CN109214422B - Parking data repairing method, device, equipment and storage medium based on DCGAN - Google Patents
Parking data repairing method, device, equipment and storage medium based on DCGAN Download PDFInfo
- Publication number
- CN109214422B CN109214422B CN201810869067.7A CN201810869067A CN109214422B CN 109214422 B CN109214422 B CN 109214422B CN 201810869067 A CN201810869067 A CN 201810869067A CN 109214422 B CN109214422 B CN 109214422B
- Authority
- CN
- China
- Prior art keywords
- parking lot
- parking
- data
- empty rate
- sample
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention is suitable for the technical field of data repair, and provides a parking data repair method, a parking data repair device, parking data repair equipment and a parking data repair storage medium based on DCGAN, wherein the parking data repair method comprises the following steps: the method comprises the steps of clustering parking lots in an urban parking lot set into parking lot clusters, wherein the parking lots comprise sample parking lots and parking lots to be repaired, generating corresponding empty rate curve images according to parking data of the sample parking lots, training a generation type countermeasure network based on deep convolution according to the empty rate curve images of the sample parking lots in the parking lot clusters to generate empty rate curve simulation images of each parking lot cluster, processing and mapping the empty rate curve simulation images into corresponding one-dimensional empty rate data, and performing data repair on the parking lots to be repaired in each parking lot cluster according to the one-dimensional empty rate data, so that accurate data repair is performed on the parking lots to be repaired when prior knowledge of the parking lots to be repaired is not relied on, urban-level parking data repair is achieved, and data acquisition cost of the parking lots is saved.
Description
Technical Field
The invention belongs to the technical field of data repair, and particularly relates to a parking data repair method, device, equipment and storage medium based on DCGAN.
Background
Transportation is an important part of modern economy, but along with the development of social economy, the contradiction between the increasing traffic demand and the urban traffic load is increasingly prominent. The parking difficulty problem is prominent, and the intelligent parking system is one of the most effective methods for solving the problem and reducing the parking time cost.
Any machine learning algorithm needs a large amount of sample data as a support, and the development of an intelligent parking system also needs a large amount of parking data of parking lots in a certain area. Conventional methods for obtaining parking data include searching for public data sets, or collecting by purchasing and installing sensors, etc., and these conventional methods have high economic cost and time cost, and it is difficult to obtain parking data of all parking lots in a city, resulting in a large portion of the parking lots in the city lacking parking data. In addition, construction damage, line faults, processing errors and the like also cause partial parking lots to have incomplete parking data. Therefore, in the case of large-range parking data loss, how to improve the integrity, effectiveness and predictability of the parking data in the city range on the premise that the economic cost is controllable provides effective support for related machine learning algorithms and traffic analysis models is a problem to be considered in the design and training of the current intelligent parking algorithm.
The above problem can be expressed as model distortion caused by data missing, and data enhancement techniques are generally adopted in machine learning to deal with the data missing problem, i.e. data generation (or repair) techniques in a popular sense. The latest application result of the technology is mainly reflected in the style migration and the repair of the image, and the research on the traffic data repair is relatively deficient. At present, the research in the field of traffic data repair is mainly performed by interpolation methods, such as a data repair method based on a lagrange interpolation method, a data repair method based on a newton interpolation method, and a data repair method based on a piecewise linear interpolation method. However, historical parking data for most parking lots is difficult to obtain due to economic cost, installation construction, and economic property.
Disclosure of Invention
The invention aims to provide a DCGAN-based parking lot data repairing method, a DCGAN-based parking lot data repairing device, DCGAN-based parking lot data repairing equipment and a storage medium, and aims to solve the problems that in the prior art, the cost for collecting parking lot parking data in an urban range is high, and the dependency of the existing data repairing method on the historical parking lot parking data is high.
In one aspect, the present invention provides a DCGAN-based parking lot data repairing method, including the following steps:
acquiring a pre-collected urban parking lot set and an urban geographic interest point set, wherein parking lots in the urban parking lot set comprise a sample parking lot and a parking lot to be repaired;
clustering the parking lots into a plurality of parking lot clusters according to the urban geographic interest point set, and generating an empty rate curve image of the sample parking lot according to the parking data of the sample parking lot;
constructing a generating countermeasure network based on deep convolution, and training the generating countermeasure network based on deep convolution according to the empty rate curve image of the sample parking lot in the parking lot cluster to generate an empty rate curve simulation image corresponding to each parking lot cluster;
and processing the empty rate curve simulation image corresponding to each parking lot cluster, mapping the processed image into corresponding one-dimensional empty rate data, and performing parking data repair on the parking lot to be repaired in each parking lot cluster according to the one-dimensional empty rate data corresponding to each parking lot cluster.
In another aspect, the present invention provides a DCGAN-based parking lot data patching device, including:
the system comprises a collected data acquisition unit, a data processing unit and a data processing unit, wherein the collected data acquisition unit is used for acquiring a pre-collected urban parking lot set and an urban geographic interest point set, and parking lots in the urban parking lot set comprise sample parking lots and parking lots to be repaired;
the data processing unit is used for clustering the parking lots into a plurality of parking lot clusters according to the urban geographic interest point set and generating an empty rate curve image of the sample parking lot according to the parking data of the sample parking lot;
the network construction training unit is used for constructing a generating countermeasure network based on deep convolution, and training the generating countermeasure network based on deep convolution according to the empty rate curve image of the sample parking lot in the parking lot cluster so as to generate an empty rate curve simulation image corresponding to each parking lot cluster; and
and the parking data repairing unit is used for processing the empty rate curve simulation image corresponding to each parking lot cluster, mapping the empty rate curve simulation image into corresponding one-dimensional empty rate data, and repairing the parking data of the parking lot to be repaired in each parking lot cluster according to the one-dimensional empty rate data corresponding to each parking lot cluster.
