TWI617998B - System and method for car number identification data filtering - Google Patents

Abstract

The invention discloses a system and a method for filtering vehicle number identification data, which is based on the identified series of vehicle numbers, based on the correlation between intersections and the historical track data of the vehicle number, combined with the concept of time window cutting. The historical behavior characteristic calculation and the vehicle number similarity comparison analysis are performed, and the unreasonable identification number is automatically filtered out.

Description

System and method for car number identification data filtering

The invention belongs to a system and a method for filtering vehicle number identification data, in particular to a network service provider (ISP) for utilizing equipment quality measurement, user declaration, repair history record, and manual test. Sources are combined to establish a classification model for the number of high-dimensional parameters.

The intersection license plate image automatic identification system is a common urban traffic and security system. It mainly records the intersection traffic image through the monitor deployed at the intersection, and then uses the license plate number identification function to convert the traffic image into a series of car numbers. After grasping the traffic image of the intersection and the identified vehicle number data, the police unit will be able to analyze the useful vehicle traffic trajectory information for crime investigation and early warning. However, the intersection monitor may be affected by various environmental factors (such as rain, light, wind and other factors), causing the recorded traffic image to be unclear or too blurred. These blurred traffic images will cause the vehicle number to be misidentified. A big increase, and the wrong identification results will result in the analysis of the wrong driving track results, but the burden of police handling and investigation.

Generally, in order to improve the probability that the correct car number is recognized, the license plate number identification function usually analyzes and identifies all the video frames in a specific time interval, and identifies a plurality of car numbers with high confidence values. First cut into a lot of car traffic images, for example, cut in seconds After cutting the image, each segment of the car stream image is converted into a plurality of still image frames. For example, after being converted into 16 still image frames, the license plate recognition module separately identifies the 16 still image frames and respectively generates a car number. All the car numbers will be counted and the confidence value (quantity /16) will be calculated. Then the car number with the confidence value greater than the threshold value will be used as the car number identification result of the car traffic image, and finally two sets of car numbers, such as ABC128, will be generated. , ABC1234. Although the above identification technology can ensure that the intersection number of the intersection can be recognized as much as possible, it is obvious that this method will generate redundant wrong number. For example, if ABC1234 is the vehicle number that passes through the intersection, then ABC128 It is the wrong car number, and this wrong car number will affect the results of the trajectory analysis. For example, the car number of the vehicle is ABC128. In order to solve the problem of redundant car numbers, in the prior art, by adding a third-party identification system, for example, an ETC identification system, the two-party comparison is performed, and the read ETC id is obtained through the ETC database to obtain the car number corresponding to the id. And compare with the car number identified by the car identification system, filter the unreasonable car number, leaving the correct car number. However, although the third-party identification system can improve the correctness of the car number identification, it requires a very high cost of system construction, and it is not obvious that all vehicles will install ETC electronic tags. Therefore, in practice, the above system can only be built in some areas, and it is not yet possible to promote it to all intersections of counties and cities.

In view of the shortcomings derived from the above-mentioned conventional methods, the invention has been improved and innovated, and after years of painstaking research, it has finally successfully developed a system and method for filtering the identification data of the vehicle number.

The present invention provides a system and method for filtering vehicle number identification data. In order to solve the above problems, the present invention discloses a vehicle identification data filtering. The system and method do not need to pass through a third-party identification system, such as ETC, etc., that can analyze a series of consecutive car numbers identified by the car identification system, and perform real-time operations through the intersection map, the historical track of the car number, and the like. Filter out the unreasonable car number and leave only a reasonable car number. In addition to saving the cost of construction at the intersection, it is only necessary to establish an intersection identification system. It is not necessary to add a third-party identification system, such as ETC. It can also improve the correct trajectory analysis rate and maintain the correct trajectory analysis results. Sex, can reduce the amount of data analysis.

