A method for identifying vehicles in the passenger-seeking state on a road section based on taxi GPS data

Abstract

The present invention discloses a method for identifying vehicles in a passenger-seeking state on a road section based on taxi GPS data. The present invention uses the taxi GPS data and map road data to screen out vehicles in passenger-seeking behavior and normal driving vehicles by calculating a speed threshold. The present invention provides a higher quality sample for estimating the vehicle speed of the road section by screening out vehicles in passenger-seeking behavior on the road section and discarding the data of vehicles in this mode in the sample, which can help the related work of vehicle behavior identification to a certain extent.

Description

Technical Field

[0001] The invention belongs to the technical field of data mining, and in particular relates to a method for distinguishing vehicles in a passenger-seeking state on a road section based on taxi GPS data, and improving the quality of vehicle speed samples. Background Art

[0002] In recent years, intelligent transportation based on floating vehicle data has become one of the research hotspots among the majority of followers. During this period, many new technologies have emerged and have been used by researchers in transportation systems, demonstrating their practical application value.

[0003] As a key parameter of many intelligent transportation systems, the accuracy of road section speed is crucial for intelligent transportation systems. One of the key factors affecting the accuracy of estimating the average speed of a road section using floating car GPS data is that the driving modes of vehicles on the road section are diverse. For example, there may be vehicles in the customer-seeking mode on the road section, that is, vehicles that drive at a low speed in a certain period of time on the road section to find customers. But in fact, the speed of vehicles in the customer-seeking mode is intentionally controlled by the driver and cannot reflect the actual speed of vehicles on the road section. Therefore, in some application scenarios, it is necessary to exclude vehicles in the customer-seeking mode and only count the speed of normal driving vehicles. Therefore, in order to accurately estimate the speed of sections that include and do not include the customer-seeking mode, it is necessary to first filter out the driving behavior of the vehicle. The GPS data of taxis may be inaccurate due to factors such as hardware. To address this problem, some existing technologies use the spatial distance and time difference of taxi trajectory points to calculate the vehicle speed instead of the GPS vehicle speed. Based on the vehicle speed information, some potentially valuable information can be mined. However, how to filter out vehicles that engage in passenger-seeking behavior on a road section and discard the data of vehicles in this mode in the sample to provide higher quality samples for estimating the speed of vehicles on the road section is a technical problem that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to solve the technical problem that vehicles in the state of looking for customers cannot be identified in the prior art, and to provide a method for identifying vehicles in the state of looking for customers in a road section based on taxi GPS data.

[0005] The specific technical solutions adopted by the present invention are as follows:

[0006] A method for identifying vehicles in a passenger-seeking state on a road section based on taxi GPS data, the steps of which are as follows:

[0007] S1. Obtain a historical driving data set of taxis operating in a target area, wherein the historical driving data set is composed of road speed subsets corresponding to different road sections in the target area, and the road speed subset of each road section contains a speed list of different taxis during driving on the road section, wherein the speed list is composed of instantaneous speed sampling points regularly uploaded by a vehicle-mounted terminal device of the vehicle, and each instantaneous speed sampling point is synchronously recorded by the vehicle-mounted terminal device with a corresponding vehicle number, as well as the GPS positioning coordinates and passenger status when recording the instantaneous speed sampling point;

[0008] S2, pre-processing each road speed subset in the historical driving data set to remove abnormal instantaneous speed samples caused by abnormal upload of the vehicle terminal device;

[0009] S3, for each speed subset of the road section processed by S2, the number of taxis whose passenger-carrying status changes from empty to loaded during driving on the road section is counted, and the speed subsets of the road section whose number is not higher than the set number threshold are eliminated;

[0010] S4, for each speed subset of each road section retained after processing in S3, calculate the distance between the GPS positioning coordinates corresponding to each instantaneous speed sampling point and the nearest traffic light in front of the vehicle in the road section, and eliminate all abnormal instantaneous speed samples in the deceleration zone before the traffic light;

[0011] S5. For each speed subset of each road section processed by S4, calculate the ratio between the average value of the instantaneous speed sampling points of all the vehicles with no passengers and the average value of the instantaneous speed sampling points of all the vehicles with passengers, sort all the speed subsets of the road section according to the ratio, and select the top-ranked ones as candidate speed subsets of the road section;