In another aspect, the present invention further provides a computing device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps of the DCGAN-based parking lot data patch method.
In another aspect, the present invention further provides a computer-readable storage medium, which stores a computer program, and the computer program, when executed by a processor, implements the steps of the DCGAN-based parking lot data patch method.
The method comprises the steps of collecting an urban parking lot set and an urban geographic interest point set in advance, clustering parking lots in the urban parking lot set into parking lot clusters according to the urban geographic interest point set, generating empty rate curve images of the sample parking lots according to parking data of the sample parking lots, training a generating countermeasure network based on deep convolution according to the empty rate curve images of the sample parking lots in the parking lot clusters, generating empty rate curve simulation images corresponding to each parking lot cluster through the trained generating countermeasure network based on the deep convolution, processing the empty rate curve simulation images and mapping the empty rate curve simulation images into corresponding one-dimensional empty rate data, and conducting parking data repairing on the parking lots to be repaired in each parking lot cluster according to the one-dimensional empty rate data, so that accurate data repairing can be conducted on the parking lots to be repaired when the prior knowledge of the parking lots to be repaired is not relied on And (4) repairing, so that city-level parking data repairing is realized, and the data acquisition cost of the parking lot is further saved.
Drawings
Fig. 1 is a flowchart illustrating an implementation of a DCGAN-based parking lot data patching method according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a DCGAN-based parking lot data patching device according to a second embodiment of the present invention;
fig. 3 is a schematic structural diagram of a preferred parking lot data patching device based on DCGAN according to a second embodiment of the present invention; and
fig. 4 is a schematic structural diagram of a computing device according to a third embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The following detailed description of specific implementations of the present invention is provided in conjunction with specific embodiments:
the first embodiment is as follows:
fig. 1 shows an implementation flow of a DCGAN-based parking lot data patching method according to an embodiment of the present invention, and for convenience of description, only the parts related to the embodiment of the present invention are shown, which are detailed as follows:
in step S101, a pre-collected urban parking lot set and an urban geographic interest point set are obtained, where parking lots in the urban parking lot set include a sample parking lot and a parking lot to be repaired.
The embodiment of the invention is suitable for a data patching platform, a system and equipment. The pre-collected urban parking lot set comprises a plurality of parking lots in a city and relevant information of each parking lot, parking lots lack of parking data exist in the parking lots, parking lots not lack of parking data exist in the parking lots, and parking data repair needs to be carried out on the parking lots lack of parking data in order to support research and development of the intelligent parking system. The information on the parking lot to be repaired may include a geographical location of the parking lot to be repaired in the city and a scale of the parking lot to be repaired, and the information on the sample parking lot may include a geographical location of the sample parking lot in the city, a scale of the sample parking lot, and parking data of the sample parking lot.
In the embodiment of the present invention, the pre-collected city geographic Interest Point set includes geographic locations of geographic Interest points (POI for short, such as parking lots, shopping malls, stations, and schools) in a city, and these information may be obtained on a map website through a crawler program, where the geographic locations may be longitude and latitude information.
In step S102, the parking lots are clustered into a plurality of parking lot clusters according to the city geographical interest point set, and an empty rate curve image of the sample parking lot is generated according to the parking data of the sample parking lot.
In the embodiment of the invention, people often find a parking lot closest to a destination in life, if the closest parking lot has no parking space, a parking lot a little far away is considered, but the distance between the finally selected parking lot and the destination is not infinite, namely, a maximum distance acceptable to people exists, and the maximum distance is called tolerance radius. Therefore, a circle is formed by taking the parking lot as the center of the circle and taking the tolerance radius as the radius of the circle, all the POIs in the circle have influence on the parking data of the parking lot, and the influence of the POIs outside the circle on the parking data of the parking lot is negligible, so that two parking lots with similar number of the surrounding POIs are probably similar.
Preferably, when the parking lots are clustered into a plurality of parking lot clusters according to the urban geographical interest point set, a tolerance radius value is preset, a circle is formed by taking the parking lots as a center of the circle and taking the tolerance radius as a circle radius, POIs of which the geographical positions appear in the circle are obtained from the urban geographical interest point set, the POIs are POIs around the parking lots, a spatial feature corresponding to each parking lot is constructed according to the POIs around each parking lot, if n POIs are mainly found in the current city (or n POIs are collected from the urban geographical interest point set), the spatial feature corresponding to each parking lot is an n-dimensional feature vector, all the parking lots in the urban parking lot set are clustered according to the spatial feature corresponding to each parking lot and a preset Density-Based spatial clustering algorithm (Density-Based clustering of application with Noise, referred to as can dbs), a plurality of clusters are obtained, thereby effectively improving the classification effect of the parking lot. For convenience of description, the clustered clusters are referred to herein as parking lot clusters.