A system for filtering vehicle identification data for the above purpose, comprising a plurality of intersection identification systems, which generate a series of identification numbers of the vehicle number and transmit the data to the vehicle number filtering system; the vehicle number filtering system is the receiving intersection The car identification system transmits the identification number of the car number, and uses the data of the database to calculate and filter out the wrong number of each intersection; the database is the data needed to provide the car number filtering analysis; and a cache system, The vehicle number data generated by the temporary vehicle number filtering system and the corresponding vehicle number identification event data are used.

The car number filtering system additionally includes a car number identification event receiving module, which is to receive the car number identification event data provided by the front end intersection car identification system; a car number set generating module is to identify each intersection car The series of car number identification events provided by the system are cut according to the time window, and a series of car number sets are generated and written into the cache system; a car number behavior characteristic calculation module is to access the data of the database for calculating the car. The reasonable car number evaluation value of each car number in the number set; a car number similarity calculation module is to calculate the similarity degree of any two car numbers, and generate a similar number of car numbers; a car number grouping module is According to the similar values of the two or two car numbers, the car number in the car number set is divided into one or more groups; and a reasonable car number selection module is based on the selection mechanism to select a reasonable car number from each group of the car number set. .

The database is a database of related road maps, which is to provide whether the two intersections are adjacent and the intersection identification system and the intersection are related. Material, and provide this information for the car number behavior characteristic calculation module for calculation; and a car number history track database, is to record the historical track data of each car number, and provide information to the car number behavior feature calculation module For calculation purposes.

A method for filtering vehicle identification data includes: step 1: receiving vehicle identification event data transmitted by a plurality of intersection identification systems; and step 2, sorting according to chronological order and using a preset value of the time window, the vehicle number The identification event data is converted into a plurality of sets of car numbers; step 3: determining whether the number of car numbers in the car number set is 1, if 1, then writing the corresponding car number identification event into the vehicle history track according to the received car number Database; Step 4: If it is greater than 1, then each set of car number data and corresponding car number identification event data is written into the cache system; Step 5: Perform feature behavior calculation for each car number in the car number set Calculate the reasonable car number evaluation value of the car number through the data of the intersection-related database and the historical track database of the vehicle number; Step 6. Calculate the similarity of the car number for each car number of each car number set, and The car number of the two or two car numbers is similar; step seven, the car number is divided into one or more groups according to the similar values of the two car numbers in the car number set, and the car number in each group is evaluated by the car number from large to small. After sorting, each group only leaves For the car number of the first two car number evaluation values, if the car number evaluation value is 0, do not take it; Step 8: Determine the number of car numbers in each group. If it is 1, according to the received car number, the corresponding car will be used. The number identification event is written into the vehicle history trajectory database; step IX. If 2, the vehicle number evaluation value of the two vehicle numbers in the group is determined. Whether it is the same, if yes, according to the received car number, the corresponding car number identification event is written into the vehicle history track database; Step 10, if no, calculate the two car numbers in the group (car number A) The length of the car code with the car number B); Step 11: Determine whether the lengths of the two car numbers (car number A and car number B) are different. If yes, the corresponding car will be based on the received car number. The number identification event is written into the vehicle history track database; step 12, if no, it is calculated whether the car number A is the left or right subset of the car number B completely matched, and if yes, the car number B is output, and according to The received car number is written into the vehicle history track database by the corresponding car number identification event; step 13 if no, the car number A is output and the corresponding car number identification event is written according to the received car number. Enter the vehicle history track database.

The system and method for filtering the vehicle number identification data provided by the invention have the following advantages when compared with other conventional technologies: 1. Using the intersection correlation data and the historical track data of the vehicle number, real-time calculation of a vehicle number at a specific intersection The probability value (reasonable car number evaluation value) appears; 2. The car number similarity algorithm is used for real-time operation, and the car number identified in the specific time window is grouped; 3. The car number in the group is unreasonable through the selection mechanism. The vehicle number is filtered; 4. The invention does not need to pass the third-party identification technology (for example, ETC), and only needs to use the historical track data of the vehicle number to combine the associated data between the intersections to filter the wrong number of the vehicle, and only retain the reasonable vehicle. number.