[0012] S6, clustering all instantaneous vehicle speed sampling points with empty vehicle status in all candidate road speed subsets into 2 clusters, and taking the weighted sum of the two cluster centers obtained by clustering as the speed threshold;

[0013] S7. For the taxis operating on each road section in the target area, the instantaneous speed of each taxi within the set time interval is compared with the speed threshold. If a taxi outside the deceleration zone before the traffic light is empty and its instantaneous speed is continuously less than the speed threshold, and there is no other taxi with passengers on the same road section and in the same time period whose instantaneous speed is continuously less than the speed threshold, then the taxi is identified as a vehicle seeking passengers.

[0014] Preferably, in S2, the method for identifying the abnormal instantaneous vehicle speed samples caused by the abnormal upload of the vehicle terminal device in the road section vehicle speed subset is as follows:

[0015] S21. Traverse the speed list corresponding to each taxi in the road speed subset. If there is an i-th instantaneous speed sampling point v in the speed list, i is 0 and there are z-1 consecutive 0s after the sampling point, and v i+z ≠0, then calculate the speed mutation threshold v in the speed list GPS_e for:

[0016]

[0017] Where: R t-1 represents the instantaneous speed of the same taxi collected at the moment before the instantaneous speed sampling point v1, z≥1;

[0018] S22, judge v i+z ≥v GPS_e Is it true? If true, then v i 、v i+z And the instantaneous vehicle speed sampling points between the two are determined to be abnormal instantaneous vehicle speed samples caused by abnormal uploading of the vehicle-mounted terminal device.

[0019] Preferably, the deceleration zone before the traffic light is a road section within a range of 30 to 50 m before the traffic light.

[0020] Preferably, the clustering adopts a k-means clustering algorithm.

[0021] Preferably, the calculation formula of the speed threshold is:

[0022]

[0023] Where: represents the speed threshold obtained by weighted summation after the i-th clustering, and Respectively represent the two cluster centers obtained after the i-th clustering, where and They are and The number of samples in the corresponding cluster.

[0024] Preferably, in S6, clustering is repeated multiple times with the number of clusters being 2 but the initial cluster centers being random, and a weighted sum of the two cluster centers obtained in each clustering is taken as a speed threshold, and the average of the speed thresholds obtained from multiple clusterings is taken as the final speed threshold.

[0025] Preferably, the quantity threshold is 0.

[0026] Preferably, the historical driving data set is obtained by fusing recorded data from an on-board terminal device installed on a taxi and road network map data.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention uses the speed of taxis, GPS data and map road data to calculate the speed threshold to screen out vehicles that are seeking customers and vehicles that are driving normally. The present invention provides a higher quality sample for estimating the speed of the road section by screening out vehicles that are seeking customers on the road section and discarding the data of vehicles in this mode in the sample, which can help the related work of vehicle behavior recognition to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is an example diagram of abnormal speed samples of customer-seeking behavior;

[0030] Figure 2 It is the data structure table of the historical driving data set;

[0031] Figure 3 It is a line graph of different vehicle speeds on two road sections. DETAILED DESCRIPTION

[0032] The present invention is further described and illustrated below in conjunction with the accompanying drawings and specific embodiments. The technical features of each embodiment of the present invention can be combined accordingly without conflicting with each other.

[0033] In a preferred embodiment of the present invention, a method for identifying vehicles in a seeking-customer state on a road section based on taxi GPS data is provided, the main purpose of which is to identify the speed characteristics of vehicles in a seeking-customer state from a driving data set uploaded by a large number of operating taxis through on-board terminal devices, and to find out vehicles in a seeking-customer state based on the identified speed characteristics, so as to avoid such vehicles from affecting the speed statistics of the entire road section, and to provide higher quality samples for estimating the road section speed.

[0034] like Figure 1 The figure shows a typical abnormal speed sample generated when a taxi is engaged in passenger-seeking behavior. In the non-passenger-seeking state, the vehicle's driving speed on the road section is often faster, while in the passenger-seeking state, its driving speed is often intentionally controlled by the driver to be in a lower range for a period of time. The present invention automatically identifies the passenger-seeking behavior of a taxi based on this speed feature.