Further preferably, when all parking lots in the urban parking lot set are clustered according to the spatial features corresponding to each parking lot and the DBSCAN, the core point threshold value MinPts in the DBSCAN is determined according to the data scale of the data set formed by the spatial features of all parking lots, and the field radius epsilon of the density in the DBSCAN is determined according to the mean value of the euclidean distances between the parking lots randomly selected from the urban parking lot set, so that the classification effect of the parking lots is effectively improved.
As an example, in a specific experimental process, 10 parking lots are randomly extracted from 310 parking lots, the mean value of euclidean distances of the parking lots in the dimensional space (the number of types of POI) is calculated 7 to be 22, 22 is set as the domain radius epsilon of the density in DBSCAN, 49 is taken as the core point threshold value MinPts in DBSCAN, and the 310 parking lots are clustered into 6 classes, that is, 6 parking lot clusters by the DBSCAN.
In the embodiment of the invention, when the empty rate curve image of the sample parking lot is generated according to the parking data of the sample parking lot, the parking data of the sample parking lot is accurate time information of vehicles entering and exiting the sample parking lot, the number of the idle parking spaces of the sample parking lot at different moments can be obtained according to the accurate time information of the vehicles entering and exiting the sample parking lot, and the number of the idle parking spaces of the sample parking lot at different moments can be normalized by considering that the scales of the different parking lots are possibly different (that is, the total number of the parking spaces is possibly different), so that the idle parking space rate of the sample parking lot at different moments is obtained.
In the embodiment of the invention, two-dimensional coordinates corresponding to the empty rate of the sample parking lot at different moments are calculated according to preset curve parameters and the number of sampling points, the two-dimensional coordinates form an empty rate curve corresponding to the sample parking lot, and then an empty rate curve image of the sample parking lot is obtained.
Preferably, the sample parking lot time tpThe empty rate of
Sample parking lot time tpThe empty rate of (a) corresponds to a two-dimensional coordinate of
Wherein the time sequence of the sample parking lot is expressed as
For sample parking lot at time tpThe Total number of the idle parking spaces is the Total number of the parking spaces in the sample parking lot, L is the length of a curve in a curve parameter, H is the height of the curve in the curve parameter, and J is the number of sampling points.
Preferably, since POIs of different parking lots in the same parking lot cluster are certainly similar after clustering, but parking data of different parking lots in the same parking lot cluster are not necessarily similar after clustering, the similarity degree of the empty rate curves of different sample parking lots in the same parking lot cluster is calculated through the pearson correlation coefficient after clustering is finished, so as to judge whether the empty rate curves of different parking lots in the same parking lot cluster are similar. For example, when the pearson correlation coefficient of the empty rate curves of the two sample parking lots is greater than 0.5, the parking data of the two sample parking lots are considered to have strong correlation, and if the number of the parking lot alignment combinations having strong correlation in one parking lot cluster exceeds 70% of the number of all the parking lot alignment combinations in the parking lot cluster, the parking data of different parking lots in the parking lot cluster are considered to be similar.
In step S103, a generative confrontation network based on deep convolution is constructed, and the generative confrontation network based on deep convolution is trained according to the empty rate curve images of the sample parking lots in the parking lot clusters to generate an empty rate curve simulation image corresponding to each parking lot cluster.
In the embodiment of the present invention, a preset Generative Adaptive Network (GAN) is extended according to a preset supervised Convolutional Neural Network (CNN) to obtain a Deep Convolutional based Generative adaptive network (DCGAN), where the number of network layers of the supervised Convolutional neural network and the Generative adaptive network, and the number of units in each layer are not limited. Compared with GAN before expansion, the DCGAN constructed has the following main changes in model structure: the pooling layer is replaced with a convolutional layer, wherein the pooling layer in the discriminator is replaced with a convolutional layer with step size, and the pooling layer in the generator is replaced with a convolutional layer with micro-step size; batch normalization is employed on both the discriminator and the generator; and removing the full connection layer. Preferably, the output layer of the generator in the DCGAN is connected with a filter for the normality check, and only the empty rate curve simulation image conforming to the normality check is output, so as to ensure the effect of the empty rate curve simulation image generated by the generator in the DCGAN. Further preferably, the filter for the normality check adopts a agonst algorithm (D' agonstino-Pearsontest) to improve the effect of the normality check on the empty rate curve simulation image.
In the embodiment of the invention, the empty rate curve images of the sample parking lots in the parking lot cluster are input into the DCGAN as training data, the DCGAN is trained, and the trained DCGAN generates the empty rate curve simulation images corresponding to the parking lot cluster (namely, the empty rate curve simulation images corresponding to the parking lots to be repaired in the parking lot cluster, and for convenience of distinguishing, the empty rate curve simulation images generated by the DCGAN are referred to as the empty rate curve simulation images), so that the empty rate curve simulation images corresponding to each parking lot cluster can be generated.
In step S104, the empty rate curve simulation image corresponding to each parking lot cluster is processed and mapped to corresponding one-dimensional empty rate data, and parking data repair is performed on the parking lot to be repaired in each parking lot cluster according to the one-dimensional empty rate data corresponding to each parking lot cluster.