110‧‧‧ intersection identification system

120‧‧‧Car number filtration system

121‧‧‧Car number identification event receiving module

122‧‧‧Car number generation module

123‧‧‧Car number behavior characteristic calculation module

124‧‧‧Car number similarity calculation module

125‧‧‧Car number grouping module

126‧‧‧ Reasonable car number selection module

130‧‧‧Database

131‧‧‧ intersection map database

132‧‧‧Car number history track database

140‧‧‧Cache System

S210~S280‧‧‧Process

Please refer to the detailed description of the present invention and the accompanying drawings, which can further understand the technical content of the present invention and its effect. The related drawings are: FIG. 1 is a structural diagram of a system and method for filtering vehicle identification data according to the present invention; 2 is a flow chart of a system and method for filtering vehicle identification data; FIG. 3 is a schematic diagram of a system and method for filtering vehicle identification data; FIG. 4 is a system and method for filtering vehicle identification data according to the present invention; Schematic diagram of car number similarity calculation.

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Hereinafter, the present invention will be further described with reference to the accompanying drawings: FIG. 1 is a structural diagram of a system and method for filtering vehicle identification data, and a system for filtering vehicle identification data, which includes a plurality of intersection identification systems 110. In order to generate a series of car number identification event data, and transmit it to the car number filtering system 120; the car number filtering system 120 is a series of car number identification events transmitted by the receiving intersection car identification system 110, and utilizes the database 130. The data is calculated and filtered out to identify the wrong number of each intersection; the database 130 is the data required for providing the car number filtering analysis; and a cache system 140 is the car number set data generated by the temporary car number filtering system 120. With the corresponding car number Identify event data.

The car number filtering system 120 further includes a car number identification event receiving module 121, which is to receive the car number identification event data provided by the front end intersection car identification system 110; a car number set generating module 122 is to A series of car identification events provided by the intersection identification system 110 are cut according to the time window, and a series of car numbers are generated and written to the cache system 140; a car number behavior feature calculation module 123 is an access database. The data of 130 is used to calculate a reasonable car number evaluation value (PV value) of each car number in the car number set; a car number similarity calculation module 124 is to calculate the similarity degree of any two car numbers, and generate one The car number is similar to the value; a car number grouping module 125 is based on the similar values of the two car numbers, and the car number in the car number set is divided into one or more groups; and a reasonable car number selection module 126 is selected according to the selection. The mechanism picks out the reasonable number from each group of the car number set.

The database 130 further includes an intersection map database 131, which is to provide whether the two intersections are adjacent and the intersection identification system 110 and the intersection correspond to the data, and provide the information to the vehicle number behavior characteristic calculation module. 123 is used for calculation; and a car number history track database 132 is used to record the past historical track data of each car number, and provide the data to the car number behavior feature calculation module 123 for calculation.

In the embodiment of the intersection, a car number, a roadway identification system number, a car number identification time and additional identification data can constitute a car number identification event. A series of car identification event data will be transmitted to the car number identification event receiving module under the car number filtering system via the network.

The car number filtering system is connected to a plurality of intersection car identification systems, a database and a cache system for receiving a series of car number identification events of a plurality of intersection car identification systems, and selecting a reasonable car number identification event, The car number identification event is written to the database.

The car number identification event receiving module is connected with a plurality of intersection car identification systems for receiving a series of car number identification events transmitted by the intersection car identification system, and classifying a series of car number identification events according to the intersection car identification system number. After that, the car number identification events in the same category are sorted according to the time from far to near.

The car number generation module is connected to the cache system, and is used to classify and sort the car number recognition event receiving module by a series of car number identification events according to the time window value, and convert into a plurality of car number sets. In this embodiment, the car number set includes a car number set number, an intersection car identification system number, a car number set (containing at least one car number), and a corresponding car number identification event set; the generated car number set is written to the cache. system.

The car number behavior characteristic calculation module is used to calculate the probability value of the car number as a reasonable car number. The vehicle number behavior characteristic calculation module calculates the reasonable vehicle number evaluation value (PV value) of the vehicle number by reading the data of the intersection correlation map database and the vehicle number history trajectory database.

The car number similarity calculation module is connected to calculate the similarity degree of the two car numbers and generate similar values of the car numbers.

The car number grouping module is connected to the cache system and is used to group the car numbers in the car number set. The car number grouping module reads each car number set in the cache system, and calculates the similarity value of the car number of the two car numbers in the car number set in the car number set through the car number similarity calculation module, and according to the car number The similarity value divides the car number in the car number set into one or more groups, and generates the car number group number group data, and each group data contains one or more car numbers.

A reasonable license plate number selection module, connected to the cache system, is used to select a reasonable car number. The reasonable license plate number selection module obtains the car number group group result data generated by the car number grouping module, and analyzes each group data. Select a reasonable car number and write the car number identification event data corresponding to the car number into the car number history track database.

The database is connected to the car number filtering system and is used to provide the data required for car number filtering calculation and analysis, and to store reasonable car number identification event data.

The intersection correlation map database is used to provide the correspondence between the intersection and the intersection identification system and the adjacent relationship between the intersections. In this embodiment, the intersection map database can store the correspondence between the intersection and the intersection identification system (the intersection number, the corresponding intersection identification system number), and the adjacent intersection information (the intersection number, the adjacent intersection number). But it is not limited to this.

The vehicle number history track database is used to store the historical car number identification event data of each vehicle. In this embodiment, the information that can be stored in the vehicle number history track database includes the car number, the intersection identification system number, the identification time, and additional identification data, but is not limited thereto.

A cache system for storing car number data. In this embodiment, the car number set data that can be stored by the cache system includes a car number set including a car number set number, an intersection car identification system number, a car number set (including at least one car number), and a corresponding car number identification event set. But it is not limited to this.

Please refer to FIG. 2, which is a flowchart of a system and method for filtering vehicle number identification data, and a method for filtering vehicle number identification data, which comprises: Step 1: S210 receives a number transmitted by a plurality of intersection identification systems. Identify the event data; Step 2: S220 sorts according to the chronological order and uses the time window preset value to convert the car number identification event data into multiple sets of car numbers; Step 3: S221 determines whether the number of car numbers in the car number set is 1 If it is 1, the S270 writes the corresponding car number identification event into the vehicle history track database according to the received car number; Step 4: If it is greater than 1, the S230 writes each set of vehicle number data and the corresponding vehicle number identification event data into the cache system; in step 5, S240 performs characteristic behavior calculation for each vehicle number in the vehicle number set. Calculate the reasonable car number evaluation value of the car number through the data of the intersection-related database and the historical track database of the vehicle number; Step 6: S250 calculates the similarity of the car number for each car number of each car number set, and The car number of the two or two car numbers is similar; step seven, S260 divides the car number into one or more groups according to the similar values of the two car numbers in the car number set, and the car number in each group is evaluated by the car number. After the small sorting, each group only leaves the car number of the evaluation value of the first two major car numbers. If the car number evaluation value is 0, it is not taken; in step 8, S261 judges the number of car numbers in each group. If it is 1, then S280 According to the received car number, the corresponding car number identification event is written into the vehicle historical trajectory database; step IX, if 2, S262 determines whether the car number evaluation values of the two car numbers in the group are the same, if Yes, the S280 writes the corresponding car number identification event to the car according to the received car number. Historical trajectory database; Step 10, if no, S270 calculates the car code length of the two car numbers (car number A and car number B) in the group; step XI, S271 judges two car numbers (car number) Is the length of A and car number B) different? If yes, S280 writes the corresponding car number identification event into the vehicle history track database according to the received car number; step 12, if no, then S272 Calculate whether the car number A is the left or right subset of the car number B that exactly matches, and if so, the car number B, And S280 writes the corresponding vehicle number identification event into the vehicle historical trajectory database according to the received vehicle number; step 13, if not, outputs the vehicle number A, and S270 according to the received vehicle number, corresponding to The car number identification event is written into the vehicle history track database.