[0035] The specific implementation process of the method for identifying vehicles in a passenger-seeking state on a road section based on taxi GPS data in this embodiment is described in detail below.

[0036] S1. Obtain a historical driving data set of taxis operating in a target area, wherein the historical driving data set is composed of road speed subsets corresponding to different road sections in the target area, and the road speed subset of each road section contains a speed list of different taxis during driving on the road section, and the so-called speed list is composed of instantaneous speed sampling points regularly uploaded by the vehicle-mounted terminal device of the vehicle, and each instantaneous speed sampling point is synchronously recorded by the vehicle-mounted terminal device with the corresponding vehicle number and the GPS positioning coordinates and passenger status when recording the instantaneous speed sampling point.

[0037] In this embodiment, the historical driving data set is obtained by fusing the recorded data of the vehicle-mounted terminal device installed on the taxi and the road network map data. The so-called vehicle-mounted terminal device of the taxi refers to a terminal device installed on the taxi for sending vehicle status information to the remote monitoring center in real time. The specific model is not limited, and it needs to have GPS positioning function and the monitoring function of the vehicle's own direction and passenger status. In this embodiment, the data recorded and uploaded by the vehicle-mounted terminal device of the taxi includes the taxi's serial ID, longitude and latitude coordinates, upload time, etc., and its instantaneous speed can be determined based on the time difference between adjacent longitude and latitude coordinates and upload time. The serial ID of each taxi is unique, and its corresponding license plate number can be confirmed based on the serial ID. For each taxi, the track point data recorded is continuous in time, and a track point represents a record of floating vehicle data, which can be recorded as TP. The track point corresponds to the longitude and latitude coordinates, data upload time, instantaneous vehicle speed, and passenger status at different times, expressed as: TP DATA Indicates the time of the record, TP GPS Indicates the recorded longitude and latitude coordinates of the vehicle, that is, its longitude and latitude information, TP STATUS Indicates the passenger status of the vehicle. A value of 1 indicates that there are passengers, and a value of 0 indicates that the vehicle is empty. V Indicates the instantaneous speed of the recorded vehicle. Based on the continuous trajectory point data recorded by the vehicle, combined with the road network map data of the operating area and the starting point information of each road section, the trajectory point data of all vehicles can be divided according to the road section. A section of a city road is recorded as section R, and the section it passes through during the driving process can be determined according to the positioning information in the trajectory points of different vehicles. When a taxi passes through a road section, it will regularly upload its instantaneous speed at fixed intervals, and each uploaded instantaneous speed will be recorded as an instantaneous speed sampling point. A taxi passes through a section R. i In the process of , all the uploaded instantaneous speed sampling points can be called the vehicle on the road section R i There will be a series of taxis passing through a road section, so a road section R i The list of speeds of all taxis passing through is called the road segment speed subset i is the road segment number. The entire historical driving data set can be represented as a set of road segment speed subsets V for all road segments. R , like Figure 2 In the data structure of the historical driving data set, the primary key is represented by the road section ID. Each road section ID corresponds to a road section speed subset. A road section speed subset further uses the license plate number as a secondary key. Each secondary key contains a speed list {v1, v2, v3, ...}. The speed list consists of the instantaneous speed sampling points v uploaded periodically by the vehicle terminal device when the vehicle passes through the road section. j composition.

[0038] When the taxi's onboard terminal device uploads GPS and other data, data delays and omissions may occur due to hardware or network factors. The instantaneous speed of the vehicle is actually calculated by two adjacent longitude and latitude coordinates and the time difference of the records. Therefore, if there is a delay in uploading GPS positioning data, the speed during the delay stage will be recorded as 0, and the distance the vehicle moves during the data delay will be superimposed on the instantaneous speed corresponding to the first data uploaded after the data delay occurs, resulting in an abnormally large value for the speed value. Therefore, based on the calculation of instantaneous speed by trajectory distance and time difference, we need to first screen the instantaneous speed samples generated by abnormal GPS uploads.

