CN114639240B - Vehicle effective travel time extraction method based on first-in first-out rule anomaly detection - Google Patents

Abstract

The invention discloses a method for extracting effective travel time of a vehicle based on first-in first-out rule anomaly detection, which comprises the following steps: 1. obtaining number plate data detected by a bayonet electronic police at a road network intersection, and matching upstream and downstream number plates; 2. screening out the number plate data of the complete passing road section; 3. extracting a time stamp sequence when the vehicle passes through an upstream intersection according to the sequence of the traffic flow reaching a downstream intersection; 4. screening abnormal values in the timestamp sequence by using a vehicle first-in first-out rule, and regenerating the timestamp sequence which accords with the first-in first-out rule; 5. and matching the number plate data with the newly generated time stamp sequence, and calculating the travel time value of the vehicle in the sequence to obtain the travel time of the effective road section of the vehicle. The invention improves the scientificity and accuracy of vehicle travel time pretreatment.

Description

Vehicle effective travel time extraction method based on first-in first-out rule anomaly detection

Technical Field

The invention belongs to the technical field of intelligent traffic.

Background

The reliable travel time is an important index for reflecting the running state of the urban traffic, and can be used for estimating intersection delay and queuing length, guiding signal control optimization, revealing discontinuous flow running rules, identifying urban road congestion areas and providing data support for traffic management decision schemes and effect evaluation. In recent years, traffic information acquisition technologies and equipment coverage rates are rapidly developed, and travel time acquisition channels are greatly enriched by using travel information acquisition technologies represented by floating car technologies, bluetooth technologies, automatic license plate recognition technologies and automatic vehicle recognition technologies. The automatic license plate recognition technology can be used for recording the driving information of the vehicle on different intersection sections, and the license plate data becomes a common data source for acquiring the observed value of the high-quality road section travel time by virtue of the continuity and the advantages of large sample recording. However, as the vehicles may have driving behaviors such as temporary stopping, detouring and overspeed in the road section which cannot reflect the traffic flow operation characteristics, and the entrance and exit in the road section can significantly influence the vehicle operation speed, the abnormal values which cannot reflect the operation characteristics of most vehicles in the traffic flow in the road section exist in the detected vehicle travel time, and the accurate travel time validity preprocessing from the number plate data has very important research significance.

The existing vehicle travel time preprocessing method can be roughly divided into the following steps: (1) An absolute threshold method, which sets the upper and lower limit values of travel time according to road segment physical or traffic control limits; (2) The quantile value method is to keep the travel time value within a certain distance range from the center of the sample in a specific statistical time window. The subjectivity of the upper and lower limit values of the absolute threshold value method is strong, and the effect is good when the overtime or undersize obvious travel time errors are eliminated; the quantile value method can adapt to the difference of travel time of different traffic flows, but the whole statistical distribution of the travel time is not easy to capture only by means of proportionally rejecting data outside intervals. The essence of the two methods is that the threshold is set to remove outlier data outside the interval, and although the effect of eliminating the travel time under abnormal driving behaviors is achieved in terms of results, the effective travel time of partial outliers may be lost, and the traffic flow operation characteristics cannot be completely reflected. In addition, the method for setting the threshold value depends on certain traffic control experience, and the extraction result is lack of scientificity and credibility.

Disclosure of Invention

The purpose of the invention is as follows: in order to solve the problems in the prior art, the invention provides a method for extracting the effective travel time of the vehicle by using an FIFO rule.

The technical scheme is as follows: the method for extracting the effective travel time of the vehicle based on the first-in first-out rule abnormity detection specifically comprises the following steps:

step 1: obtaining number plate data of road network intersections, wherein the number plate data comprises: the method comprises the following steps of (1) intersection number, lane number, date, license plate number, intersection detection time and road section number; adding upstream and downstream intersection numbers in the number plate data according to the data acquisition equipment and the intersection comparison table; matching license plate numbers passing through road network intersections with license plate data, wherein each license plate number is matched with one piece of license plate data to obtain a data set m ₁ ；

Step 2: for data set m ₁ In any lane, in the data set m ₁ Screening out the number plate numbers which completely pass through the upstream and downstream crossings of the lane and the corresponding number plate data,obtain a data set m ₂ ；

And step 3: m is to be ₂ The data in the sequence are sorted in ascending order according to the detection time of the downstream intersection; m after sorting ₂ Extracting the license plate number corresponding to the upstream intersection detection time and the upstream intersection detection time to obtain an upstream timestamp sequence l ₁ ；

And 4, step 4: based on the sequence l ₁ Generating the longest ascending sequence l with sequentially increasing upstream time stamps according with FIFO rule by using the longest ascending subsequence algorithm ₂ Is prepared by ₁ Taking the residual upstream time stamps and the number plate data corresponding to the residual upstream time stamps as abnormal data;

and 5: based on the longest increasing sequence l ₂ At m is ₂ And extracting the number plate data which accords with the FIFO rule, calculating the travel time of the vehicle, and taking the travel time as the effective travel time.