It can be seen from the above process that in the embodiment, a series of car number identification event data transmitted by each intersection car identification system is received at S210, and each car number identification event data may include an intersection car identification system number, a car number, and a Car identification time and additional identification data (model, car color, etc.), but not limited to this; all received car identification events will be grouped according to the intersection identification system number, the car number with the same intersection identification system number The identification events will be in the same group, and then the car number identification events in each group will be sorted from far to near according to the car identification time.

The S220 car number set generation module receives a series of group-numbered car number identification events transmitted by the car number identification event receiving module, and cuts and converts the plurality of sets of car number set data according to the time window value. The time window value can be adjusted according to actual needs. In this embodiment, the preset time is 1 minute, and a series of car number identification events will be in minutes, and the car number identification event data is cut according to the identification time in the car number identification event. For a plurality of car number sets, a series of car number identification events are converted into car number sets. As shown in Table 1 and Table 2, the car number identification event data of Table 1 is converted into two sets of car numbers in Table 2. S221 then judges the number of car numbers in the car number set. If the number of car numbers in the car number set is 1 (such as the car number set 2 in Table 2), then go directly to step S270; if the car number in the car number set If the number is greater than 1, the process proceeds to the next step S230.

Table 1 A series of car number identification event data sections after grouping

Step S230, according to the car number set data provided in step S220, all the received car number set data is written into the cache system. In this embodiment, the car number set data written to the cache system needs to include at least the car number set number, the intersection car identification system number, the intersection number, the car number set, and the corresponding car number identification event set, but is not limited thereto.

Step S240, reading a set of car numbers from the cache system, and calculating a reasonable car number evaluation value (PV value) for all the car numbers in the car number set. In this embodiment, the PV value calculation is performed by using the intersection map database and the vehicle historical trajectory database, and the trajectory data of the vehicle number is read for the past 7 days for analysis, but is not limited to reading only the past 7 days. data. Taking the car number set 1 in Table 2 as an example, the PV values of the respective car numbers are calculated based on the historical trajectory data of the car number A, the car number B, and the car number C in the past 7 days. The PV value calculation formula is as shown in the following formula (1):

among them Represents the number of days in which the car number p has appeared in the past 7 days and the number of car identification events on that day is 1. It represents the number of days in which the car number p has appeared in the past 7 days and the number of car number identification events on that day is greater than 1. It represents the number of times that the car number p has passed at the intersection i of the past 7 sunrises. On behalf of the car number p , the score of the first adjacent intersection inter(i, NB1) of this intersection i in the past 7 sunrises, On behalf of the car number p , the score of the second adjacent intersection inter(i, NB2) of this intersection i in the past 7th sunrise, the concept of the adjacent intersection is shown in Figure 3, as shown in intersection 1 in Figure 3. At the intersection of the first floor of intersection 1, there are intersection 2, intersection 3 and intersection 4, and the adjacent intersection of the second floor of intersection 1 has intersection 5, intersection 6, intersection 7, intersection 8 and intersection 9. In addition, w ₁ , w ₂ , and w ₃ are weight values used to adjust the calculation result of the PV value. Through the above formula calculation, a reasonable car number evaluation value PV(p, i) of the car number p at the intersection i can be obtained.

In addition, The score calculation can refer to the following formula:

among them It is used to calculate the number of times that the car number p has been recognized by the car identification system corresponding to the first adjacent intersection of the intersection i in the past 7 days. The license plate number is used to calculate whether the license plate number p in the past 7 days of track just before the intersection i is the corresponding vehicle system identification to resolution, then the adjacent intersection intersection i corresponding vehicle identification has been identified to the system In addition, Path(p, 1) and Path(p, 2) in the formula respectively represent the cut set of the historical track of the car number p , and the representation and meaning thereof are shown in Table 3 below. For example, one of the days of the car number p in the past 7 days passes through the intersection 1, 2, 3, 4, 5, 2, 1, respectively, then Path(p, 1) represents the collection of the track into a group of single-point intersections. Path(p, 2) is a collection of trajectories cut into a group of two intersections.

