot structure embedded features of (a)The method specifically comprises the following steps:
taking one floor of the parking building as a basic unit, wherein the outdoor parking lot is treated as one floor and is shared by the parking buildingsLayer, time->The parking area structure of each layer can be modeled as a graph structure +.>,/>Wherein->Represents a set of parking areas, let->Layer commonness->Each region is defined as +.>Is->Wherein->Represents the parking lot->All areas of the layer;>representation->Positional relationship between individual parking areas at the moment, wherein ∈ ->Is area->The weights between the two are initialized according to the following rules: if->And->Adjacent, then->On the contrary, the->;
By passing throughCalculating the relation weight between each parking area according to the parking quantity of each area at the moment, and setting +.>Time parking area->Is +.>Definitions->Difference vector from other region->In the formula +.>Represents->Number of layer areas>Using disparity vector->Is associated with self-attention mechanism>Weight between other nodes->Update by the weight +.>Middle->Weight between regions->Wherein->Is->Weights in between;
map after updatingInput into GCN to get->Time->Embedded features of layers->Calculating attention weights between different layers using a self-attention mechanism>Obtain->Parking lot structure embedded feature->;
Step S23, each time is setParking lot structure embedded feature->Along with the feature vector->Training a parking space occupation prediction model as input data; the method comprises the following steps:
will each time instantFeature vector +.>Embedding features with corresponding parking structures>Integration to obtain input feature vector->;
The parking space occupation prediction model adopts a deep neural network based on a transducer as a main network, and inputs characteristic vectorsInputting a parking space occupation prediction model, and obtaining a feature map ++using 1×1 convolution on the features passing through the transform decoder>Wherein->Representing the total number of parking floors>And->Height and width of the feature map; for->After flattening the characteristics of the channel, obtaining a final prediction interval through a linear layer, and finally iterating 100 times by using an MSE loss function as a loss function training model;
s3, inputting the current parking space data into a trained parking space occupation prediction model to obtain a parking space occupation prediction interval, and returning the result to the user through the corresponding APP.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310490841.4A CN116187602B (en) | 2023-05-05 | 2023-05-05 | Parking space occupation prediction method for parking lot |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310490841.4A CN116187602B (en) | 2023-05-05 | 2023-05-05 | Parking space occupation prediction method for parking lot |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116187602A CN116187602A (en) | 2023-05-30 |
CN116187602B true CN116187602B (en) | 2023-07-14 |
Family
ID=86436872
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310490841.4A Active CN116187602B (en) | 2023-05-05 | 2023-05-05 | Parking space occupation prediction method for parking lot |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116187602B (en) |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115223132B (en) * | 2021-11-10 | 2023-10-27 | 广州汽车集团股份有限公司 | Empty space recognition method and system and computer readable storage medium |
CN114802296A (en) * | 2022-03-17 | 2022-07-29 | 浙江工业大学 | Vehicle track prediction method based on dynamic interaction graph convolution |
CN115223019B (en) * | 2022-07-14 | 2023-05-19 | 清华大学 | All-day parking space detection method based on fusion of camera and laser radar |
CN115643285A (en) * | 2022-10-14 | 2023-01-24 | 成都秦川物联网科技股份有限公司 | Smart city parking lot recommendation method, internet of things system, device and storage medium |
CN115985132A (en) * | 2022-12-29 | 2023-04-18 | 浙大城市学院 | Roadside parking availability prediction method based on graph neural network |
-
2023
- 2023-05-05 CN CN202310490841.4A patent/CN116187602B/en active Active
Non-Patent Citations (1)
Title |
---|
基于卷积神经网络的环视车位检测研究;杨科;张效宇;徐维庆;;时代汽车(04);第1-7页 * |
Also Published As
Publication number | Publication date |
---|---|
CN116187602A (en) | 2023-05-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110991713B (en) | Irregular area flow prediction method based on multi-graph convolution sum GRU | |
CN113724504B (en) | Urban area traffic prediction system and method oriented to vehicle track big data | |
CN110047291B (en) | Short-term traffic flow prediction method considering diffusion process | |
CN111311122B (en) | Charging station construction method for smart city | |
US20240054321A1 (en) | Traffic prediction | |
CN111242292B (en) | OD data prediction method and system based on deep space-time network | |
CN106017496A (en) | Real-time navigation method based on road condition | |
CN112863182B (en) | Cross-modal data prediction method based on transfer learning | |
CN111242395B (en) | Method and device for constructing prediction model for OD (origin-destination) data | |
JP2021182408A (en) | Method for predicting water level in sewage pipe sewer | |
US20230057564A1 (en) | Methods for managing cleaning routes in smart cities, internet of things systems, and storage mediums | |
CN109543895A (en) | A kind of transit network planning method out based on taxi passenger flow conversion | |
CN116913088A (en) | Intelligent flow prediction method for expressway | |
CN116187602B (en) | Parking space occupation prediction method for parking lot | |
WO2023276420A1 (en) | Distributed multi-task machine learning for traffic prediction | |
CN113837148A (en) | Pedestrian trajectory prediction method based on self-adjusting sparse graph transform | |
CN116089744B (en) | Hospital parking lot recommendation method based on transform dynamic time-space association | |
CN113327417A (en) | Traffic flow prediction method based on 3D dynamic space-time residual convolution associated network | |
CN115762147B (en) | Traffic flow prediction method based on self-adaptive graph meaning neural network | |
Ara et al. | Ride Hailing Service Demand Forecast by Integrating Convolutional and Recurrent Neural Networks. | |
CN115660157A (en) | Smart city charging pile construction planning method, internet of things system, device and medium | |
CN114139773A (en) | Public traffic flow prediction method based on space-time graph convolutional network | |
CN117252404B (en) | Method and system for measuring and calculating population and post scale of city updating unit | |
CN111242394B (en) | Method and system for extracting spatial correlation characteristics | |
Ye et al. | Urban Traffic Dynamic OD Prediction Based on Multi-Source Data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |