CN110598758A - Training modeling method, vehicle charging method, management system, and storage medium - Google Patents
Training modeling method, vehicle charging method, management system, and storage medium Download PDFInfo
- Publication number
- CN110598758A CN110598758A CN201910782795.9A CN201910782795A CN110598758A CN 110598758 A CN110598758 A CN 110598758A CN 201910782795 A CN201910782795 A CN 201910782795A CN 110598758 A CN110598758 A CN 110598758A
- Authority
- CN
- China
- Prior art keywords
- license plate
- vehicle
- information
- image data
- parking lot
- 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.)
- Pending
Links
Classifications
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07B—TICKET-ISSUING APPARATUS; FARE-REGISTERING APPARATUS; FRANKING APPARATUS
- G07B15/00—Arrangements or apparatus for collecting fares, tolls or entrance fees at one or more control points
- G07B15/02—Arrangements or apparatus for collecting fares, tolls or entrance fees at one or more control points taking into account a variable factor such as distance or time, e.g. for passenger transport, parking systems or car rental systems
Abstract
The invention relates to a parking lot vehicle characteristic training modeling method, a vehicle charging method, a management system and a storage medium, wherein the parking lot vehicle characteristic training modeling method comprises the following steps: acquiring first image data at an entrance of a parking lot, identifying first license plate information, acquiring first characteristic information, and adding the first characteristic information into a field record; acquiring second image data at the exit of the parking lot, identifying second license plate information, judging whether the second license plate information can be successfully matched with one first license plate information in the field record or not, and acquiring second characteristic information when the second license plate information is successfully matched with one first license plate information in the field record; judging whether the first license plate information corresponding to the first characteristic information with the highest similarity with the second characteristic information is the same as the second license plate information corresponding to the second characteristic information; if not, training the general vehicle model to establish the special vehicle model. By implementing the technical scheme of the invention, the accurate special vehicle model can be established by continuously correcting the general vehicle model so as to improve the accuracy of feature matching.
Description
Technical Field
The invention relates to the field of Intelligent Transportation (ITS), in particular to a parking lot vehicle characteristic training modeling method, a vehicle charging method, a management System and a storage medium.
Background
With the development of the parking lot entrance and exit identification technology, the vehicle identification accuracy rate is greatly improved, and the mainstream identification modes comprise license plate identification, card swiping identification and the like at present, wherein the card swiping identification user experience is poor, the user gradually exits the market, and the license plate identification accuracy rate also reaches the bottleneck, so that the license plate identification accuracy rate is hardly improved. With the increasing growth of urban vehicles, parking lot managers also need to improve the entrance and exit identification accuracy to reduce the charging problem and congestion pressure caused by entrance and exit identification errors, and improving the vehicle identification accuracy of the entrance and exit of the parking lot is a problem to be solved urgently at present.
Disclosure of Invention
The invention aims to solve the technical problem of providing a parking lot vehicle feature training modeling method, a vehicle charging method, a management system and a storage medium aiming at the defect of low vehicle identification accuracy rate of a parking lot entrance and exit in the prior art.
The technical scheme adopted by the invention for solving the technical problems is as follows: a parking lot vehicle feature training modeling method is constructed, and comprises the following steps:
a driving-in step: acquiring first image data of an incoming vehicle at an entrance of a parking lot, identifying first license plate information of the incoming vehicle according to the first image data, inputting the first image data into a general vehicle model, acquiring first characteristic information of the incoming vehicle according to an output result of the general vehicle model, and adding the first license plate information and the first characteristic information into a presence record;
a step of exiting: acquiring second image data of a vehicle at an exit of a parking lot, identifying second license plate information of the vehicle according to the second image data, judging whether the second license plate information can be successfully matched with one first license plate information in the existing record, inputting the second image data into the general vehicle model when the second license plate information is successfully matched with one first license plate information in the existing record, and acquiring second characteristic information of the vehicle according to an output result of the general vehicle model;
and (3) feature matching: determining the similarity between the second characteristic information and each first characteristic information of the field records, judging whether the first license plate information corresponding to the first characteristic information with the highest similarity is the same as the second license plate information corresponding to the second characteristic information, and entering a characteristic training step if the first license plate information is not the same as the second license plate information;
a characteristic training step: and taking the first image data and the second image data as training samples, and training the general vehicle model according to the training samples to establish a special vehicle model.
Preferably, in the feature matching step, if the first license plate information corresponding to the first feature information with the highest similarity is different from the second license plate information corresponding to the second feature information, the method further includes:
recording different times, calculating the accuracy of feature matching, judging whether the total number of the training samples reaches a preset number, and/or judging whether the accuracy of the feature matching reaches a preset value, and taking the current special vehicle model as a final special vehicle model when the total number of the training samples reaches the preset number, and/or the accuracy of the feature matching reaches the preset value.
Preferably, determining the similarity of the second characteristic information and each first characteristic information of the presence record comprises:
and determining the similarity between the second characteristic information and each first characteristic information of the field record according to a cosine similarity algorithm.
Preferably, the entering step further comprises:
and judging whether the vehicle enters the field or not according to the first image data, and if so, identifying first license plate information of the vehicle entering the field according to the first image data.
Preferably, the first image data comprises a plurality of incoming images, and,
inputting the first image data into a general vehicle model, and acquiring first characteristic information of the incoming vehicle according to an output result of the general vehicle model, including:
screening the plurality of entrance images according to an image noise estimation algorithm, inputting the screened entrance images into a general vehicle model, and acquiring first characteristic information of the entrance vehicle according to an output result of the general vehicle model;
and/or the presence of a gas in the gas,
the second image data includes a plurality of outgoing images, and,
inputting the second image data into a general vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the general vehicle model, including:
and screening the plurality of departure images according to an image noise estimation algorithm, inputting the screened departure images into a general vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the general vehicle model.