In addition, The score calculation can refer to the following formula:

among them It is used to calculate the number of times the car number p has been identified by the car identification system corresponding to the second adjacent intersection of the intersection i in the past 7 days, as shown in the following formula:

and Whether the license plate number is used to calculate the license number p in the track immediately in the last 7 days according to the time sequence corresponding to the intersection i, i of the first layer adjacent to the junction intersection, and a second layer adjacent to the junction intersection identified vehicle i The number of times the system recognizes, as shown in the following formula:

For the content represented by Path(p, 3) , refer to the description of Table 3.

In step S250, a pairwise similarity calculation is performed for the car numbers in each set of car numbers, and a car number similarity value is generated. In the present embodiment, the car number similarity value is calculated as follows: sim ( A, B ) = αLCS ( A, B ) + BLOC ( A, B ) , where α + β =1 and α, β 0

The vehicle similarity value is calculated by using the Longest common subsequence (LCS) and the location calculation (LOC) based similarity factors. The calculation method of the LCS value is shown in FIG. 4A. Each car number will be cut into a series of car code character strings, for example, car number A is BC1284, car code character string is listed as [B, C, 1, 2, 8, 4], car code length is 6 , and the car number B is ABC1234, the car code character string is listed as [A, B, C, 1, 2, 3, 4], the car code length = 7; and the maximum similar character string can be obtained through LCS calculation. Listed as [B, C, 1, 2, 4], the length is 5, and the LCS similarity value LCS (A, B) of car number B and car number A is 5/6. The LOC similarity calculation method is shown in the part of FIG. 4B. Based on the car code character string of the car number A, the car code character string of the car number B is aligned to the left to the car code word of the car number A. Yuan string, compare the number of the same position character in the two series of columns (result = 0); also compare the car code character string of car number B to the car code character string of right-numbered car number A, compare The number of the same string of characters in the same string (result = 5), from the left LOC ^left (A, B) and the right LOC ^right (A, B) two of the two scores, such as the following formula Shown: LOC ( A,B )=max{ LOC ^left ( A,B ) ,LOC ^right ( A,B )}

It can be concluded that the LOC similarity value LOC(A, B) of car number B and car number A is 5/6.

Step S260, grouping the car numbers in the car number set, and the grouping basis is referring to the similarity value of the two car numbers. In this embodiment, if the car number similarity value is greater than 0.5, it will be divided into the same group; step S261 will determine the number of the car number, if the number of car number = 1, then go to step S280, otherwise go to step S262; step S262 The car number PV value comparison is performed. If the PV values of the two car numbers are the same, the car number is output, and the process proceeds to step S280, otherwise the process proceeds to step S270.

In step S270, the car code length of each car number in the group is calculated, and the process proceeds to step S271. Step S271, comparing the relationship between the length of the car code and the PV value of the car number, and supposing that the car numbers sorted by the PV value from high to low in the group are the car number A and the car number B, and step S271 compares whether the two car numbers are different, such as the car number. Whether the length of the A car code is greater than or equal to the car code length of the car number B, if yes, the car number A is output and proceeds to step S280; otherwise, for the received car number (the car number is 1 or more groups), respectively, the cache system is obtained. After identifying the event data, the car number is written into the car number. Historical track database.

In the above step S272, it is calculated whether the car code [B, C, 1, 2, 8, 4] of the car number A is the left of the car number B car code [A, B, C, 1, 2, 3, 4] or The right complete subset, where the left or right complete subset meaning is as follows, assuming the car number B as an example, the 5 code character string [A, B, C, 1, 2] is the left complete of the car number B The subset, [B, C, 1, 2, 3, 4] is the right full subset of the car number B; if the car number A is the left or right complete subset of the car number B, the car number B is output and the steps are entered. S280, otherwise the car number A is outputted and proceeds to step S280.

The detailed description of the present invention is intended to be illustrative of a preferred embodiment of the invention, and is not intended to limit the scope of the invention. The patent scope of this case.

To sum up, this case is not only innovative in terms of technical thinking, but also has many of the above-mentioned functions that are not in the traditional methods of the past. It has fully complied with the statutory invention patent requirements of novelty and progressiveness, and applied for it according to law. Approved this invention patent application, in order to invent invention, to the sense of virtue.