[0039] S2. Preprocess each speed subset of each road section in the above historical driving data set to eliminate abnormal instantaneous speed samples caused by abnormal upload of the vehicle terminal device.

[0040] In this embodiment, the method for identifying the abnormal instantaneous vehicle speed samples caused by abnormal upload of the vehicle terminal device in the road section vehicle speed subset is as follows:

[0041] S21. Traverse the speed list {v1, v2, v3, ...} corresponding to each taxi in the road speed subset. If there is an i-th instantaneous speed sampling point v in the speed list, i is 0 and there are z-1 consecutive 0s after this sampling point, and v i+z ≠0, then this string of sampling points may be caused by abnormal taxi GPS data upload, and it is necessary to calculate the speed mutation threshold v in the speed list. GPS_e For subsequent logical judgment. Since the taxi GPS data upload abnormality does not last long, often only a few sampling points apart, the vehicle speed mutation threshold v GPS_eThe calculation logic of is to use the relative stability of the vehicle speed in a short period of time during driving, and then use the speed before the 0 value to evaluate the rationality of the non-0 value speed that appears after the 0 value data string. Among them, when the first speed value added to the speed list is 0, that is, v1 = 0, the speed threshold is the instantaneous speed of the same taxi collected at the previous moment of the instantaneous speed sampling point v1; and when v i =0 and i>=2, the speed threshold is the average of the instantaneous speed sampling points before the first speed value is 0. The specific calculation formula is:

[0042]

[0043] Where: R t-1 It represents the instantaneous speed of the same taxi collected at the moment before the instantaneous speed sampling point v1, z≥1.

[0044] S22. Since the abnormally large value is generally the first non-zero value after a series of 0 values, the speed value can be used to determine the first non-zero value v after this series of 0 values. i+z Is it a mutation value caused by GPS upload anomaly, that is, judging v i+z ≥v GPS_e Is it true? If true, then v i 、v i+z And the instantaneous vehicle speed sampling points between the two are determined to be abnormal instantaneous vehicle speed samples caused by abnormal uploading of the vehicle-mounted terminal device.

[0045] For example, as shown in the speed list in Table 1 below, the third instantaneous vehicle speed sampling point has a value of 0, and there is another 0 value in the subsequent instantaneous vehicle speed sampling point, while the fifth instantaneous vehicle speed sampling point has a non-zero value. Therefore, the calculation can be performed according to formula (1):

[0046] v GPS_e =(32.52+31.29) / 2*3=95.715<127.37, so the speed of 127.37 can be determined as a speed mutation value, and the corresponding data of the 3rd to 5th sampling points need to be eliminated.

[0047] Table 1. Sample table of some GPS uploaded abnormal data

[0048] Vehicle number time Vehicle speed (km / h) 10110 2015-08-28 09:55:28 32.52 10111 2015-08-28 09:55:58 31.29 10112 2015-08-28 09:56:28 0 10113 2015-08-28 09:56:58 0 10114 2015-08-28 09:57:28 127.37 10115 2015-08-28 09:57:58 33.58

[0049] After processing each speed list of each road section vehicle speed subset in the historical driving data set, the road section vehicle speed subset without sample data generated by GPS upload anomalies can be obtained.

[0050] S3: For each speed subset of the road section processed by S2, count the number of times the passenger status changes from empty (TP STATUS=0)To customers (TP STATUS =1) The number of taxi vehicles transformed. In this embodiment, the road segment set R A ={R1,R2,......,R n}, which corresponds to the number of taxis with status changes obtained by statistics They are Based on the above statistical result M, if the number of taxis with changing passenger status in a certain section of road If the number of taxis is too small, the sample will not be representative. Therefore, a threshold can be set to set the number of taxis with passenger status changes. The speed subset of the road sections that is not higher than the set quantity threshold is eliminated and does not participate in the subsequent statistical process. The quantity threshold can be selected according to the actual situation. In this embodiment, the quantity threshold can be set to 0, that is, only the road sections where the taxis have no changes in the passenger status are eliminated, and the rest of the road sections are retained.