The present invention also constructs a vehicle charging method for a parking lot, including:
acquiring first image data of an incoming vehicle at an entrance of a parking lot, identifying first license plate information of the incoming vehicle according to the first image data, inputting the first image data into a special vehicle model, acquiring first characteristic information of the incoming vehicle according to an output result of the special vehicle model, and adding the first license plate information and the first characteristic information into a presence record, wherein the special vehicle model is acquired according to the parking lot vehicle characteristic training modeling method;
acquiring second image data of a vehicle at an exit of a parking lot, identifying second license plate information of the vehicle according to the second image data, judging whether the second license plate information can be successfully matched with one first license plate information in the field record, inputting the second image data to the special vehicle model when the matching is not successful, and acquiring second characteristic information of the vehicle according to an output result of the special vehicle model;
determining the similarity between the second characteristic information and each first characteristic information of the presence record, and judging the first characteristic information with the highest similarity;
and acquiring the driving-in/driving-out time according to the first characteristic information and the second characteristic information, and calculating parking cost according to the driving-in/driving-out time so as to deduct fees for related accounts.
The invention also constitutes a parking lot management system comprising a processor implementing the steps of the parking lot vehicle feature training modeling method described above, and/or the steps of the parking lot vehicle charging method described above, when executing a stored computer program.
The invention also constitutes a storage medium storing a computer program which, when executed by a processor, carries out the steps of the above-described parking lot vehicle feature training modeling method, and/or the steps of the above-described parking lot vehicle charging method.
The present invention also constructs a parking lot management system including:
the system comprises an image acquisition module, a storage module and a display module, wherein the image acquisition module is used for acquiring first image data of an incoming vehicle at an entrance of a parking lot; acquiring second image data of the vehicles on the scene at the exit of the parking lot;
the license plate recognition module is used for recognizing first license plate information of the incoming vehicle according to the first image data and adding the first license plate information into a field record; identifying second license plate information of the vehicle on the spot according to the second image data, and judging whether the second license plate information can be successfully matched with one first license plate information in the record on the spot;
the characteristic identification module is used for inputting the first image data into a general vehicle model, acquiring first characteristic information of the entering vehicle according to an output result of the general vehicle model, and adding the first characteristic information into a presence record; when the second license plate information can be successfully matched with the first license plate information in the field record, inputting the second image data into a general vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the general vehicle model;
the feature matching module is used for determining the similarity between the second feature information and each piece of first feature information of the field records, judging whether the first license plate information corresponding to the first feature information with the highest similarity is the same as the second license plate information corresponding to the second feature information or not, and entering the feature training module if the first license plate information is not the same as the second license plate information;
and the characteristic training module is used for taking the first image data and the second image data as training samples and training the general vehicle model according to the training samples so as to establish a special vehicle model.
Preferably, if the first license plate information corresponding to the first feature information with the highest similarity is different from the second license plate information corresponding to the second feature information, the parking lot management system further includes:
and the statistical module is used for recording different times, calculating the accuracy of feature matching, judging whether the total number of the training samples reaches a preset number, and/or judging whether the accuracy of the feature matching reaches a preset value, and taking the current special vehicle model as a final special vehicle model when the total number of the training samples reaches the preset number, and/or the accuracy of the feature matching reaches the preset value.
Preferably, a billing module is also included, and,
the characteristic identification module is further used for inputting the first image data into a special vehicle model and acquiring first characteristic information of the incoming vehicle according to an output result of the special vehicle model; when the second license plate information is unsuccessfully matched with any first license plate information in the field records, inputting the second image data into a special vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the special vehicle model;
the feature matching module is further configured to determine similarity between the second feature information and each piece of first feature information of the presence record, and determine the first feature information with the highest similarity;
the charging module is used for acquiring the driving-in/driving-out time according to the first characteristic information and the second characteristic information, and calculating parking cost according to the driving-in/driving-out time so as to deduct fees for related accounts.
According to the technical scheme provided by the invention, in the early stage of parking lot operation (automatic training modeling stage), the characteristics of the personalized environment of the parking lot, the direction and the angle of the coming vehicle and the like can be automatically and efficiently deeply learned according to the environmental characteristics of different parking lots, and the personalized environmental characteristics of the parking lot are eliminated, so that the general vehicle model is continuously corrected, an accurate special vehicle model can be established, and the accuracy of feature matching is improved.
Furthermore, after a special vehicle model is established, the accuracy of feature matching is obviously improved, so that when an entering vehicle and a leaving vehicle are matched, two matching modes of license plate matching and feature matching can be combined, when the license plate matching mode fails, the feature matching mode is adopted, the dual-mode linkage mode can achieve the purpose of accurate matching, accurate charging and releasing of the vehicles are completed, an unattended parking lot scene is really built, and the passing efficiency of the parking lot at an entrance and an exit is greatly improved.
Drawings
In order to illustrate the embodiments of the invention more clearly, the drawings that are needed in the description of the embodiments will be briefly described below, it being apparent that the drawings in the following description are only some embodiments of the invention, and that other drawings may be derived from those drawings by a person skilled in the art without inventive effort. In the drawings:
FIG. 1 is a flow chart of a first embodiment of a parking lot vehicle feature training modeling method of the present invention;
FIG. 2 is a flow chart of a first embodiment of a vehicle charging method for a parking lot according to the present invention;
FIG. 3 is a logical block diagram of a parking lot management system according to a first embodiment of the present invention;
fig. 4 is a logical structure diagram of a parking lot management system according to a second embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It is noted that, with the development of artificial intelligence, although some vehicle identification at the entrance and exit of parking lots has started to introduce a vehicle feature identification method, at present, a large number of pictures are collected through various channels, and then are trained by using a general data sample, and a basic general vehicle model is established, and then the general model is applied to each parking lot. However, in practical application, it is found that the environments of the parking lots are different, the lane angles, the lane widths and the like are not unified, and although the general vehicle model has a certain matching degree, the practical effect is not good, so that the general vehicle model cannot well meet the use requirements of the different parking lots, and the accuracy of the feature extraction and the matching of the vehicle is not high.
Fig. 1 is a flowchart of a first embodiment of the parking lot vehicle feature training and modeling method according to the present invention, where the parking lot vehicle feature training and modeling method of the embodiment includes the following steps:
step S11: acquiring first image data of an entering vehicle at an entrance of a parking lot, identifying first license plate information of the entering vehicle according to the first image data, inputting the first image data into a general vehicle model, acquiring first characteristic information of the entering vehicle according to an output result of the general vehicle model, and adding the first license plate information and the first characteristic information into a presence record, preferably, adding the first image data into the presence record;
exit step S12: acquiring second image data of a vehicle at an exit of a parking lot, identifying second license plate information of the vehicle according to the second image data, judging whether the second license plate information can be successfully matched with one first license plate information in the existing record, inputting the second image data into the general vehicle model when the second license plate information is successfully matched with one first license plate information in the existing record, and acquiring second characteristic information of the vehicle according to an output result of the general vehicle model;
feature matching step S13: determining the similarity between the second characteristic information and each first characteristic information of the field records, judging whether the first license plate information corresponding to the first characteristic information with the highest similarity is the same as the second license plate information corresponding to the second characteristic information, and entering a characteristic training step if the first license plate information is not the same as the second license plate information;
a feature training step S14: and taking the first image data and the second image data as training samples, and training the general vehicle model according to the training samples to establish a special vehicle model.
In this embodiment, it should be noted that:
firstly, because the access & exit department in present parking area all is provided with the camera, when sensing the vehicle and advancing/going out of the court, all can carry out the image shooting to the vehicle to acquire the image data of access & exit department.
Secondly, regarding the license plate information recognition of the vehicle, the implementation manner thereof includes but is not limited to: in the traditional image processing machine learning method, deep learning method and the like, mature license plate recognition models exist in the current market, the recognition accuracy rate is about 99.5%, and details are not repeated here. Regarding the characteristic information identification of the vehicle, the realization mode is as follows: a general vehicle model (base model) is introduced into the parking management system in advance, and the general vehicle model is obtained by modeling a standard lane, including but not limited to: VGG, ResNet, SE-ResNet, SSD, YOLO, CornetNet, and the like. Although the effect of vehicle feature recognition using the general vehicle model is not good in the early stage, the general vehicle model still has a certain matching degree. After the image data of the vehicle at the entrance and the exit are obtained, the corresponding image data can be input into the general vehicle model, and then the characteristic information of the corresponding vehicle can be obtained according to the output of the general vehicle model.
And finally, when the first image data and the second image data need to be trained, the environmental features and the vehicle outline part are firstly removed, then the features of the vehicle are extracted, and then the features of the vehicle are automatically trained, so that the special vehicle model is established.
According to the technical scheme of the embodiment, in the early stage (automatic training modeling stage) of parking lot operation, the characteristics of the personalized environment of the parking lot, the direction and the angle of the coming vehicle and the like can be automatically and efficiently deeply learned according to the environmental characteristics of different parking lots, the personalized environmental characteristics of the parking lot are eliminated, and therefore the general vehicle model is continuously corrected, an accurate special vehicle model can be established, and the accuracy of feature matching is improved.
Further, in the step S12 of going out, if it is determined that the second license plate information of the vehicle on the spot cannot be successfully matched with any of the first license plate information in the record on the spot, that is, the license plate matching is not successful, the original manual processing mode may be switched, and the feature matching of the vehicle is not performed this time.
Further, in the feature matching step S13, if it is determined that the first license plate information corresponding to the first feature information having the highest similarity to the second feature information of the departure vehicle is the same as the second license plate information corresponding to the second feature information, that is, the second feature information of the departure vehicle can be matched with one of the first feature information in the presence record and is consistent with the result of license plate matching, it is indicated that the model is accurate, and the first image data and the second image data do not need to be trained, and only the number of times of feature matching needs to be updated.
In an optional embodiment, in the feature matching step S13, if the first license plate information corresponding to the first feature information with the highest similarity is not the same as the second license plate information corresponding to the second feature information, the parking lot vehicle feature training modeling method of the present invention further includes:
recording different times, calculating the accuracy of feature matching, judging whether the total number of the training samples reaches a preset number, and/or judging whether the accuracy of the feature matching reaches a preset value, and taking the current special vehicle model as a final special vehicle model when the total number of the training samples reaches the preset number, and/or the accuracy of the feature matching reaches the preset value.
In this embodiment, whether the modeling is completed or not is determined by judging the total number of the training samples and/or judging the accuracy of the feature matching, and preferably, the modeling is determined to be completed only when the total number of the training samples and the accuracy of the feature matching meet requirements. When the training samples are gradually increased to be large enough, the feature matching accuracy rate is higher and higher, and the set sample value number and accuracy rate are reached (the values can be set), the automatic training modeling is considered to be completed. Of course in other embodiments, the modeling may be determined to be complete by determining whether a training period has expired, such as two months. It should be understood that whether the modeling is completed may also be determined by other artificially set conditions, and the above example is only for example and should not be construed as a specific limitation.