In the embodiment of the invention, after the empty rate curve simulation image corresponding to each parking lot cluster is obtained, the image processing operations such as graying, binarization, noise reduction and the like are carried out on the empty rate curve image, and the processed empty rate curve image is mapped into corresponding one-dimensional empty rate data (time sequence of empty rates). According to the one-dimensional empty rate data corresponding to the parking lot cluster and the scale of the parking lot to be repaired in the parking lot cluster, the number of the idle parking spaces of the parking lot to be repaired at different moments in the parking lot cluster can be calculated, and then the parking data repairing of each parking lot to be repaired in each parking lot cluster is achieved. When the processed empty rate curve image is mapped into corresponding one-dimensional empty rate data, the inverse relationship of the mapping relationship from the empty rate of the sample parking lot at different moments to the empty rate curve image can be adopted for realization.
Preferably, the formula for graying the simulation image of the empty rate curve is as follows:
R2＝G2＝B2＝R1*a1+G1*a2+B1*a3wherein R is2、G2And B2Simulating RGB values, R, after graying of an image for an empty rate curve1、G1And B1Simulating the RGB value before graying of the image for the empty rate curve, a1、a2And a3Is a preset graying parameter.
Preferably, when the grayed empty rate curve simulation image is binarized, a threshold is preset, each pixel point of the empty rate curve simulation image is traversed, the pixel point with the pixel value exceeding the threshold is set to be white, otherwise, the pixel point is set to be black, and therefore the processing effect of the empty rate curve simulation image is improved.
Preferably, the noise reduction processing on the empty rate curve simulation image comprises burr point removal and outlier removal, and considering that the burr points of the curve in the empty rate curve simulation image are caused by too dense pixel points and the density of the pixel points needs to be reduced, the average filtering is used for removing the burr points so as to improve the processing effect of the burr points on the empty rate curve simulation image. Further preferably, the formula for removing the burr points using the mean filtering is:
Preferably, when the outliers are removed from the empty rate curve simulation image, the outliers on the empty rate curve simulation image are detected by adopting a distance-based anomaly detection algorithm, so that the outlier detection effect is improved. Further preferably, the distance-based anomaly detection algorithm is a Local anomaly Factor LOF algorithm (Local Outlier Factor) to improve the Outlier detection effect.
In the embodiment of the invention, the parking lots are clustered into the parking lot clusters, the empty rate curve image of the sample parking lot is generated according to the parking data of the sample parking lot, training a generating countermeasure network based on deep convolution according to the empty rate curve image of the sample parking lot in the parking lot cluster, generating an empty rate curve simulation image corresponding to each parking lot cluster by a trained generation countermeasure network based on deep convolution, processing the empty rate curve simulation images and mapping the empty rate curve simulation images into corresponding one-dimensional empty rate data, so as to carry out parking data repair on the parking lots to be repaired in each parking lot cluster according to the one-dimensional space rate data, therefore, accurate data repair is carried out on the parking lot to be repaired when the priori knowledge of the parking lot to be repaired is not relied on, city-level parking data repair is achieved, and the data acquisition cost of the parking lot is saved.
Example two:
fig. 2 shows a structure of a DCGAN-based parking lot data patch apparatus according to a third embodiment of the present invention, and for convenience of description, only the parts related to the third embodiment of the present invention are shown, where the parts include:
the system comprises a collected data acquisition unit 21, which is used for acquiring a pre-collected urban parking lot set and an urban geographic interest point set, wherein parking lots in the urban parking lot set comprise sample parking lots and parking lots to be repaired.
In the embodiment of the invention, the pre-collected urban parking lot set comprises a plurality of parking lots in a city and relevant information of each parking lot, and the parking lots comprise parking lots to be repaired and sample parking lots. The information on the parking lot to be repaired may include a geographical location of the parking lot to be repaired in the city and a scale of the parking lot to be repaired, and the information on the sample parking lot may include a geographical location of the sample parking lot in the city, a scale of the sample parking lot, and parking data of the sample parking lot.
In the embodiment of the invention, the pre-collected city geographic interest point set comprises geographic positions of geographic interest Points (POIs) in a city, the information can be acquired on a map website through a crawler program, and the geographic positions can be longitude and latitude information.
And the data processing unit 22 is configured to cluster the parking lots into a plurality of parking lot clusters according to the city geographic interest point set, and generate an empty rate curve image of the sample parking lot according to the parking data of the sample parking lot.
In the embodiment of the invention, people often find a parking lot closest to a destination in life, if the closest parking lot has no parking space, a parking lot a little far away is considered, but the distance between the finally selected parking lot and the destination is not infinite, namely, a maximum distance acceptable to people exists, and the maximum distance is called tolerance radius. Therefore, a circle is formed by taking the parking lot as the center of the circle and taking the tolerance radius as the radius of the circle, all the POIs in the circle have influence on the parking data of the parking lot, and the influence of the POIs outside the circle on the parking data of the parking lot is negligible, so that two parking lots with similar number of the surrounding POIs are probably similar.