Further, the license plate numbers completely passing through the upstream and downstream intersections of a certain lane in the step 2 meet the following conditions:

the detection time of the vehicle at the upstream and downstream intersections of the lane is not empty, and the detection time interval delta t of the upstream and downstream intersections is less than or equal to a preset time threshold.

Further, the step 3 specifically includes:

s31 data set m ₂ The number plate data in (1) is reordered according to the detection time of the downstream intersection: data set m ₂ The detection time of each upstream and downstream intersection is converted into the cumulative seconds from 0 th day to the corresponding detection time, and m is calculated according to the cumulative seconds of the downstream intersections ₂ The data in the sequence are sorted according to the ascending order of the accumulated seconds of the downstream intersection;

s32, extracting the accumulated seconds of the detection time of the upstream intersection and the corresponding license plate number from the data set after the ascending sequencing to obtain an upstream detection timestamp sequence l ₁ (ii) a And taking the accumulated seconds corresponding to the detection time of the upstream intersection as a passing time stamp.

Further, the step 4 specifically includes:

s41: with the sequence l ₁ As an initial transit time stamp of the incremental sequence;

s42: if the current vehicle passing timestamp j is smaller than the maximum vehicle passing timestamp in the incremental sequence, replacing the maximum vehicle passing timestamp in the incremental sequence with the current vehicle passing timestamp;

otherwise, adding the current passing timestamp into the incremental sequence, and updating the number of the passing timestamps in the incremental sequence at the same time;

s43: sequence l ₁ After traversing is finished, obtaining an increasing sequence;

s44: based on the sequence l ₁ Finding out the license plate number corresponding to each passing time stamp in the increasing sequence, and taking each passing time stamp and the corresponding license plate number in the increasing sequence as the longest increasing sequence l ₂ 。

Further, the step 5 specifically includes:

s51: at m ₂ In which only l is retained ₂ The number plate data corresponding to the number plate exists; obtain the number plate data set m ₃ ；

S52, mixing m ₃ And taking the travel time of any one vehicle in the section between the corresponding upstream and downstream intersections as the effective travel time of the vehicle according with the FIFO rule.

Has the advantages that: the invention takes the data of the number plate of the vehicle leaving the lane level as input, screens out the effective travel time which can reflect the travel rule of the traffic flow road section by utilizing the FIFO rule, and carries out the effective pretreatment of the travel time by a method which does not need to set a threshold value. The scheme does not need subjective judgment of components and road section physical prior information, and reduces the workload of actual travel time pretreatment. In addition, the invention extracts the effective travel time by a data driving method, can reserve a certain sample of the outlier effective travel time, and obtains the travel time which is more in line with the traffic operation characteristics of the road section.

Drawings

FIG. 1 is a flow chart of the present invention.

FIG. 2 is a schematic diagram of the run time of the mismatch FIFO rule.

FIG. 3 is a graph comparing probability distributions of forward and backward travel times using the method of the present invention; wherein plot (a) is a probability distribution plot of travel time without the use of the present invention; fig. (b) is a travel time probability distribution chart according to the present invention.

Detailed Description

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an embodiment of the invention and, together with the description, serve to explain the invention and not to limit the invention.

The embodiment shown in fig. 1 provides a method for extracting a valid travel time of a vehicle based on an abnormality detection of a fifo rule, comprising the following steps:

step 1: obtaining number plate data detected by a bayonet electronic police at a road network intersection, matching the number plate data of intersections of upstream and downstream roads according to the number plate number according to a bayonet electronic police device and an intersection comparison table to obtain a number plate data set m ₁ ；

S11, road network intersection number plate data are obtained, the number plate data fields mainly used comprise intersection numbers, entrance lane numbers, dates, detection time, license plate numbers and road section numbers, and the road section number fields are used for adding upstream and downstream intersection number fields.

S12, matching the same license plate number with upstream and downstream intersections, adding an upstream vehicle detection time field in the intersection number data to obtain a license plate data set m containing the license plate number, the downstream intersection detection time and the upstream intersection detection time field ₁ . Number plate data set m ₁ As shown in table 1:

TABLE 1

Step 2: according to the number plate data set m ₁ Screening out the vehicle number plate data of the complete passing road section to obtain a number plate data set m ₂ (ii) a And selecting the license plate number corresponding to the license plate data meeting the following condition A.

Condition a: the detection time of the upstream and downstream intersections is not empty, and the detection time interval delta t of the upstream and downstream intersections is less thanEqual to a preset time threshold. In this embodiment, the time threshold is 5 minutes, and the number plate data set m ₂ As shown in table 2:

TABLE 2

And step 3: number plate data set m ₂ Extracting detection time information, and generating an upstream detection time stamp sequence l by taking a downstream detection time as a reference ₁ The method specifically comprises the following steps: number plate data set m ₂ Reordering by downstream detection time, converting the upstream and downstream detection times into cumulative seconds from the time of day 0 to the corresponding detection time (for example, the upstream detection timestamp of the data snenn 7 × 8 is 74737153, which represents the time of 7, and is converted into the number of seconds passed from 0) ₁ The serial number label of (1); according to the accumulated seconds of the downstream intersection, m is divided ₂ The data in (1) is sorted in ascending order of cumulative seconds of downstream intersection (and also in order of arrival of vehicles).

Sorted number plate dataset m ₂ In the middle, the up stream time accumulated second number field and the license plate number field are extracted from the lane and recorded as an up stream detection time stamp sequence l ₁ . Upstream detection timestamp sequence l ₁ Shown in Table 3:

TABLE 3

Time stamp/second	License plate number
		27999.676	Threen 7 x 8
28001.838	Threo E88A W
		28002.478	Threo E5L1 x 5
28006.641	Threo ES1Z 9
		28000.717	Threo E55J 5
28008.082	Threef 36 × 5
		28008.482	Threo E5H1 x 8
28013.926	Threo E766E
		27996.473	Threo EMH20
28015.927	Threo EN5R 9

And 4, step 4: according to the sequence l ₁ Generating the sequence l using a longest ascending subsequence algorithm ₁ Longest ascending sequence l of sequential increase of medium time stamps (i.e. maximum number of vehicles according to FIFO rule) ₂ And eliminating the time stamp of the abnormal vehicle.

S41: with the sequence l ₁ Is started, initializedIs the first timestamp, traverses the entire sequence l ₁ By searching through dichotomy, looking up the current sequence l ₁ The position j-1 of the ith passing time stamp in the existing increasing sequence M is marked as M (j-1) to represent the j-1 passing time stamp in the increasing sequence M, and the passing time stamp is in the sequence l ₁ Is the ith.

S42: store M (j-1) in sequence l ₁ Position P (i), M (j) is updated by the following conditions:

(1) Timestamp if the current time of the passing (also sequence l) ₁ The (i + 1) th passing timestamp) is smaller than the maximum timestamp in the existing increasing sequence, the maximum timestamp in the increasing sequence is replaced by the current timestamp, and the current timestamp is dynamically stored in the sequence l ₁ And l ₂ The position of (1);

(2) If the current passing timestamp is greater than or equal to the maximum timestamp in the existing increasing sequence, taking the current timestamp as the last bit of the increasing sequence, and updating the number of the increasing sequence to be the number of the increasing sequence; and dynamically saving the current time stamp in the sequence l ₁ And l ₂ The position in (1);

s43: after traversing is finished, starting from a time stamp corresponding to M (L-1) in the longest incremental sequence, and sequentially converting the sequence L according to the relation of P (i) = M (j-1) ₁ P (M (L-1)), P (M (L-1)))) \8230'; time stamp concatenation of locations (also finding each time stamp corresponding to sequence L starting from the last time stamp of the sequence once for each time stamp ₁ Time stamp in (c) and the generated complete time stamp sequence is inverted back and forth with respect to sequence l ₁ After the license plate number corresponding to the timestamp sequence is obtained through matching, the license plate number is recorded as a longest increasing sequence l ₂ As shown in table 4:

TABLE 4

Time stamp/second	License plate number
		27999.676	Threen 7 x 8
28001.838	Threo E88A W
		28002.478	Threo E5L1 x 5
28006.641	Threo ES1Z 9
		28008.082	Threef 36 × 5
28008.482	Threo E5H1 x 8
		28013.926	Threo E766E
28015.927	Threo EN5R 9

And 5: according to the sequence l ₂ Number plate data set m ₂ Matching and generating a number plate data set m conforming to the FIFO rule ₃ Number plate data set m ₃ The method comprises the following steps of calculating and extracting effective travel time information of the vehicle:

s51: number plate data set m ₂ With the sequence l ₂ Matching is done with upstream time stamps, only sequence l is retained ₂ The corresponding number plate data in the license plate data set generating a number plate data set m ₃ ，m ₃ As shown in table 5:

TABLE 5

S52: according to the number plate data set m ₃ And calculating the time interval delta t of each vehicle passing through the vehicle at the upstream and downstream intersections as effective travel time, namely the effective road section travel time according with FIFO (first-in first-out) rule.