In an alternative embodiment, the entering step S11 further includes:
and judging whether the vehicle enters the field or not according to the first image data, and if so, identifying first license plate information of the vehicle entering the field according to the first image data.
In this embodiment, after the first image data of the vehicle entering the parking lot entrance is obtained, it is determined whether the vehicle belongs to the vehicle to be admitted, for example, by identifying the vehicle type of the first image data, the vehicle type includes but is not limited to the following: the model identification method comprises the steps of identifying a model of a vehicle, wherein the model of the vehicle can be directly obtained by sending first image data into a pre-established model of the vehicle, and the model of the vehicle can be obtained directly, wherein the model module of the vehicle is based on a SqueezeNet basic network model and can be obtained by training sample data (for example, a model data set collected by 4000 entrances and exits of a parking lot) according to a caffe frame. And then, judging whether the identified vehicle type belongs to the vehicle type of the access entrance, and only when the identified vehicle type belongs to the access vehicle, identifying the first license plate information of the vehicle and releasing the vehicle. If the identified vehicle type belongs to the vehicle type which is forbidden to enter, the vehicle owner can be reminded of forbidding to enter by the public information.
In an optional embodiment, the determining the similarity between the second characteristic information and each first characteristic information of the presence record includes:
and determining the Similarity between the second characteristic information and each first characteristic information of the presence record according to a Cosine Similarity (Cosine Similarity) algorithm.
In this embodiment, the cosine similarity algorithm is used to calculate the similarity between the second feature information of the departing vehicle and the first feature information of the entering vehicle, and whether the two feature information match is determined by setting a reasonable similarity threshold.
The cosine similarity algorithm is explained as follows: the cosine similarity is also called cosine similarity, and the similarity of two vectors is evaluated by calculating the cosine value of the included angle of the two vectors. Cosine similarity maps vectors into a vector space according to coordinate values, the most common of which is a two-dimensional space. The similarity between the two vectors is then measured by measuring the cosine of the angle between them. The cosine of an angle of 0 degrees is 1, while the cosine of any other angle is not greater than 1, and its minimum value is-1. The cosine of the angle between the two vectors thus determines whether the two vectors point in approximately the same direction. When the two vectors have the same direction, the cosine similarity value is 1; when the included angle of the two vectors is 90 degrees, the value of the cosine similarity is 0; the cosine similarity has a value of-1 when the two vectors point in completely opposite directions. The result is independent of the length of the vector, only the pointing direction of the vector. Cosine similarity is commonly used in the positive space, and thus gives values between 0 and 1.
The cosine value between two vectors can be found by using the euclidean dot product formula:
A·B＝||A||||B||cosθ
wherein A, B is two vectors, | a | | is the absolute value of a, | | B | | is the absolute value of B, and θ is the included angle of the vector A, B.
For a given two attribute vectors a and B, the remaining chord similarity is given by the dot product and the vector length, as follows:
wherein A isiRepresenting the respective components of vector A, BiRepresenting the components of vector a, similarity is given in the range-1 to 1, where-1 means that the two vectors point in exactly the opposite direction, 1 means that their points are identical, 0 usually means that they are independent, and the value between them means an intermediate similarity or dissimilarity.
Additionally, for text matching, attribute vectors A and B are typically word frequency vectors in the document. Cosine similarity can be viewed as a way to normalize the length of a file during comparison. In the case of information retrieval, since the frequency of one word (TF-IDF weight) cannot be a negative number, the cosine similarity of the two documents ranges from 0 to 1. Also, the angle between the frequency vectors of the two words cannot be greater than 90 °. The cosine values range between [ -1,1], the closer the value is to 1, the closer the directions of the two vectors are represented; the closer they approach-1, the more opposite their direction; close to 0 means that the two vectors are nearly orthogonal.
When the similarity between the second characteristic information of the vehicle leaving the parking lot and the first characteristic information of the vehicle entering the parking lot is determined through cosine similarity, the obtained vehicle characteristic is double-type long text data, so the cosine similarity calculation method can be just used for calculating the similarity of cosine included angles of texts, two vectors are established for the two texts according to the data of the two texts, and the cosine values of the two vectors are calculated, so that the similarity condition of the two texts in a statistical method can be known.
Further, in order to improve the accuracy of license plate recognition and feature recognition, when a vehicle arrives at an entrance, the corresponding vehicle is subjected to image capturing for multiple times, then several images with the best image quality are screened from the multiple images, and feature recognition is performed on the screened images.
Based on this, in an optional embodiment, the first image data comprises a plurality of incoming pictures and the second image data comprises a plurality of outgoing pictures. Furthermore, it is possible to provide a liquid crystal display device,
the entry step S11 further includes:
screening the plurality of entrance images according to an image noise estimation algorithm, inputting the screened entrance images into a general vehicle model, and acquiring first characteristic information of the entrance vehicle according to an output result of the general vehicle model;
the exiting step S12 includes:
and screening the plurality of departure images according to an image noise estimation algorithm, inputting the screened departure images into a general vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the general vehicle model.
