CN115099308A - Vehicle next position prediction method based on segmented track clustering - Google Patents
Vehicle next position prediction method based on segmented track clustering Download PDFInfo
- Publication number
- CN115099308A CN115099308A CN202210598479.8A CN202210598479A CN115099308A CN 115099308 A CN115099308 A CN 115099308A CN 202210598479 A CN202210598479 A CN 202210598479A CN 115099308 A CN115099308 A CN 115099308A
- Authority
- CN
- China
- Prior art keywords
- center point
- cluster
- cluster center
- track
- sub
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000011218 segmentation Effects 0.000 claims abstract description 14
- 238000006243 chemical reaction Methods 0.000 claims description 23
- 230000007704 transition Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 abstract 1
- 230000008901 benefit Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides a vehicle next position prediction method based on segmented track clustering, which comprises the following steps: s1: trajectory segmentation, given vehicle v a Original track sequenceWherein the content of the first and second substances,indicating a vehicle v a At t b Extracting a key position, namely a track inflection point, in the track at the longitude and latitude coordinate position of the moment so as to obtain a piecewise linear sub-track, wherein B is the total number of the sequence; s2: sub-track clustering, namely, clustering sub-tracks with similar characteristics, namely, clustering sub-tracks with distances and angles smaller than preset values on the same road or the same direction, and acquiring the position and time information of the center point of each cluster; s3: based on the segmented clustering result of the historical track data, moving time difference mode is carried out according to cluster center pointAnd a type which performs next cluster center point prediction, that is, next position prediction, using the shortest moving time from the passing cluster center point to the candidate next cluster center point and its actual moving time difference.
Description
Technical Field
The invention relates to the technical field of intelligent transportation, in particular to a vehicle next position prediction method based on segmented track clustering.
Background
With the continuous development of Internet of Vehicles (IoV) technology and the wide application of mobile GPS devices and wireless sensors, people increasingly benefit from various Location Based Services (LBS), such as entertainment Services, advertisement delivery, parking space information, and traffic congestion conditions. The core of these services is to collect location information, and most basically, to obtain the next location of the current trip, thereby providing better service to the user.
Vehicle next position prediction is usually based on a large amount of historical trajectory data, and the most common method for next position prediction is based on a markov model, which, however, due to the complexity of time and space, predicts the next position using only the position just passed through, but ignores the position passed earlier in the trajectory. The user often selects a route using the shortest time from the departure point to the destination, and predicts the next location by comparing the difference between the shortest time from the passed location to the next location and its actual time. On the other hand, the historical track data has a sparsity problem, which means that the track of the vehicle is not completely the same due to various artificial factors or factors such as actual weather and roads, and the coverage of the existing historical track data cannot meet the requirement of inquiring the track, so that the accuracy of prediction is influenced.
Disclosure of Invention
The invention aims to provide a vehicle next position prediction method based on segmented track clustering, which can consider the earlier passing position in a track, and simultaneously solve the problem of sparsity of historical track data by comparing the shortest time from the passing position to the next position with the actual time difference, thereby improving the accuracy of next position prediction.
The technical scheme is as follows: the invention provides a vehicle next position prediction method based on segmented track clustering, which comprises the following steps:
s1: trajectory segmentation, given vehicle v a Original track sequenceWherein,indicating a vehicle v a At t b Extracting a key position, namely a track inflection point, in the track at the longitude and latitude coordinate position of the moment so as to obtain a piecewise linear sub-track, wherein B is the total number of the sequence;
s2: sub-track clustering, namely, clustering sub-tracks with similar characteristics, namely, clustering sub-tracks with distances and angles smaller than preset values on the same road or the same direction, and acquiring the position and time information of the center point of each cluster;
s3: based on the segmented clustering result of the historical track data, according to a cluster center point moving time difference model, the next cluster center point prediction, namely the next position prediction, is carried out by using the shortest moving time from the passing cluster center point to the candidate next cluster center point and the actual moving time difference.
Further, the neutron trajectory clustering method in S2 is as follows:
s21: initializing a sub-track cluster set C, and enabling the sub-track set S to be { S } 1 ,s 2 ,…,s k ,…,s K In which s is k Representing the kth sub-track segmented based on S1, dividing the first term S 1 Set as a new cluster c 1 And adds it to cluster C, when C ═ C 1 }={(s 1 ) And K is the total number of the sub-tracks;
s22: traversing other sub-tracks S in the set S in sequence 2 ,…s k ,…,s K Comparing the similarity of the sub-tracks and the existing clusters, when K is x, x is more than or equal to 2 and less than or equal to K, obtaining the sub-tracks which are line segments consisting of uniformly filled points based on a track division DP algorithm, and sequentially calculating the current sub-tracks s x With the existing clusterc i Average closest point pair distance ANPPD of, denoted asIt refers to sub-track s x From each point to a cluster c i Average distance of nearest points, wherein sub-track s x Should be less than or equal to the length of the cluster c i The length of the center track, denoted as | s x |≤|c i L, |; calculating to obtain the current sub-track s x With the existing cluster c i Angle therebetween; i is more than or equal to 1 and less than or equal to | C |, wherein | C | is the number of elements of the sub-track clustering set C;
s23: based on the method, the current sub-track s is found x With the existing cluster c i Smallest between
S24: if it is smallestIf the Angle is smaller than the set ANPPD threshold and the corresponding Angle is smaller than the set Angle threshold, the sub-track s is determined x Added to the current cluster c i The preparation method comprises the following steps of (1) performing; and recalculates the current cluster c i Center, specifically, determining the vector direction of the center of the cluster, assuming cluster c i ={s m ,...,s n 1 ≦ m ≦ n ≦ K, then each sub-track is represented as a vector, respectivelyThe vector direction of the cluster center is derived from the following equation:
determining a center vectorStarting from the current cluster c i The centers of the starting points of all the sub-tracks and the end point are the current cluster c i The centers of all the sub-track end points are uniformly filled with central lines between the start points and the end points, the number of filling points is the average number of points of each sub-track in the cluster, and the current cluster c is obtained i The center midpoint, i.e., the cluster center point CC i Coordinate position and time information are included;
s25: if it is smallestGreater than the set ANPPD threshold, the current sub-trajectory s should be set x Created as a new cluster and added to cluster set C.
Further, the specific method of step 3 is as follows:
s31: constructing a weighted cluster center point transfer graph by using the average moving time of each cluster center point conversion; considering the road network as a graph, the cluster center point CC i Corresponding to nodes of the graph, the conversion between the center points of each cluster is regarded as an edge of the graph, the weight of the edge is the average moving time of the conversion of the center points of the clusters, and the average moving time of the conversion of the center points of the clusters, namely the MS of a given moving road section, is obtained through the average running time of the road section in the historical track data of the vehicle i =CC i →CC i+1 Wherein, MS i Represented as a directed segment, containing two consecutive cluster center points CC i And CC i+1 Calculating the average moving timeFurther consider that the road conditions in different time periods are different, resulting inWill vary from time slot to time slot, and the mean moving time of the cluster center point transition at the k-th time slot is expressed asAssuming that the size of the time slot is set to 1 hour, then there are 24 time slots in a day, 1 ≦ k ≦ 24, k is an integer, and 24 snapshots are provided for the weighted cluster center point transfer graph;
s32, calculating the shortest moving time between all the cluster center point pairs; given a starting cluster center point CC i Firstly, selecting a kth weighted cluster center point conversion graph according to the time period of arrival of a user, and calculating a target cluster center point CC based on a shortest path algorithm j Minimum travel time ofAnd calculating the shortest moving time between any two possible cluster center points in an off-line mode in advance according to the weighted cluster center point conversion graph, and storing the values in an effective data structure;
s33, estimating the actual moving time between the passing cluster center point and the candidate next cluster center point;