[0051] After the above initial elimination process is completed, a subset of road speeds with vehicles with passenger status changes is retained. These road speed subsets contain taxi samples with passenger status changes, and these samples can be used to find the speed characteristics of taxis in the passenger-seeking state. However, it should be noted that taxis will also slow down when approaching traffic lights during driving. If these speed samples are also included in the subsequent statistical scope, it will affect the statistical results, so they need to be further eliminated through S4.

[0052] S4, for each speed subset of each road section retained after processing by S3, calculate the distance between the GPS positioning coordinates corresponding to each instantaneous speed sampling point and the nearest traffic light in front of the vehicle in the road section, and eliminate all abnormal instantaneous speed samples located in the deceleration interval before the traffic light, that is, eliminate all abnormal instantaneous speed samples with a spacing less than a distance threshold. The distance threshold represents the range of the deceleration interval before the traffic light, and the specific value can be adjusted according to the actual situation. In this embodiment, the distance threshold is set to 50m, that is, considering the influence of the traffic lights in the intersection area and other conditions, the area within 50m of the vehicle from the center point of the traffic light intersection in front is regarded as the deceleration interval before the traffic light, and the instantaneous speed sampling point data whose GPS positioning coordinates are located in the deceleration interval before the traffic light in the speed list are all eliminated. It should be noted that the area within 50m here needs to be combined with the driving direction of the vehicle on the road section. During the driving process of the vehicle, the distance from the next traffic light in the driving direction is less than 50m and it is regarded as entering the deceleration interval before the traffic light.

[0053] S5. For each speed subset of each road section processed by S4, calculate the number of vehicles with empty passenger status (TP STATUS=0) the average value of the instantaneous vehicle speed sampling points v avg1 , and calculate all passenger states as passenger (TP STATUS =1) the average value of the instantaneous vehicle speed sampling points v avg0 , and then calculate the ratio sp between the two means:

[0054]

[0055] The larger the value of sp is, the more likely it is that taxis looking for customers will appear on the road section. Therefore, all the speed subsets of the road sections can be sorted according to the ratio and the top k ones can be selected as the candidate speed subsets of the road sections. The specific value of k can be adjusted according to actual needs. Generally, a certain percentage of the top speed subsets of the road sections can be selected. All the candidate speed subsets of the road sections can be merged into a set V0.

[0056] S6. Cluster all the instantaneous speed sampling points in the subset of vehicle speeds of all candidate sections, i.e., the set V0, where the vehicle is empty and has a passenger status. The clustering in this embodiment adopts the k-means clustering algorithm. Generally speaking, the speed sample set V0 contains two parts, one part is the data sample of normal driving, whose speed value is higher, and the other part is the data sample of driving in the state of seeking passengers, whose speed value is lower. Therefore, the k-means clustering algorithm divides all the instantaneous speed sampling points in the set V0 into two categories, i.e., the number of clusters k=2 is set, and the Euclidean distance is used for calculation. The two cluster centers obtained by clustering represent the above-mentioned normal driving speed and the driving speed in the state of seeking passengers. The cluster center of normal driving and the cluster center of driving in the state of seeking passengers are represented by v1 and v2 respectively, then v1>v2. After the weighted summation of the two cluster centers, a speed threshold between v1 and v2 can be obtained. The speed threshold represents the speed that a vehicle in the state of seeking passengers should have, and can be used to identify vehicles in the state of seeking passengers.

[0057] Since the position of the cluster center initially selected during the clustering process will affect the final cluster center position to a certain extent, in this embodiment, it is necessary to repeat the clustering multiple times with the number of clusters being 2 but the initial cluster center being random. The weighted sum of the two cluster centers obtained each time is taken as a speed threshold. and Respectively represent the two cluster centers obtained after the i-th clustering, where and They are and The number of samples in the corresponding cluster, then the speed threshold calculated by weighted summation after the i-th clustering is:

[0058]

[0059] Finally, the average of the speed thresholds obtained by multiple clustering can be taken as the final speed threshold v fp , and then conduct subsequent passenger-seeking vehicle identification. The basic logic of passenger-seeking vehicle identification is: based on the vehicle's real-time positioning GPS data, calculate the vehicle's real-time speed sample, and obtain sample data within a period of time from it. If the vehicle is unloaded (TP STATUS =0), and the vehicle speed is continuously less than v for a long time fp , then it can be determined that the unloaded vehicle is in the passenger-seeking state. However, considering that there may be road congestion and other conditions in real scenes, which may cause the vehicle to be in a state of long-term low-speed driving even if it is not in the passenger-seeking state, the present invention also needs to add a judgment condition for judging whether the long-term low-speed driving of the vehicle is a collective behavior or an individual behavior. The specific judgment and identification method is shown in S7.