Preferably, when the parking lots are clustered into a plurality of parking lot clusters according to the urban geographic interest point set, a tolerance radius value is preset, the parking lots are used as the center of a circle, the tolerance radius is used as the radius of the circle to form a circle, POIs with geographic positions appearing in the circle are obtained from the urban geographic interest point set, the POIs are POIs around the parking lots, the spatial features corresponding to each parking lot are constructed according to the POIs around each parking lot, if n types of main POIs exist in the current city, the spatial features corresponding to each parking lot are n-dimensional feature vectors, all the parking lots in the urban parking lot set are clustered according to the spatial features corresponding to each parking lot and a preset density-based spatial clustering algorithm (DBSCAN), and a plurality of clusters are obtained, so that the classification effect of the parking lots is effectively improved. For convenience of description, the clustered clusters are referred to herein as parking lot clusters.
Further preferably, when all parking lots in the urban parking lot set are clustered according to the spatial features corresponding to each parking lot and the DBSCAN, the core point threshold value MinPts in the DBSCAN is determined according to the data scale of the data set formed by the spatial features of all parking lots, and the field radius epsilon of the density in the DBSCAN is determined according to the mean value of the euclidean distances between the parking lots randomly selected from the urban parking lot set, so that the classification effect of the parking lots is effectively improved.
In the embodiment of the invention, when the empty rate curve image of the sample parking lot is generated according to the parking data of the sample parking lot, the parking data of the sample parking lot is accurate time information of vehicles entering and exiting the sample parking lot, the number of the idle parking places of the sample parking lot at different moments can be obtained according to the accurate time information of the vehicles entering and exiting the sample parking lot, and the number of the idle parking places of the sample parking lot at different moments can be normalized to obtain the empty rate of the sample parking lot at different moments by considering that the scales of different parking lots are possibly different.
In the embodiment of the invention, two-dimensional coordinates corresponding to the empty rate of the sample parking lot at different moments are calculated according to preset curve parameters and the number of sampling points, the two-dimensional coordinates form an empty rate curve corresponding to the sample parking lot, and then an empty rate curve image of the sample parking lot is obtained.
Preferably, the sample parking lot time tpThe empty rate of
Sample parking lot time tpThe empty rate of (a) corresponds to a two-dimensional coordinate of
Wherein the time sequence of the sample parking lot is expressed as
For sample parking lot at time tpThe Total number of the idle parking spaces is the Total number of the parking spaces in the sample parking lot, L is the length of a curve in a curve parameter, H is the height of the curve in the curve parameter, and J is the number of sampling points.
Preferably, since POIs of different parking lots in the same parking lot cluster are certainly similar after clustering, but parking data of different parking lots in the same parking lot cluster are not necessarily similar after clustering, the similarity degree of the empty rate curves of different sample parking lots in the same parking lot cluster is calculated through the pearson correlation coefficient after clustering is finished, so as to judge whether the empty rate curves of different parking lots in the same parking lot cluster are similar.
And the network construction training unit 23 is configured to construct a deep convolution-based generative confrontation network, and train the deep convolution-based generative confrontation network according to the empty rate curve image of the sample parking lot in the parking lot cluster to generate an empty rate curve simulation image corresponding to each parking lot cluster.
In the embodiment of the invention, the preset generative countermeasure network (GAN) is extended according to the preset supervised Convolutional Neural Network (CNN) to obtain the deep convolution based generative countermeasure network (DCGAN), and the network layer number of the supervised convolutional neural network and the generative countermeasure network and the unit number of each layer are not limited. Compared with GAN before expansion, the DCGAN constructed has the following main changes in model structure: the pooling layer is replaced with a convolutional layer, wherein the pooling layer in the discriminator is replaced with a convolutional layer with step size, and the pooling layer in the generator is replaced with a convolutional layer with micro-step size; batch normalization is employed on both the discriminator and the generator; and removing the full connection layer. Preferably, the output layer of the generator in the DCGAN is connected with a filter for the normality check, and only the empty rate curve simulation image conforming to the normality check is output, so as to ensure the effect of the empty rate curve simulation image generated by the generator in the DCGAN. Further preferably, the filter for the normality check adopts a degaussing-Pearson test (D' agonstino-Pearson test) to improve the effect of the normality check on the empty rate curve simulation image.
In the embodiment of the invention, the empty rate curve images of the sample parking lots in the parking lot clusters are input into the DCGAN as training data to train the DCGAN, and the trained DCGAN generates the empty rate curve simulation images corresponding to the parking lot clusters, so that the empty rate curve simulation images corresponding to each parking lot cluster can be generated.
And the parking data repairing unit 24 is configured to process the empty rate curve simulation image corresponding to each parking lot cluster, map the empty rate curve simulation image into corresponding one-dimensional empty rate data, and repair parking data of the parking lot to be repaired in each parking lot cluster according to the one-dimensional empty rate data corresponding to each parking lot cluster.
In the embodiment of the invention, after the empty rate curve simulation image corresponding to each parking lot cluster is obtained, the image processing operations such as graying, binarization, noise reduction and the like are carried out on the empty rate curve image, and the processed empty rate curve image is mapped into corresponding one-dimensional empty rate data. According to the one-dimensional empty rate data corresponding to the parking lot cluster and the scale of the parking lot to be repaired in the parking lot cluster, the number of the idle parking spaces of the parking lot to be repaired at different moments in the parking lot cluster can be calculated, and then the parking data repairing of each parking lot to be repaired in each parking lot cluster is achieved. When the processed empty rate curve image is mapped into corresponding one-dimensional empty rate data, the inverse relationship of the mapping relationship from the empty rate of the sample parking lot at different moments to the empty rate curve image can be adopted for realization.