given a query trajectory sequence QT ═ P [ [ (P) 1 ,h 1 ),...,(P i ,h i ),...,(P n ,h n )]Wherein (P) i ,h i ) Position information P including longitude and latitude representing ith trace point i And time stamp information h i Since the next position prediction is performed based on the cluster set, the given query trajectory sequence needs to be converted into a query cluster center point sequence QCC ═ CC 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]Wherein (CC) i ,t i ) Location information CC including longitude and latitude representing the ith cluster center point i And corresponding time stamp information t i The conversion of the query track sequence QT to the query cluster center point sequence QCC is based on the position information coordinate P in the given query track i And the cluster center point CC acquired in S24 i Coordinate information one-to-one mapping, track point P i With cluster center point CC nearest to it i Converted one by one and time stamp information h i Also from the cluster center point CC i Attached time stamp information t i Replacing one by one; based on the QCC, the cluster center point CC passing through can be estimated and calculated i To the next cluster center point CC n+1 I is more than or equal to 1 and less than or equal to n:
wherein, t n Is the current cluster center point CC n Time stamp information of t i Is the passing cluster center point CC i Timestamp information of (2), current cluster center point CC n To the next cluster center point CC n+1 Is unknown, using the average moving timeInstead, n represents the subscript of the current cluster center point;
s34, predicting the central point of the next cluster, namely the next position, by using the difference between the actual moving time and the shortest moving time;
based on the obtained query cluster center point sequence QCC ═ CC [ (CC) ] 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]And its candidate next cluster center point CC n+1 If from the passing cluster center point CC i To the next cluster center point CC n+1 Minimum travel time ofLess than its actual movement time at a (CC i ~CC n+1 QCC), then CC is indicated n+1 Not the expected next cluster center point; in order to subsequently calculate the probability of reaching the center point of each candidate next cluster, the shortest moving time and the actual moving time from the center point of each passing cluster in the cluster center point sequence to the center point of the candidate next cluster need to be calculated and compared;
obtaining cluster center point CC from all the passing clusters based on S32 method i To the candidate next cluster center point CC n+1 The sum of the shortest moving time of (a) is:
wherein k is a k weighted cluster center point transition graph determined according to the time period;
based on S33, all passing cluster center points CC in the query cluster center point sequence QCC i To the candidate next cluster center point CC n+1 The sum of the actual movement times of (a) is:
in summary, given the query trajectory QT, the query cluster center point sequence QCC ═ CC is obtained by conversion 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]And calculating the probability of the center point of the next cluster of the access candidate as follows:
wherein, | QCC | represents the number of cluster center points in the query sequence QCC, and z (QCC) is a normalization term, and the calculation formula is:
where m represents the number of candidate next cluster center points,a function f is represented that acts on the center point of the r-th candidate next cluster, and the function f is the inverse function.
Has the advantages that: compared with the prior art, the technical scheme of the invention has the following beneficial technical effects:
the invention provides a vehicle next position prediction method based on segmented track clustering, which comprises the steps of finding out key points of a historical track and carrying out segmentation to obtain a segmented linear sub-track; then, clustering sub-tracks, and classifying similar sub-tracks by taking the direction and the distance as judgment conditions; by means of segmentation and clustering of the historical tracks, the problem of sparsity of historical track data is solved; then, based on the segmented clustering result, a cluster center point moving time difference model is provided, and next position prediction is carried out by using the shortest moving time from the passing cluster center point to the candidate next cluster center point and the actual moving time difference.
Drawings
FIG. 1 is a flow chart of a method for predicting a next vehicle position based on a segmented trajectory clustering according to the present invention;
FIG. 2 is a specific example of the ANPPD in the method S2 for predicting the next vehicle position based on the segmented trajectory clustering according to the present invention;
fig. 3 is an illustrative example of a user query cluster center point sequence in the method S3 for predicting the next vehicle position based on segmented trajectory clustering according to the present invention.
Detailed Description
The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings.
The invention provides a flow chart of a vehicle next position prediction method based on segmented track clustering, which is shown in figure 1 and comprises the following steps:
s1: trajectory segmentation, given vehicle v a Original track sequenceWherein,indicating vehicle v a At t b Extracting a key position, namely a track inflection point, in the track at the longitude and latitude coordinate position of the moment so as to obtain a piecewise linear sub-track, wherein B is the total number of the sequence;
s2: sub-track clustering, namely, clustering sub-tracks with similar characteristics, namely, clustering sub-tracks with distances and angles smaller than preset values on the same road or the same direction, and acquiring the position and time information of the center point of each cluster;
s3: based on the segmented clustering result of the historical track data, according to a cluster center point moving time difference model, the next cluster center point prediction, namely the next position prediction, is carried out by using the shortest moving time from the passing cluster center point to the candidate next cluster center point and the actual moving time difference.
In a specific implementation, in the method for predicting a next position of a vehicle based on segmented trajectory clustering provided by the present invention, the trajectory linear segmentation process based on the DP algorithm in S1 may include the following steps:
s11: starting point M in a sequence of trajectories to be divided s (x s ,y s ) And end point M e (x e ,y e ) Connected in a straight line and from which the linear equation can be derived:
(y s -y e )x+(x e -x s )y+(x s y e -x e y s )=0
s12: obtaining the vertical distances d from all other points in the track to the straight line, and finding the maximum distance d in the distances max . For any point M (x) m ,y m ) From M to connection M s And M e The vertical distance of the straight line of (a) can be obtained by the following formula:
s13: will d max And compared with a preset division threshold STH. If d is max And if the STH is less than or equal to the STH, the segmentation is not needed, and only the starting point and the end point become key positions. Then, uniformly filling line segments between the starting point and the end point, wherein the number of the filling points is the same as the number of the original points between the two key positions;
s14: otherwise, if d max If > STH, segmentation is required to find the farthest point of the line connecting the start point and the end point, i.e. d max Corresponding points of track, thisOne divides the trajectory into two parts. One is from the first pointSub-tracks to this point, and one from this point to the last pointThe sub-track of (2). Then the two sub-tracks return to step S11 to perform the dividing operation respectively until d max Not greater than STH. At this point, the segmentation of the entire trajectory is complete.
The segmentation threshold STH, the value of which determines the segmentation effect of the track. Generally, the division point of the trajectory should be a position where the vehicle turns, and a minimum turning radius R of the road is introduced in consideration of safety of vehicle travel. When the vehicle is travelling at speed v, the curve radius must be greater than or equal to R; otherwise it is unsafe for the passengers because the road cannot provide enough centripetal force F to prevent the moving vehicle from skidding. The minimum turning radius R can be calculated by the following formula:
where μ is the lateral force coefficient and i is the ultra-high grade, both of which are used to measure road conditions. Therefore, R can be made to be the division threshold STH if the distance d is vertical max If the number of sub tracks is larger than R, the original track should be divided into a plurality of sub tracks. The transverse force coefficient of a common road is 0.02, the ultra-high gradient is 0.08, the vehicle speed is 0-120km/h, so the average value of v can be 60km/h, the segmentation threshold STH can be fixed, and the track segmentation is carried out based on the threshold.
In a specific implementation, in the method for predicting a next position of a vehicle based on segmented track clustering provided by the present invention, the sub-track clustering process based on the single-pass clustering algorithm in S2 may include the following steps:
s21: initializing a sub-track cluster set C, and enabling the sub-track set S to be { S } 1 ,s 2 ,...,s k ,...,s K In which s is k Representing the kth sub-track segmented based on S1, the first term S 1 Set as a new cluster c 1 And adds it to cluster C, when C ═ C 1 }={(s 1 ) K is the total number of the sub tracks;
s22: sequentially traversing other sub-tracks S in the set S 2 ,...s k ,...,s K Comparing the similarity of the sub-tracks and the existing clusters, when K is x, x is more than or equal to 2 and less than or equal to K, obtaining the sub-tracks which are line segments consisting of uniformly filled points based on a track division DP algorithm, and sequentially calculating the current sub-tracks s x With the existing cluster c i Average closest point pair distance ANPPD of, denoted asIt refers to sub-track s x Each point in to a cluster c i Average distance of nearest points, wherein sub-track s x Should be less than or equal to the length of the cluster c i The length of the center track, denoted as | s x |≤|c i L, |; calculating to obtain the current sub-track s x With the existing cluster c i Angle therebetween; i is more than or equal to 1 and less than or equal to | C |, and | C | is the number of elements of the sub-track clustering set C;
FIG. 2 is a graph of when x |≠|c i Specific examples of the meaning of ANPPD when, say, x is 1, and then cluster c i Center locus of s 2 I.e. from short sub-tracks s 1 To a long sub-track s of each point in 2 Average distance of the closest point in (a). As the closest point pair is connected as shown in the solid line part of the figure, the euclidean distances from (1, 1) to (0.8, 2.2), (2, 2) to (1.6, 3.4) and (3, 3) to (2.4, 4.6) are divided into 1.22, 1.46 and 1.71, where the formula for calculating the euclidean distance between two points is as followsThus, it can be calculated
S23: based on the method, the current sub-track s is found x With the existing cluster c i Smallest between
S24: if it is smallestIf the Angle is smaller than the set ANPPD threshold and the corresponding Angle is smaller than the set Angle threshold, the sub-track s is determined x Adding to the current cluster c i Performing the following steps; and recalculates the current cluster c i Center, specifically, the vector direction of the cluster center, assuming cluster c i ={s m ,...,s n 1 ≦ m ≦ n ≦ K, then each sub-track is represented as a vector, respectivelyThe vector direction of the cluster center is derived from the following equation:
determining two end points of the central vector, starting from the current cluster c i The center of the starting point of all the sub-tracks and the end point of the sub-tracks are the current cluster c i The center of the end point of all the sub-tracks is uniformly filled with the central line between the start point and the end point, the number of filling points is the average number of points of each sub-track in the cluster, and the current cluster c is obtained i The center midpoint, i.e., the cluster center point CC i Coordinate position and time information are included;
s25: if it is smallestGreater than the set ANPPD threshold, the current sub-trajectory s should be set x Is created as a new cluster and added toIn cluster C.
In practical implementation, in the method for predicting the next position of the vehicle based on the segmented track clustering provided by the invention, S3 includes the following steps:
s31: constructing a weighted cluster center point transfer graph by using the average moving time of each cluster center point conversion; considering the road network as a graph, the cluster center point CC i Corresponding to nodes of the graph, the conversion between the center points of each cluster is regarded as an edge of the graph, the weight of the edge is the average moving time of the conversion of the center points of the clusters, and the average moving time of the conversion of the center points of the clusters, namely the MS of a given moving road section, is obtained through the average running time of the road section in the historical track data of the vehicle i =CC i →CC i+1 Wherein, MS i Represented as a directed segment containing two consecutive cluster center points CC i And CC i+1 Calculating the average moving timeFurther consider that the road conditions in different time periods are different, resulting inWill vary from time slot to time slot, and the mean moving time of the cluster center point transition at the k-th time slot is expressed asAssuming that the size of the time slot is set to be 1 hour, 24 time slots exist in one day, k is more than or equal to 1 and less than or equal to 24, k is an integer, and 24 snapshots are provided for the weighted cluster center point transfer graph;
s32, calculating the shortest moving time between all the cluster center point pairs; given an initial cluster center point CC i Firstly, selecting a kth weighted cluster center point conversion graph according to the time period of arrival of a user, and calculating a target cluster center point CC based on a shortest path algorithm j Minimum travel time ofAnd, the graph is converted in advance according to the weighted cluster center point toCalculating the shortest moving time between any two possible cluster center points in an off-line mode, and storing the values in an effective data structure;
s33, estimating the actual moving time between the passing cluster center point and the candidate next cluster center point;
given a query track sequence QT ═ P [ [ (P) 1 ,h 1 ),...,(P i ,h i ),...,(P n ,h n )]Wherein (P) i ,h i ) Position information P including longitude and latitude representing ith trace point i And time stamp information h i Since the next position prediction is performed based on the cluster set, the given query trajectory sequence needs to be converted into a query cluster center point sequence QCC ═ CC 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]Wherein (CC) i ,t i ) Location information CC including longitude and latitude representing ith cluster center point i And corresponding time stamp information t i The conversion of the query track sequence QT to the query cluster center point sequence QCC is based on the position information coordinate P in the given query track i And the cluster center point CC acquired in S24 i Coordinate information one-to-one mapping, track point P i With cluster center point CC nearest to it i Converted one by one and time stamp information h i Also from the cluster center point CC i Attached time stamp information t i Replacing one by one; based on the QCC, the cluster center point CC passing through can be estimated and calculated i To the next cluster center point CC n+1 I is more than or equal to 1 and less than or equal to n:
wherein, t n Is the current cluster center point CC n Time stamp information of t i Is the passing cluster center point CC i Timestamp information of (2), current cluster center point CC n To the next cluster center point CC n+1 Is unknown, by averagingMoving timeInstead, n represents the subscript of the current cluster center point;
s34, predicting the central point of the next cluster, namely the next position, by using the difference between the actual moving time and the shortest moving time;
based on the query cluster center point sequence QCC ═ CC [ (CC) obtained above 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]And its candidate next cluster center point CC n+1 If from the passing cluster center point CC i To the next cluster center point CC n+1 Minimum travel time ofLess than its actual movement time at a (CC i ~CC n+1 QCC), then CC is indicated n+1 Not the expected next cluster center point; in order to subsequently calculate the probability of reaching the center point of each candidate next cluster, the shortest moving time and the actual moving time from the center point of each passing cluster in the cluster center point sequence to the center point of the candidate next cluster need to be calculated and compared;
obtaining cluster center point CC from all the passing clusters based on S32 method i To the candidate next cluster center point CC n+1 The sum of the shortest moving time of (a) is:
wherein k is a k weighted cluster center point transition graph determined according to the time period;
based on S33, all passing cluster center points CC in the query cluster center point sequence QCC i To the candidate next cluster center point CC n+1 Is:
in summary, given the query trajectory QT, the query cluster center point sequence QCC ═ CC is obtained by conversion 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]And calculating the probability of the next cluster center point of the access candidate as follows:
wherein, | QCC | represents the number of cluster center points in the query sequence QCC, and z (QCC) is a normalization term, and the calculation formula is:
wherein m represents the number of candidate next cluster center points,a function f is represented that acts on the center point of the r-th candidate next cluster, and the function f is the inverse function.
Illustrative example of a user querying a sequence of cluster center points is shown in FIG. 3, the cluster center point CC visited by the user 1 @8:01、CC 2 @8:02、CC 7 @8:04 and CC 6 @8:07, and 4 candidate next cluster center points, it is apparent that the user prefers to select CC more than time considerations 5 And CC 9 Becomes the next cluster center point. Suppose CC is selected first 9 The actual movement time (e.g., Δ t) between any two cluster center points can be calculated accordingly a (CC 1 ~CC 9 QCC) | (8:09-8:01) ═ 8 minutes). Using the shortest path algorithm, the shortest moving time between any two cluster center points within the respective 9 th time periods can be calculated (here, 60 minutes is set as the size of the time period, and 0:00-0:59 is defined as the first time period, for example, the shortest moving time between the two cluster center points
Claims (3)
1. A vehicle next position prediction method based on segmented track clustering is characterized by comprising the following steps:
s1: trajectory segmentation, given vehicle v a Original track sequenceWherein,indicating vehicle v a At t b Extracting a key position, namely a track inflection point, in the track at the longitude and latitude coordinate position of the moment so as to obtain a piecewise linear sub-track, wherein B is the total number of the sequence;
s2: sub-track clustering, namely, clustering sub-tracks with similar characteristics, namely, clustering sub-tracks with distances and angles smaller than preset values on the same road or the same direction, and acquiring the position and time information of the center point of each cluster;
s3: based on the segmented clustering result of the historical track data, according to a cluster center point moving time difference model, the next cluster center point prediction, namely the next position prediction, is carried out by using the shortest moving time from the passing cluster center point to the candidate next cluster center point and the actual moving time difference.