[0060] S7: For each taxi running on each road section in the target area, compare the instantaneous speed of each taxi within the set time interval with the speed threshold v. fp For comparison, if a taxi outside the deceleration zone before the traffic light is empty and its instantaneous speed is continuously less than the speed threshold v fp , and at the same time, there is no other taxi with passengers whose instantaneous speed is continuously less than the speed threshold v on the same road section and in the same time period. fp , the taxi is identified as a passenger-seeking vehicle.

[0061] It should be noted that, in the present invention, “the instantaneous speed is continuously less than the speed threshold v fp The judgment condition of "" needs to ensure that the vehicle is outside the deceleration zone before the traffic light to avoid misjudgment caused by the deceleration of the vehicle at the traffic light. The definition of the deceleration zone before the traffic light is as described above. In this embodiment, the area within 50m from the center point of the traffic light intersection ahead is taken as the deceleration zone before the traffic light.

[0062] In addition, the judgment condition "the instantaneous speed is continuously less than the speed threshold v fp This can be assisted by presetting a continuous time interval during which all instantaneous speeds are less than the speed threshold v fp , can be used as a feature of a passenger-seeking vehicle to avoid misjudgment caused by a short-term deceleration of the vehicle. The duration time interval can be set as needed, generally between 10 seconds and 30 seconds.

[0063] In addition, in the present invention, it is necessary to further screen other taxis with passengers in the same section and in the same time period, in order to prevent misjudgment due to road congestion. Because when the road is congested, regardless of whether the taxi is looking for passengers, its speed is very low, especially the taxi with passengers. Therefore, although the taxi outside the deceleration zone before the traffic light is empty and the instantaneous speed is continuously less than the speed threshold v fp , but if at this time there is a taxi with passengers on the same road section whose instantaneous speed is also continuously less than the speed threshold v fp , then it is likely that the vehicles are collectively slowing down due to traffic congestion. Figure 3 The figure shows the speed of some actual taxis in the two sections. The speeds of the vehicles in (a) are very low, so they are probably group slow-speed behaviors caused by traffic congestion and are not considered as passenger-seeking vehicles. In (b), only the instantaneous speed of vehicle No. 69695 is continuously less than the speed threshold v. fp , so it is considered a passenger-seeking vehicle.

[0064] The present invention can identify vehicles in the passenger-seeking state from a large amount of taxi data through the above-mentioned passenger-seeking vehicle identification method, and remove its speed samples from the data, so that the remaining speed samples can reflect the actual speed of the vehicle on the road. This embodiment selects two consecutive days of vehicle historical GPS data for validity verification, and also performs operations such as road section selection and intersection area data removal. After that, the passenger-seeking vehicle identification method is used to identify the passenger-seeking vehicle and remove its corresponding data. Table 2 shows the comparison results of the average road speed of the vehicle before and after removing the passenger-seeking vehicle data. It can be concluded that the vehicle speed after processing is closer to the actual speed of the road section, which improves the quality of the speed sample to a certain extent.