Preferably, the formula for graying the simulation image of the empty rate curve is as follows:
R2＝G2＝B2＝R1*a1+G1*a2+B1*a3wherein R is2、G2And B2Simulating RGB values, R, after graying of an image for an empty rate curve1、G1And B1Simulating pre-graying of images for empty rate curvesRGB value, a1、a2And a3Is a preset graying parameter.
Preferably, when the grayed empty rate curve simulation image is binarized, a threshold is preset, each pixel point of the empty rate curve simulation image is traversed, the pixel point with the pixel value exceeding the threshold is set to be white, otherwise, the pixel point is set to be black, and therefore the processing effect of the empty rate curve simulation image is improved.
Preferably, the noise reduction processing on the empty rate curve simulation image comprises burr point removal and outlier removal, and considering that the burr points of the curve in the empty rate curve simulation image are caused by too dense pixel points and the density of the pixel points needs to be reduced, the average filtering is used for removing the burr points so as to improve the processing effect of the burr points on the empty rate curve simulation image. Further preferably, the formula for removing the burr points using the mean filtering is:
Preferably, when the outliers are removed from the empty rate curve simulation image, the outliers on the empty rate curve simulation image are detected by adopting a distance-based anomaly detection algorithm, so that the outlier detection effect is improved. Further preferably, the distance-based anomaly detection algorithm is a local anomaly factor LOF algorithm to improve outlier detection effects.
Preferably, the data processing unit 22 comprises:
the geographic interest point determining unit 321 is configured to determine geographic interest points around each parking lot in the urban parking lot set in the urban geographic interest point set according to a preset tolerance radius; and
and the parking lot clustering unit 322 is configured to construct a spatial feature corresponding to each parking lot according to the geographic interest points around each parking lot, and perform high-dimensional clustering on the parking lots according to the spatial feature corresponding to each parking lot and a preset density-based spatial clustering algorithm.
Preferably, the network construction training unit 23 includes:
the network construction unit 331 is configured to extend the preset generative countermeasure network according to the preset supervised convolutional neural network to construct a generative countermeasure network based on deep convolution, where an output layer in the generative countermeasure network based on deep convolution is connected to a filter for normality check.
In the embodiment of the invention, the parking lots are clustered into the parking lot clusters, the empty rate curve image of the sample parking lot is generated according to the parking data of the sample parking lot, training a generating countermeasure network based on deep convolution according to the empty rate curve image of the sample parking lot in the parking lot cluster, generating an empty rate curve simulation image corresponding to each parking lot cluster by a trained generation countermeasure network based on deep convolution, processing the empty rate curve simulation images and mapping the empty rate curve simulation images into corresponding one-dimensional empty rate data, so as to carry out parking data repair on the parking lots to be repaired in each parking lot cluster according to the one-dimensional space rate data, therefore, accurate data repair is carried out on the parking lot to be repaired when the priori knowledge of the parking lot to be repaired is not relied on, city-level parking data repair is achieved, and the data acquisition cost of the parking lot is saved.
In the embodiment of the present invention, each unit of the DCGAN-based parking lot data patch apparatus may be implemented by a corresponding hardware or software unit, and each unit may be an independent software or hardware unit, or may be integrated into a software or hardware unit, which is not limited herein.
Example three:
fig. 4 shows a structure of a computing device provided in a third embodiment of the present invention, and for convenience of explanation, only a part related to the third embodiment of the present invention is shown.
In the embodiment of the invention, the parking lots are clustered into the parking lot clusters, the empty rate curve image of the sample parking lot is generated according to the parking data of the sample parking lot, training a generating countermeasure network based on deep convolution according to the empty rate curve image of the sample parking lot in the parking lot cluster, generating an empty rate curve simulation image corresponding to each parking lot cluster by a trained generation countermeasure network based on deep convolution, processing the empty rate curve simulation images and mapping the empty rate curve simulation images into corresponding one-dimensional empty rate data, so as to carry out parking data repair on the parking lots to be repaired in each parking lot cluster according to the one-dimensional space rate data, therefore, accurate data repair is carried out on the parking lot to be repaired when the priori knowledge of the parking lot to be repaired is not relied on, city-level parking data repair is achieved, and the data acquisition cost of the parking lot is saved.
Example four:
in an embodiment of the present invention, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps in the above-described method embodiments, e.g., steps S101 to S104 shown in fig. 1. Alternatively, the computer program realizes the functions of the units in the above-described apparatus embodiments, such as the functions of the units 21 to 24 shown in fig. 2, when executed by the processor.
In the embodiment of the invention, the parking lots are clustered into the parking lot clusters, the empty rate curve image of the sample parking lot is generated according to the parking data of the sample parking lot, training a generating countermeasure network based on deep convolution according to the empty rate curve image of the sample parking lot in the parking lot cluster, generating an empty rate curve simulation image corresponding to each parking lot cluster by a trained generation countermeasure network based on deep convolution, processing the empty rate curve simulation images and mapping the empty rate curve simulation images into corresponding one-dimensional empty rate data, so as to carry out parking data repair on the parking lots to be repaired in each parking lot cluster according to the one-dimensional space rate data, therefore, accurate data repair is carried out on the parking lot to be repaired when the priori knowledge of the parking lot to be repaired is not relied on, city-level parking data repair is achieved, and the data acquisition cost of the parking lot is saved.