In this embodiment, regarding the noise estimation algorithm, the following is explained:
in order to ensure the quality of the images used for training, a noise estimation algorithm (Filter-based approach Using arithmetric estimation) is used for filtering the images with poor quality, a model is adopted for noise estimation, and the following brief introduction is made for the model. The algorithm model is based on the fact that an image is sensitive to a noise statistic of a Laplacian Mask (Laplacian Mask) because an image edge structure has strong second-order difference characteristics, and the algorithm carries out convolution operation through a kernel consisting of two Laplacian masks. The laplacian mask is a filter matrix which is established according to a laplacian operator and has a plurality of gradient directions such as horizontal and vertical directions, and is used for detecting edge features of the image in all directions. A typical matrix form is 3 x 3 dimensions. In practical application, the weight and dimension of the array elements can be selected according to the intensity and width of the image edge, and the total sum of the array elements is kept to be zero.
Then directly through a convolution of the image:
where W, H is the width and height of the image, I (x, y) is the value of each pixel of the image, LAIs the pramipexole operator, as a convolution kernel.
However, the above formula is improved by once multiplying each pixel by a power, and the result obtained by the algorithm is calculated once by the following formula, and the overall performance is good.
Further, because the general vehicle model is continuously corrected to establish the special vehicle model according to the environmental characteristics of the parking lot such as lane angles, widths and the like in the early stage of operation of the parking lot, at the moment, the accuracy rate of characteristic matching is obviously improved, therefore, after the special vehicle model is established, when an incoming vehicle is matched with an outgoing vehicle, two matching modes of license plate matching and characteristic matching can be combined, when the license plate matching mode fails, the characteristic matching mode is adopted, the dual-mode linkage mode can achieve the aim of accurate matching, thereby completing accurate charging and releasing of vehicles, really setting up scenes of unattended parking lots, and greatly improving the passing efficiency of the entrances and exits of the parking lots.
Fig. 2 is a flowchart of a first embodiment of a vehicle charging method for a parking lot according to the present invention, the vehicle charging method of the embodiment including:
s21, acquiring first image data of an incoming vehicle at an entrance of a parking lot, identifying first license plate information of the incoming vehicle according to the first image data, inputting the first image data into a special vehicle model, acquiring first characteristic information of the incoming vehicle according to an output result of the special vehicle model, and adding the first license plate information and the first characteristic information into a field record, wherein the special vehicle model is acquired according to the parking lot vehicle characteristic training modeling method;
s22, acquiring second image data of a vehicle at an exit of a parking lot, identifying second license plate information of the vehicle according to the second image data, judging whether the second license plate information can be successfully matched with one first license plate information in the field record, inputting the second image data into the special vehicle model when the second license plate information is not successfully matched with the first license plate information in the field record, and acquiring second characteristic information of the vehicle according to an output result of the special vehicle model;
s23, determining the similarity between the second characteristic information and each first characteristic information of the field records, and judging the first characteristic information with the highest similarity;
and S24, acquiring the driving-in/driving-out time according to the first characteristic information and the second characteristic information, and calculating parking cost according to the driving-in/driving-out time so as to deduct fees for related accounts.
In this embodiment, in step S22, if the second license plate information can be successfully matched with the first license plate information in the presence record, the license plate is considered to be successfully matched, and in this case, in step S24, the driving-in/driving-out time may be obtained according to the first license plate information and the second license plate information, and then the charging and deduction may be performed. And if the second license plate information cannot be successfully matched with any first license plate information in the field records, the license plate matching is considered to be unsuccessful, the license plate recognition is wrong, and further the vehicle features can be adopted for matching and passing.
By implementing the embodiment of the invention, after the special vehicle model is established, the accuracy of feature matching is obviously improved, so that when the vehicles entering the parking lot are matched with the vehicles leaving the parking lot, two matching modes of license plate matching and feature matching can be combined, and when the license plate matching mode fails, the feature matching mode is adopted, and the dual-mode linkage mode can achieve the aim of accurate matching, thereby completing accurate charging and releasing of the vehicles, really building an unattended parking lot scene, and greatly improving the passing efficiency of the entrance and the exit of the parking lot.
The invention also constitutes a parking lot management system comprising a processor and a memory storing a computer program, said processor realizing the steps of the parking lot vehicle feature training modeling method described above, and/or the steps of the parking lot vehicle charging method described above, when executing the stored computer program.
The invention also constitutes a storage medium storing a computer program which, when executed by a processor, carries out the steps of the above-described parking lot vehicle feature training modeling method, and/or the steps of the above-described parking lot vehicle charging method.
Fig. 3 is a logical structure diagram of a parking lot management system according to a first embodiment of the present invention, the parking lot management system of this embodiment includes: the system comprises an image acquisition module 11, a license plate recognition module 12, a feature recognition module 13, a feature matching module 14 and a feature training module 15. The image acquisition module 11 is used for acquiring first image data of an incoming vehicle at an entrance of a parking lot; acquiring second image data of the vehicles on the scene at the exit of the parking lot; the license plate recognition module 12 is configured to recognize first license plate information of the incoming vehicle according to the first image data, and add the first license plate information to a field record; identifying second license plate information of the vehicle on the spot according to the second image data, and judging whether the second license plate information can be successfully matched with one first license plate information in the record on the spot; the feature identification module 13 is configured to input the first image data to a general vehicle model, acquire first feature information of the incoming vehicle according to an output result of the general vehicle model, and add the first feature information to a presence record; when the second license plate information can be successfully matched with the first license plate information in the field record, inputting the second image data into a general vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the general vehicle model; the feature matching module 14 is configured to determine similarity between the second feature information and each first feature information of the presence record, preferably determine similarity according to a cosine similarity algorithm, and determine whether the first license plate information corresponding to the first feature information with the highest similarity is the same as the second license plate information corresponding to the second feature information, and if not, execute a feature training module; and the characteristic training module is used for taking the first image data and the second image data as training samples and training the general vehicle model according to the training samples so as to establish a special vehicle model.