2. The method for predicting the next position of the vehicle based on the segmented track clustering as claimed in claim 1, wherein the method for clustering the sub-tracks in S2 is as follows:
s21: initializing a sub-track cluster set C, and making the sub-track set S equal to { S } 1 ,s 2 ,...,s k ,...,s K In which s is k Representing the kth sub-track segmented based on S1, the first term S 1 Set as a new cluster c 1 And adds it to cluster C, at which time C ═ C 1 }={(s 1 ) And K is the total number of the sub-tracks;
s22: sequentially traversing other sub-tracks S in the set S 2 ,...s k ,…,s K Comparing the similarity of the sub-tracks and the existing clusters, when K is x, x is more than or equal to 2 and less than or equal to K, obtaining the sub-tracks which are line segments consisting of uniformly filled points based on a track division DP algorithm, and sequentially calculating the current sub-tracks s x With the existing cluster c i Average closest point pair distance ANPPD of (1), expressed asIt refers to sub-track s x From each point to a cluster c i Average distance of nearest points, wherein sub-track s x Should be less than or equal to the length of the cluster c i The length of the center track, denoted as | s x |≤|c i L, |; calculating to obtain the current sub-track s x With the existing cluster c i Angle therebetween; i is more than or equal to 1 and less than or equal to | C |, and | C | is the number of elements of the sub-track clustering set C;
s23: based on the method, the current sub-track s is found x With the existing cluster c i Smallest between
S24: if it is smallestIf the Angle is smaller than the set ANPPD threshold and the corresponding Angle is smaller than the set Angle threshold, the sub-track s is determined x Adding to the current cluster c i Performing the following steps; and recalculates the current cluster c i Center, specifically, the vector direction of the cluster center, assuming cluster c i ={s m ,...,s n 1 m n K, then each sub-track is represented as a vector, respectivelyThe vector direction of the cluster center is derived from the following equation:
determining two end points of the central vector, starting from the current cluster c i The center of the starting point of all the sub-tracks and the end point of the sub-tracks are the current cluster c i The centers of all the sub-track end points are uniformly filled with central lines between the start points and the end points, the number of filling points is the average number of points of each sub-track in the cluster, and the current cluster c is obtained i The center midpoint, i.e., the cluster center point CC i Coordinate position and time information are included;
3. The method for predicting the next position of the vehicle based on the segmented track clustering as claimed in claim 1, wherein the specific method in step 3 is as follows:
s31, constructing a weighted cluster center point transfer graph by using the average moving time of each cluster center point conversion; considering the road network as a graph, the cluster center point CC i Corresponding to nodes of the graph, the conversion between the center points of each cluster is regarded as an edge of the graph, the weight of the edge is the average moving time of the conversion of the center points of the clusters, and the average moving time of the conversion of the center points of the clusters, namely the MS of a given moving road section, is obtained through the average running time of the road section in the historical track data of the vehicle i =CC i →CC i+1 Wherein, MS i Represented as a directed segment, containing two consecutive cluster center points CC i And CC i+1 Calculating the average moving timeFurther consider that the road conditions in different time periods are different, resulting inWill vary from time slot to time slot, and the mean moving time of the cluster center point transition at the k-th time slot is expressed asAssuming that the size of the time slot is set to be 1 hour, 24 time slots exist in one day, k is more than or equal to 1 and less than or equal to 24, k is an integer, and 24 snapshots are provided for the weighted cluster center point transfer graph;