The computer readable storage medium of the embodiments of the present invention may include any entity or device capable of carrying computer program code, a recording medium, such as a ROM/RAM, a magnetic disk, an optical disk, a flash memory, or the like.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. A parking data repairing method based on DCGAN is characterized by comprising the following steps:
acquiring a pre-collected urban parking lot set and an urban geographic interest point set, wherein parking lots in the urban parking lot set comprise a sample parking lot and a parking lot to be repaired; wherein the parking to be repaired is a parking lot lacking parking data; the sample parking lot is a parking lot without lack of parking data;
clustering the parking lots into a plurality of parking lot clusters according to the urban geographic interest point set, and generating an empty rate curve image of the sample parking lot according to the parking data of the sample parking lot;
constructing a generating countermeasure network based on deep convolution, and training the generating countermeasure network based on deep convolution according to the empty rate curve image of the sample parking lot in the parking lot cluster to generate an empty rate curve simulation image corresponding to each parking lot cluster;
and processing the empty rate curve simulation image corresponding to each parking lot cluster, mapping the processed image into corresponding one-dimensional empty rate data, and performing parking data repair on the parking lot to be repaired in each parking lot cluster according to the one-dimensional empty rate data corresponding to each parking lot cluster.
2. The method of claim 1, wherein the step of clustering the parking lots into a plurality of parking lot clusters based on the set of urban geographic points of interest comprises:
determining geographic interest points around each parking lot in the urban parking lot set in the urban geographic interest point set according to a preset tolerance radius;
and constructing a spatial feature corresponding to each parking lot according to the geographical interest points around each parking lot, and performing high-dimensional clustering on the parking lots according to the spatial feature corresponding to each parking lot and a preset density-based spatial clustering algorithm.
3. The method of claim 1, wherein the step of generating the empty rate curve image of the sample parking lot from the parking data of the sample parking lot comprises:
calculating the empty rate of the sample parking lot at different moments according to the parking data of the sample parking lot;
and calculating two-dimensional coordinates corresponding to the empty rate of the sample parking lot at different moments according to preset curve parameters and the number of sampling points.
4. The method of claim 1, wherein the step of constructing a generative countermeasure network based on deep convolution comprises:
and expanding the preset generative countermeasure network according to a preset supervised convolutional neural network to construct the generative countermeasure network based on the deep convolution, wherein the output layer of a generator in the generative countermeasure network based on the deep convolution is connected with a filter for normality check.
5. The method of claim 1, wherein the step of processing and mapping the empty rate curve simulation image corresponding to each parking lot cluster into corresponding one-dimensional empty rate data comprises:
carrying out graying, binarization and noise reduction on the empty rate curve simulation image,
the noise reduction treatment comprises burr point removal and outlier removal;
and mapping the processed empty rate curve simulation image into corresponding one-dimensional empty rate data.
6. A DCGAN-based parking data patching device, the device comprising:
the system comprises a collected data acquisition unit, a data processing unit and a data processing unit, wherein the collected data acquisition unit is used for acquiring a pre-collected urban parking lot set and an urban geographic interest point set, and parking lots in the urban parking lot set comprise sample parking lots and parking lots to be repaired; wherein the parking lot to be repaired is a parking lot lacking parking data; the sample parking lot is a parking lot without lack of parking data;
the data processing unit is used for clustering the parking lots into a plurality of parking lot clusters according to the urban geographic interest point set and generating an empty rate curve image of the sample parking lot according to the parking data of the sample parking lot;
the network construction training unit is used for constructing a generating countermeasure network based on deep convolution, and training the generating countermeasure network based on deep convolution according to the empty rate curve image of the sample parking lot in the parking lot cluster so as to generate an empty rate curve simulation image corresponding to each parking lot cluster; and
and the parking data repairing unit is used for processing the empty rate curve simulation image corresponding to each parking lot cluster, mapping the empty rate curve simulation image into corresponding one-dimensional empty rate data, and repairing the parking data of the parking lot to be repaired in each parking lot cluster according to the one-dimensional empty rate data corresponding to each parking lot cluster.
7. The apparatus of claim 6, wherein the data processing unit comprises: the geographic interest point determining unit is used for determining geographic interest points around each parking lot in the urban parking lot set according to a preset tolerance radius; and
and the parking lot clustering unit is used for constructing the spatial characteristics corresponding to each parking lot according to the geographic interest points around each parking lot and carrying out high-dimensional clustering on the parking lots according to the spatial characteristics corresponding to each parking lot and a preset density-based spatial clustering algorithm.
8. The apparatus of claim 6, wherein the network construction training unit comprises:
the network construction unit is used for expanding a preset generative countermeasure network according to a preset supervised convolutional neural network so as to construct the generative countermeasure network based on the deep convolution, and the output layer of a generator in the generative countermeasure network based on the deep convolution is connected with a filter for normality check.