Further, if the first license plate information corresponding to the first feature information with the highest similarity is different from the second license plate information corresponding to the second feature information, the parking lot management system further comprises a statistical module 16, wherein the statistical module 16 is used for recording different times, calculating the accuracy rate of feature matching, and judging whether the total number of the training samples reaches a preset number, and/or judging whether the accuracy rate of the feature matching reaches a preset value, and taking the current special vehicle model as the final special vehicle model when the total number of the training samples reaches the preset number, and/or the accuracy rate of the feature matching reaches the preset value.
Fig. 4 is a logical structure diagram of a second embodiment of the parking lot management system according to the present invention, the parking lot management system of this embodiment includes: the system comprises an image acquisition module 11, a license plate recognition module 12, a feature recognition module 13, a feature matching module 14, a feature training module 15, a statistic module 16 and a charging module 17. Moreover, in this embodiment, after the dedicated vehicle model is established, the feature recognition module 13 is further configured to input the first image data to the dedicated vehicle model, and obtain first feature information of the incoming vehicle according to an output result of the dedicated vehicle model; when the second license plate information is unsuccessfully matched with any first license plate information in the field records, inputting the second image data into a special vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the special vehicle model; the feature matching module 14 is further configured to determine similarity between the second feature information and each first feature information of the presence record, and determine the first feature information with the highest similarity; the charging module 17 is configured to obtain the entry/exit time according to the first characteristic information and the second characteristic information, and calculate the parking fee according to the entry/exit time to deduct the fee from the relevant account.
The system further comprises an admission judgment module, wherein the admission judgment module is used for judging whether the vehicle enters the vehicle according to the first image data, and preferably, judging whether the vehicle enters the vehicle by identifying the vehicle type of the first image data. And the license plate recognition module is used for recognizing the first license plate information of the vehicle entering according to the first image data when judging that the vehicle is allowed to enter.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.
Claims (11)
1. A parking lot vehicle feature training modeling method is characterized by comprising the following steps:
a driving-in step: acquiring first image data of an incoming vehicle at an entrance of a parking lot, identifying first license plate information of the incoming vehicle according to the first image data, inputting the first image data into a general vehicle model, acquiring first characteristic information of the incoming vehicle according to an output result of the general vehicle model, and adding the first license plate information and the first characteristic information into a presence record;
a step of exiting: acquiring second image data of a vehicle at an exit of a parking lot, identifying second license plate information of the vehicle according to the second image data, judging whether the second license plate information can be successfully matched with one first license plate information in the existing record, inputting the second image data into the general vehicle model when the second license plate information is successfully matched with one first license plate information in the existing record, and acquiring second characteristic information of the vehicle according to an output result of the general vehicle model;
and (3) feature matching: determining the similarity between the second characteristic information and each first characteristic information of the field records, judging whether the first license plate information corresponding to the first characteristic information with the highest similarity is the same as the second license plate information corresponding to the second characteristic information, and entering a characteristic training step if the first license plate information is not the same as the second license plate information;
a characteristic training step: and taking the first image data and the second image data as training samples, and training the general vehicle model according to the training samples to establish a special vehicle model.
2. The parking lot vehicle feature training and modeling method according to claim 1, wherein in the feature matching step, if first license plate information corresponding to the first feature information with the highest similarity is different from second license plate information corresponding to the second feature information, the method further comprises:
recording different times, calculating the accuracy of feature matching, judging whether the total number of the training samples reaches a preset number, and/or judging whether the accuracy of the feature matching reaches a preset value, and taking the current special vehicle model as a final special vehicle model when the total number of the training samples reaches the preset number, and/or the accuracy of the feature matching reaches the preset value.
3. The parking lot vehicle feature training modeling method according to claim 1, wherein determining a similarity of the second feature information to each first feature information of the presence record includes:
and determining the similarity between the second characteristic information and each first characteristic information of the field record according to a cosine similarity algorithm.
4. The parking lot vehicle feature training modeling method of claim 1, wherein the entering step further comprises:
and judging whether the vehicle enters the field or not according to the first image data, and if so, identifying first license plate information of the vehicle entering the field according to the first image data.
5. The parking lot vehicle feature training modeling method according to claim 1, wherein said first image data includes a plurality of entrance images, and further,
inputting the first image data into a general vehicle model, and acquiring first characteristic information of the incoming vehicle according to an output result of the general vehicle model, including:
screening the plurality of entrance images according to an image noise estimation algorithm, inputting the screened entrance images into a general vehicle model, and acquiring first characteristic information of the entrance vehicle according to an output result of the general vehicle model;
and/or the presence of a gas in the gas,
the second image data includes a plurality of outgoing images, and,
inputting the second image data into a general vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the general vehicle model, including:
and screening the plurality of departure images according to an image noise estimation algorithm, inputting the screened departure images into a general vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the general vehicle model.
6. A vehicle charging method for a parking lot, comprising:
acquiring first image data of an incoming vehicle at an entrance of a parking lot, identifying first license plate information of the incoming vehicle according to the first image data, inputting the first image data into a special vehicle model, acquiring first characteristic information of the incoming vehicle according to an output result of the special vehicle model, and adding the first license plate information and the first characteristic information into a presence record, wherein the special vehicle model is acquired according to the parking lot vehicle characteristic training modeling method of any one of claims 1-5;
acquiring second image data of a vehicle at an exit of a parking lot, identifying second license plate information of the vehicle according to the second image data, judging whether the second license plate information can be successfully matched with one first license plate information in the field record, inputting the second image data to the special vehicle model when the matching is not successful, and acquiring second characteristic information of the vehicle according to an output result of the special vehicle model;
determining the similarity between the second characteristic information and each first characteristic information of the presence record, and judging the first characteristic information with the highest similarity;
and acquiring the driving-in/driving-out time according to the first characteristic information and the second characteristic information, and calculating parking cost according to the driving-in/driving-out time so as to deduct fees for related accounts.