Claims (4)

A system for filtering vehicle identification data includes: a plurality of intersection identification systems for generating a series of vehicle identification event data and transmitting the data to a vehicle number filtering system; the vehicle number filtering system receives the intersection vehicle Identifying the series of car number identification events transmitted by the system, and using the data of the database to calculate and filter out the car numbers that identify the wrong intersections. The car number filtering system includes: a car number grouping module, which is based on two or two cars. a similar value, the car number in the car number set generated based on the car number identification event data is divided into one or more groups; and a reasonable car number selection module is selected from the car number set according to the selection mechanism In the middle, pick out a reasonable car number; the database provides the information needed for the car number filtering analysis; and a caching system that temporarily stores the car number data generated by the car number filtering system and the corresponding car number identification Event data. For example, the vehicle number identification data filtering system described in claim 1 of the patent scope, wherein the vehicle number filtering system further comprises: a vehicle number identification event receiving module, which receives the vehicle number provided by the front end intersection identification system. Identifying the event data; a car number generation module is configured to cut a series of car number identification events provided by each intersection car identification system according to the time window, generate a series of car number sets and write to the cache system; The vehicle number behavior characteristic calculation module is to access the data of the database for calculating the reasonable vehicle number evaluation value of each vehicle number in the vehicle number set; a vehicle number similarity calculation module is to calculate the Renji number The similarity degree, and generate a similar value of the car number; a car number grouping module, based on the similar values of the two car numbers, the car number set The car number is divided into 1 or more groups; and a reasonable car number selection module is based on the selection mechanism to select a reasonable car number from each group of the car number set. For example, the system for filtering the identification number of the vehicle number mentioned in the first paragraph of the patent application, wherein the database further comprises: a road map of the intersection map, which is to provide whether the two intersections are adjacent and the intersection identification system and intersection Correspond to the relationship data, and provide this information to the car number behavior feature calculation module for calculation; and a car number history track database, to record the historical trajectory data of each car number, and provide information to the car number behavior characteristics calculation The module is used for calculation. A method for filtering vehicle identification data includes: step 1: receiving vehicle identification event data transmitted by a plurality of intersection identification systems; and step 2, sorting according to chronological order and using time windows to identify event data of the vehicle number Converted into multiple sets of car numbers; Step 3: Determine whether the number of car numbers in the car number set is 1, if 1, then write the corresponding car number identification event into the vehicle history track database according to the received car number; Step 4: If it is greater than 1, the data of each set of vehicle number and the corresponding identification number of the vehicle number are written into the cache system; step 5: performing characteristic behavior calculation for the vehicle number in each set of vehicle numbers, through the intersection Calculate the reasonable car number evaluation value of the car number from the data of the associated database and the historical track database of the car number; Step 6. Calculate the similarity of the car number for each car number of each car number set, and generate two or two The car number of the car number is similar to the value; Step 7. According to the similar values of the two or two car numbers in the car number set, divide the car number into 1 or more groups. The car numbers in each group are sorted from large to small, and only the first two are left in each group. The car number of the car number evaluation value is not taken if the car number evaluation value is 0; Step 8: Judging the number of car numbers in each group, if it is 1, according to the received car number, the corresponding car number identification event Write to the vehicle history track database; Step 9. If it is 2, determine whether the car number evaluation values of the two car numbers in the group are the same. If yes, the corresponding car number will be based on the received car number. The identification event is written into the vehicle historical trajectory database; step 10, if not, the car code length of the two car numbers (car number A and car number B) in the group is calculated; step XI, judging the two car numbers ( Is the length of car number A and car number B) different? If yes, the corresponding car number identification event is written into the vehicle history track database according to the received car number; step 12, if no, then Calculate whether the car number A is the left or right subset of the car number B completely matched, and if yes, output the car number B, and according to the received car Write the corresponding car number identification event into the vehicle history track database; Step 13. If no, output the car number A and write the corresponding car number identification event into the vehicle historical track data according to the received car number. Library.