The travel times that do not comply with the FIFO rule are shown in fig. 2 by dashed lines. (plate data for plate numbers threo E55J 5 and threo mh20 were excluded).

In the embodiment, before and after the license plate data of a certain road section at the early peak time is applied with the method, the histogram pair of the vehicle travel time distribution is shown in fig. 3, wherein (a) in fig. 3 is the method without applying the method, and (b) is the method adopting the method; from fig. 3, it can be seen that after the FIFO rule-based anomaly detection and validity extraction are performed, the peak of the travel time distribution is significant, the skewness is reduced, and the anomaly value with the travel time greater than 150 seconds is greatly reduced.

It should be noted that the various features described in the above embodiments may be combined in any suitable manner without departing from the scope of the invention. The invention is not described in detail in order to avoid unnecessary repetition.

Claims (3)

1. The method for extracting the effective travel time of the vehicle based on the first-in first-out rule anomaly detection is characterized by comprising the following steps:

step 1: obtaining number plate data of road network intersections, wherein the number plate data comprises: the method comprises the following steps of (1) intersection number, lane number, date, license plate number, intersection detection time and road section number; adding upstream and downstream intersection numbers in the number plate data according to the data acquisition equipment and the intersection comparison table; matching the number plate number of the road network intersection with the number plate data, each number plate numberMatching codes with a number plate data to obtain a data set m ₁ ；

Step 2: for data set m ₁ In any lane, in the data set m ₁ Screening out the license plate numbers which completely pass through the upstream and downstream intersections of the lane and the corresponding license plate data to obtain a data set m ₂ ；

And step 3: m is to be ₂ The data in the sequence are sorted in ascending order according to the detection time of the downstream intersection; m after sorting ₂ Extracting the license plate number corresponding to the upstream intersection detection time and the upstream intersection detection time to obtain an upstream timestamp sequence l ₁ ；

And 4, step 4: based on the sequence l ₁ Generating the longest ascending sequence l with sequentially increasing upstream time stamps according with FIFO rule by using the longest ascending subsequence algorithm ₂ Is prepared by ₁ Taking the residual upstream time stamps and the number plate data corresponding to the residual upstream time stamps as abnormal data;

and 5: based on the longest increasing sequence l ₂ At m ₂ Extracting the number plate data meeting FIFO rules, calculating the travel time of the vehicle, and taking the travel time as effective travel time;

the step 3 specifically comprises the following steps:

s31 data set m ₂ The number plate data in (1) is reordered according to the detection time of the downstream intersection: data set m ₂ The detection time of each upstream and downstream intersection is converted into the cumulative seconds from 0 th day to the corresponding detection time, and m is calculated according to the cumulative seconds of the downstream intersections ₂ The data in the sequence are sorted according to the ascending order of the accumulated seconds of the downstream intersection;

s32, extracting the accumulated seconds of the detection time of the upstream intersection and the corresponding license plate number from the data set after the ascending sequencing to obtain an upstream detection timestamp sequence l ₁ (ii) a Taking the accumulated seconds corresponding to the detection time of the upstream intersection as a passing time stamp;

the step 4 specifically comprises the following steps:

s41: with the sequence l ₁ As an initial transit time stamp of the incremental sequence;

s42: if the current vehicle passing timestamp j is smaller than the maximum vehicle passing timestamp in the incremental sequence, replacing the maximum vehicle passing timestamp in the incremental sequence with the current vehicle passing timestamp;

otherwise, adding the current passing timestamp into the incremental sequence, and updating the number of the passing timestamps in the incremental sequence at the same time;

s43: sequence l ₁ After traversing is finished, obtaining an increasing sequence;

s44: based on the sequence l ₁ Finding out the license plate number corresponding to each passing time stamp in the increasing sequence, and taking each passing time stamp and the corresponding license plate number in the increasing sequence as the longest increasing sequence l ₂ 。

2. The method for extracting vehicle active travel time based on fifo rule anomaly detection according to claim 1, wherein the license plate number that completely passes through the upstream and downstream intersections of a lane in step 2 satisfies the following conditions:

the detection time of the vehicle at the upstream and downstream intersections of the lane is not empty, and the detection time interval delta t of the upstream and downstream intersections is less than or equal to a preset time threshold.

3. The method as claimed in claim 1, wherein the step 5 is specifically as follows:

s51: at m ₂ In which only l is retained ₂ The number plate data corresponding to the number plate exists; obtain the number plate data set m ₃ ；

S52, mixing m ₃ And taking the travel time of any one vehicle in the section between the corresponding upstream and downstream intersections as the effective travel time of the vehicle according with the FIFO rule.

Images