7. A parking lot management system comprising a processor, characterized in that the processor, when executing a stored computer program, performs the steps of the parking lot vehicle feature training modeling method of any one of claims 1-5, and/or the steps of the vehicle charging method of a parking lot of claim 6.
8. A storage medium storing a computer program, characterized in that the computer program, when being executed by a processor, carries out the steps of the parking lot vehicle feature training modeling method of any one of claims 1 to 5, and/or the steps of the vehicle charging method for a parking lot of claim 6.
9. A parking lot management system, comprising:
the system comprises an image acquisition module, a storage module and a display module, wherein the image acquisition module is used for acquiring first image data of an incoming vehicle at an entrance of a parking lot; acquiring second image data of the vehicles on the scene at the exit of the parking lot;
the license plate recognition module is used for recognizing first license plate information of the incoming vehicle according to the first image data and adding the first license plate information into a field record; identifying second license plate information of the vehicle on the spot according to the second image data, and judging whether the second license plate information can be successfully matched with one first license plate information in the record on the spot;
the characteristic identification module is used for inputting the first image data into a general vehicle model, acquiring first characteristic information of the entering vehicle according to an output result of the general vehicle model, and adding the first characteristic information into a presence record; when the second license plate information can be successfully matched with the first license plate information in the field record, inputting the second image data into a general vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the general vehicle model;
the feature matching module is used for determining the similarity between the second feature information and each piece of first feature information of the field records, judging whether the first license plate information corresponding to the first feature information with the highest similarity is the same as the second license plate information corresponding to the second feature information or not, and entering the feature training module if the first license plate information is not the same as the second license plate information;
and the characteristic training module is used for taking the first image data and the second image data as training samples and training the general vehicle model according to the training samples so as to establish a special vehicle model.
10. The parking lot management system according to claim 9, wherein if the first license plate information corresponding to the first characteristic information having the highest similarity is not identical to the second license plate information corresponding to the second characteristic information, the parking lot management system further comprises:
and the statistical module is used for recording different times, calculating the accuracy of feature matching, judging whether the total number of the training samples reaches a preset number, and/or judging whether the accuracy of the feature matching reaches a preset value, and taking the current special vehicle model as a final special vehicle model when the total number of the training samples reaches the preset number, and/or the accuracy of the feature matching reaches the preset value.
11. The parking lot management system according to claim 9, further comprising a billing module, and,
the characteristic identification module is further used for inputting the first image data into a special vehicle model and acquiring first characteristic information of the incoming vehicle according to an output result of the special vehicle model; when the second license plate information is unsuccessfully matched with any first license plate information in the field records, inputting the second image data into a special vehicle model, and acquiring second characteristic information of the departure vehicle according to an output result of the special vehicle model;
the feature matching module is further configured to determine similarity between the second feature information and each piece of first feature information of the presence record, and determine the first feature information with the highest similarity;
the charging module is used for acquiring the driving-in/driving-out time according to the first characteristic information and the second characteristic information, and calculating parking cost according to the driving-in/driving-out time so as to deduct fees for related accounts.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910782795.9A CN110598758A (en) | 2019-08-23 | 2019-08-23 | Training modeling method, vehicle charging method, management system, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910782795.9A CN110598758A (en) | 2019-08-23 | 2019-08-23 | Training modeling method, vehicle charging method, management system, and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110598758A true CN110598758A (en) | 2019-12-20 |
Family
ID=68855515
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910782795.9A Pending CN110598758A (en) | 2019-08-23 | 2019-08-23 | Training modeling method, vehicle charging method, management system, and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110598758A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112560752A (en) * | 2020-12-23 | 2021-03-26 | 杭州趣链科技有限公司 | License plate recognition training method and device based on federal learning and related equipment |
| CN113920596A (en) * | 2021-10-17 | 2022-01-11 | 绿城科技产业服务集团有限公司 | License plate recognition data re-matching method and system for unattended parking lot |
| CN116935659A (en) * | 2023-09-12 | 2023-10-24 | 四川遂广遂西高速公路有限责任公司 | High-speed service area bayonet vehicle auditing system and method thereof |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104021375A (en) * | 2014-05-29 | 2014-09-03 | 银江股份有限公司 | Model identification method based on machine learning |
| CN105575130A (en) * | 2016-02-25 | 2016-05-11 | 浙江宇视科技有限公司 | Unattended parking implementation method and device |