[0065] Table 2: Results of the passenger-seeking vehicle before and after processing

[0066] Sample data status Sample size Processing sample size (percentage) Average vehicle speed (km / h) Before data processing 12531032 0 22.63 After data processing 11765386 6.11 25.58

[0067] The above-described embodiment is only a preferred solution of the present invention, but it is not intended to limit the present invention. A person skilled in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A method for identifying vehicles in a passenger-seeking state on a road section based on taxi GPS data, characterized in that: Here are the steps: S1. Obtain a historical driving data set of taxis operating in a target area, wherein the historical driving data set is composed of road speed subsets corresponding to different road sections in the target area, and the road speed subset of each road section contains a speed list of different taxis during driving on the road section, wherein the speed list is composed of instantaneous speed sampling points regularly uploaded by a vehicle-mounted terminal device of the vehicle, and each instantaneous speed sampling point is synchronously recorded by the vehicle-mounted terminal device with a corresponding vehicle number, as well as the GPS positioning coordinates and passenger status when recording the instantaneous speed sampling point; S2, pre-processing each road speed subset in the historical driving data set to remove abnormal instantaneous speed samples caused by abnormal upload of the vehicle terminal device; S3, for each speed subset of the road section processed by S2, the number of taxis whose passenger-carrying status changes from empty to loaded during driving on the road section is counted, and the speed subsets of the road section whose number is not higher than the set number threshold are eliminated; S4, for each speed subset of each road section retained after processing in S3, calculate the distance between the GPS positioning coordinates corresponding to each instantaneous speed sampling point and the nearest traffic light in front of the vehicle in the road section, and eliminate all abnormal instantaneous speed samples in the deceleration zone before the traffic light; S5. For each speed subset of each road section processed by S4, calculate the ratio between the average value of the instantaneous speed sampling points of all the vehicles with no passengers and the average value of the instantaneous speed sampling points of all the vehicles with passengers, sort all the speed subsets of the road section according to the ratio, and select the top-ranked ones as candidate speed subsets of the road section; S6, clustering all instantaneous vehicle speed sampling points with empty vehicle status in all candidate road speed subsets into 2 clusters, and taking the weighted sum of the two cluster centers obtained by clustering as the speed threshold; S7. For the taxis running on each road section in the target area, the instantaneous speed of each taxi within the set time interval is compared with the speed threshold. If a taxi outside the deceleration zone before the traffic light is empty and its instantaneous speed is continuously less than the speed threshold, and there is no other taxi with passengers on the same road section and in the same time period whose instantaneous speed is continuously less than the speed threshold, then the taxi is identified as a passenger-seeking vehicle. In S2, the method for identifying the abnormal instantaneous vehicle speed samples caused by the abnormal upload of the vehicle terminal device in the road section speed subset is as follows: S21. Traverse the speed list corresponding to each taxi in the road speed subset. If there is an i-th instantaneous speed sampling point v in the speed list, i is 0 and there are z-1 consecutive 0s after this sampling point, and v i+z ≠0, then calculate the speed mutation threshold v in the speed list GPS_e for: Where: R t-1 represents the instantaneous speed of the same taxi collected at the moment before the instantaneous speed sampling point v1, z≥1; S22, judge v i+z ≥v GPS_e Is it true? If true, then v i 、v i+z The instantaneous vehicle speed sampling points between the two are all determined to be abnormal instantaneous vehicle speed samples caused by abnormal upload of the vehicle terminal device; The calculation formula of the speed threshold is: Where: represents the speed threshold obtained by weighted summation after the i-th clustering, and Respectively represent the two cluster centers obtained after the i-th clustering, where and They are and The number of samples in the corresponding cluster; In S6, clustering is repeated multiple times with the number of clusters being 2 but the initial cluster centers being random. The weighted sum of the two cluster centers obtained in each clustering is taken as a speed threshold, and the average of the speed thresholds obtained from multiple clusterings is taken as the final speed threshold.

2. The method for identifying vehicles in a passenger-seeking state in a road section based on taxi GPS data as claimed in claim 1, characterized in that: The deceleration zone before the traffic light is the road section within 30 to 50 meters before the traffic light.

3. The method for identifying vehicles in a passenger-seeking state in a road section based on taxi GPS data as claimed in claim 1, characterized in that: The clustering adopts k-means clustering algorithm.

4. The method for identifying vehicles in a passenger-seeking state in a road section based on taxi GPS data as claimed in claim 1, characterized in that: The quantity threshold is 0.

5. The method for identifying vehicles in a passenger-seeking state on a road section based on taxi GPS data as claimed in claim 1, characterized in that: The historical driving data set is obtained by fusing the recorded data of the vehicle-mounted terminal device installed on the taxi and the road network map data.

Images