9. A computing device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the steps of the method according to any of claims 3 or 5 are implemented when the computer program is executed by the processor.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810869067.7A CN109214422B (en) | 2018-08-02 | 2018-08-02 | Parking data repairing method, device, equipment and storage medium based on DCGAN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810869067.7A CN109214422B (en) | 2018-08-02 | 2018-08-02 | Parking data repairing method, device, equipment and storage medium based on DCGAN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109214422A CN109214422A (en) | 2019-01-15 |
| CN109214422B true CN109214422B (en) | 2020-05-22 |
Family
ID=64987991
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810869067.7A Active CN109214422B (en) | 2018-08-02 | 2018-08-02 | Parking data repairing method, device, equipment and storage medium based on DCGAN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109214422B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110097060B (en) * | 2019-03-28 | 2021-04-06 | 浙江工业大学 | Open set identification method for trunk image |
| CN109978075B (en) * | 2019-04-04 | 2021-09-28 | 江苏满运软件科技有限公司 | Vehicle false position information identification method and device, electronic equipment and storage medium |
| CN110084158B (en) * | 2019-04-15 | 2021-01-29 | 杭州拓深科技有限公司 | Electric equipment identification method based on intelligent algorithm |
| CN110503834A (en) * | 2019-09-10 | 2019-11-26 | 上海科技大学 | The intelligent traffic administration system decision-making technique of Multiple Intersections collaboration is realized based on GAN |
| CN110751853B (en) * | 2019-10-25 | 2021-05-18 | 百度在线网络技术（北京）有限公司 | Parking space data validity identification method and device |
| WO2023004595A1 (en) * | 2021-07-27 | 2023-02-02 | 中国科学院深圳先进技术研究院 | Parking data recovery method and apparatus, and computer device and storage medium |
| CN113643564B (en) * | 2021-07-27 | 2022-08-26 | 中国科学院深圳先进技术研究院 | Parking data restoration method and device, computer equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106483842A (en) * | 2015-08-28 | 2017-03-08 | 易良碧 | A kind of high-precision intelligent is imitated nuclear signal and system and its method of work is occurred |
| CN106997380A (en) * | 2017-03-21 | 2017-08-01 | 北京工业大学 | Imaging spectrum safe retrieving method based on DCGAN depth networks |
| CN108053454A (en) * | 2017-12-04 | 2018-05-18 | 华中科技大学 | A kind of graph structure data creation method that confrontation network is generated based on depth convolution |
| CN108334941A (en) * | 2018-03-06 | 2018-07-27 | 陕西师范大学 | A kind of Bridge Crack image generation model fighting network based on production |
-
2018
- 2018-08-02 CN CN201810869067.7A patent/CN109214422B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106483842A (en) * | 2015-08-28 | 2017-03-08 | 易良碧 | A kind of high-precision intelligent is imitated nuclear signal and system and its method of work is occurred |
| CN106997380A (en) * | 2017-03-21 | 2017-08-01 | 北京工业大学 | Imaging spectrum safe retrieving method based on DCGAN depth networks |
| CN108053454A (en) * | 2017-12-04 | 2018-05-18 | 华中科技大学 | A kind of graph structure data creation method that confrontation network is generated based on depth convolution |
| CN108334941A (en) * | 2018-03-06 | 2018-07-27 | 陕西师范大学 | A kind of Bridge Crack image generation model fighting network based on production |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109214422A (en) | 2019-01-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109214422B (en) | Parking data repairing method, device, equipment and storage medium based on DCGAN | |
| CN105046235B (en) | The identification modeling method and device of lane line, recognition methods and device | |
| CN112232391B (en) | Dam crack detection method based on U-net network and SC-SAM attention mechanism | |
| CN110163213B (en) | Remote sensing image segmentation method based on disparity map and multi-scale depth network model | |
| Sohn et al. | An implicit regularization for 3D building rooftop modeling using airborne lidar data | |
| CN109033170B (en) | Data repairing method, device and equipment for parking lot and storage medium | |
| CN112287832A (en) | High-resolution remote sensing image-based urban illegal building detection method | |
| CN110942071A (en) | License plate recognition method based on license plate classification and LSTM | |
| CN111611918B (en) | Traffic flow data set acquisition and construction method based on aerial data and deep learning | |
| CN111524117A (en) | Tunnel surface defect detection method based on characteristic pyramid network | |
| CN112883820A (en) | Road target 3D detection method and system based on laser radar point cloud | |
| Yang et al. | A pattern‐based approach for matching nodes in heterogeneous urban road networks | |
| CN104899892A (en) | Method for quickly extracting star points from star images | |
| CN116503705B (en) | Fusion method of digital city multi-source data | |
| CN116222577B (en) | Closed loop detection method, training method, system, electronic equipment and storage medium | |
| CN111414878B (en) | Social attribute analysis and image processing method and device for land parcels | |
| CN116778146A (en) | Road information extraction method and device based on multi-mode data | |
| CN114283343B (en) | Map updating method, training method and device based on remote sensing satellite image | |
| CN116597270A (en) | Road damage target detection method based on attention mechanism integrated learning network | |
| CN114519819B (en) | Remote sensing image target detection method based on global context awareness | |
| Nguyen et al. | Toward real-time vehicle detection using stereo vision and an evolutionary algorithm | |
| CN112733782B (en) | Urban functional area identification method based on road network, storage medium and electronic equipment | |
| CN112507867B (en) | Lane line detection method based on EDLines line characteristics | |
| CN113643303A (en) | Three-dimensional image segmentation method based on two-way attention coding and decoding network | |
| WO2020024206A1 (en) | Dcgan-based parking data repairing method and apparatus, and device and storage medium |
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 |