| CN106778777A (en) * | 2016-11-30 | 2017-05-31 | 成都通甲优博科技有限责任公司 | A kind of vehicle match method and system |
| CN206470954U (en) * | 2017-02-20 | 2017-09-05 | 厦门中诚国信电子科技有限公司 | A kind of parking lot intelligent barrier gate system |
| CN107704833A (en) * | 2017-10-13 | 2018-02-16 | 杭州电子科技大学 | A kind of front vehicles detection and tracking based on machine learning |
| CN107730617A (en) * | 2017-02-04 | 2018-02-23 | 西安艾润物联网技术服务有限责任公司 | Unlicensed vehicle toll method and apparatus |
| CN108388888A (en) * | 2018-03-23 | 2018-08-10 | 腾讯科技（深圳）有限公司 | A kind of vehicle identification method, device and storage medium |
| US20180268238A1 (en) * | 2017-03-14 | 2018-09-20 | Mohammad Ayub Khan | System and methods for enhancing license plate and vehicle recognition |
| CN109377572A (en) * | 2018-12-12 | 2019-02-22 | 杭州华云科技有限公司 | A kind of management control method and managing device of unattended cloud parking |
| CN109657596A (en) * | 2018-12-12 | 2019-04-19 | 天津卡达克数据有限公司 | A kind of vehicle appearance component identification method based on deep learning |
-
2019
- 2019-08-23 CN CN201910782795.9A patent/CN110598758A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104021375A (en) * | 2014-05-29 | 2014-09-03 | 银江股份有限公司 | Model identification method based on machine learning |
| CN105575130A (en) * | 2016-02-25 | 2016-05-11 | 浙江宇视科技有限公司 | Unattended parking implementation method and device |
| CN106778777A (en) * | 2016-11-30 | 2017-05-31 | 成都通甲优博科技有限责任公司 | A kind of vehicle match method and system |
| CN107730617A (en) * | 2017-02-04 | 2018-02-23 | 西安艾润物联网技术服务有限责任公司 | Unlicensed vehicle toll method and apparatus |
| CN206470954U (en) * | 2017-02-20 | 2017-09-05 | 厦门中诚国信电子科技有限公司 | A kind of parking lot intelligent barrier gate system |
| US20180268238A1 (en) * | 2017-03-14 | 2018-09-20 | Mohammad Ayub Khan | System and methods for enhancing license plate and vehicle recognition |
| CN107704833A (en) * | 2017-10-13 | 2018-02-16 | 杭州电子科技大学 | A kind of front vehicles detection and tracking based on machine learning |
| CN108388888A (en) * | 2018-03-23 | 2018-08-10 | 腾讯科技（深圳）有限公司 | A kind of vehicle identification method, device and storage medium |
| CN109377572A (en) * | 2018-12-12 | 2019-02-22 | 杭州华云科技有限公司 | A kind of management control method and managing device of unattended cloud parking |
| CN109657596A (en) * | 2018-12-12 | 2019-04-19 | 天津卡达克数据有限公司 | A kind of vehicle appearance component identification method based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| ALGORITHMC: "图像噪声估计算法", 《博客园》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112560752A (en) * | 2020-12-23 | 2021-03-26 | 杭州趣链科技有限公司 | License plate recognition training method and device based on federal learning and related equipment |
| CN112560752B (en) * | 2020-12-23 | 2024-03-26 | 杭州趣链科技有限公司 | License plate recognition training method and device based on federal learning and related equipment |
| CN113920596A (en) * | 2021-10-17 | 2022-01-11 | 绿城科技产业服务集团有限公司 | License plate recognition data re-matching method and system for unattended parking lot |
| CN116935659A (en) * | 2023-09-12 | 2023-10-24 | 四川遂广遂西高速公路有限责任公司 | High-speed service area bayonet vehicle auditing system and method thereof |
| CN116935659B (en) * | 2023-09-12 | 2023-12-08 | 四川遂广遂西高速公路有限责任公司 | High-speed service area bayonet vehicle auditing system and method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110674688B (en) | Face recognition model acquisition method, system and medium for video monitoring scene | |
| CN104166841A (en) | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network | |
| CN105787466B (en) | A kind of fine recognition methods and system of type of vehicle | |
| CN112016605B (en) | Target detection method based on corner alignment and boundary matching of bounding box | |
| CN110598758A (en) | Training modeling method, vehicle charging method, management system, and storage medium | |
| CN106778659B (en) | License plate recognition method and device | |
| US11790499B2 (en) | Certificate image extraction method and terminal device | |
| CN104809452A (en) | Fingerprint identification method | |
| CN113947766B (en) | Real-time license plate detection method based on convolutional neural network | |
| CN104978567A (en) | Vehicle detection method based on scenario classification | |
| CN104809453A (en) | Authentication method based on fingerprints | |
| CN109840520A (en) | A kind of invoice key message recognition methods and system | |
| CN106156777A (en) | Textual image detection method and device | |
| CN113128507B (en) | License plate recognition method and device, electronic equipment and storage medium | |
| CN112215190A (en) | Illegal building detection method based on YOLOV4 model | |
| CN110969164A (en) | Low-illumination imaging license plate recognition method and device based on deep learning end-to-end | |
| CN109784171A (en) | Car damage identification method for screening images, device, readable storage medium storing program for executing and server | |
| CN112784834A (en) | Automatic license plate identification method in natural scene | |
| CN115880260A (en) | Method, device and equipment for detecting base station construction and computer readable storage medium | |
| Lin et al. | A lightweight, high-performance multi-angle license plate recognition model | |
| CN112597995B (en) | License plate detection model training method, device, equipment and medium | |
| CN116704490B (en) | License plate recognition method, license plate recognition device and computer equipment | |
| CN113486715A (en) | Image reproduction identification method, intelligent terminal and computer storage medium | |
| CN113537211A (en) | Deep learning license plate frame positioning method based on asymmetric IOU | |
| CN116665321A (en) | Parking lot vehicle management method based on edge nano-tube technology |
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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 12 / F, building a, Tsinghua Information Port R & D building, Keyuan Road, Xili street, Nanshan District, Shenzhen City, Guangdong Province Applicant after: Weilong Technology (Guangdong) Co.,Ltd. Address before: 518057 12-storey Building A, Tsinghua Information Port, Keyuan Road, Xili Street, Nanshan District, Shenzhen City, Guangdong Province Applicant before: WEILONG JINYI TECHNOLOGY (SHENZHEN) Co.,Ltd. |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191220 |