s32, calculating the shortest moving time between all the cluster center point pairs; given a starting cluster center point CC i Firstly, selecting a kth weighted cluster center point conversion graph according to the time period of arrival of a user, and calculating a target cluster center point CC based on a shortest path algorithm j Minimum travel time ofAnd calculating the shortest moving time between any two possible cluster center points in an off-line mode in advance according to the weighted cluster center point conversion graph, and storing the values in an effective data structure;
s33, estimating the actual moving time between the passing cluster center point and the candidate next cluster center point;
given a query trajectory sequence QT ═ P [ [ (P) 1 ,h 1 ),...,(P i ,h i ),...,(P n ,h n )]Wherein (P) i ,h i ) Position information P including longitude and latitude representing ith trace point i And time stamp information h i Since the next position prediction is performed based on the cluster set, the given query trajectory sequence needs to be converted into a query cluster center point sequence QCC ═ CC 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]Therein is disclosedIn (CC) i ,t i ) Location information CC including longitude and latitude representing the ith cluster center point i And corresponding time stamp information t i The conversion of the query track sequence QT to the query cluster center point sequence QCC is based on the position information coordinate P in the given query track i And the cluster center point CC acquired in S24 i Coordinate information mapping, tracing point P i With cluster center point CC nearest to it i Are converted one by one, and the time stamp information h i Also from the cluster center point CC i Attached time stamp information t i Replacing one by one; based on the QCC, the cluster center point CC passing through can be estimated and calculated i To the next cluster center point CC n+1 I is more than or equal to 1 and less than or equal to n:
wherein, t n Is the current cluster center point CC n Time stamp information of t i Is the passing cluster center point CC i Timestamp information of (2), current cluster center point CC n To the next cluster center point CC n+1 Is unknown, using the average moving timeInstead, n represents the subscript of the current cluster center point;
s34, predicting the central point of the next cluster, namely the next position, by using the difference between the actual moving time and the shortest moving time;
based on the obtained query cluster center point sequence QCC ═ CC [ (CC) ] 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]And its candidate next cluster center point CC n+1 If from the passing cluster center point CC i To the next cluster center point CC n+1 Minimum travel time ofLess than its actual movement time at a (CC i ~CC n+1 QCC), then CC is indicated n+1 Not the expected next cluster center point; in order to subsequently calculate the probability of reaching the center point of each candidate next cluster, the shortest moving time and the actual moving time from the center point of each passing cluster in the cluster center point sequence to the center point of the candidate next cluster need to be calculated and compared;
obtaining cluster center point CC from all the passing clusters based on S32 method i To the candidate next cluster center point CC n+1 The sum of the shortest moving time of (a) is:
wherein k is a k weighted cluster center point transition graph determined according to the time period;
based on S33, all passing cluster center points CC in the query cluster center point sequence QCC i To the candidate next cluster center point CC n+1 The sum of the actual movement times of (a) is:
in summary, given the query trajectory QT, the query cluster center point sequence QCC ═ CC is obtained by conversion 1 ,t 1 ),...,(CC i ,t i ),...,(CC n ,t n )]And calculating the probability of the next cluster center point of the access candidate as follows:
wherein, | QCC | represents the number of cluster center points in the query sequence QCC, and z (QCC) is a normalization term, and the calculation formula is:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210598479.8A CN115099308B (en) | 2022-05-30 | 2022-05-30 | Vehicle next position prediction method based on segmented track clustering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210598479.8A CN115099308B (en) | 2022-05-30 | 2022-05-30 | Vehicle next position prediction method based on segmented track clustering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115099308A true CN115099308A (en) | 2022-09-23 |
CN115099308B CN115099308B (en) | 2024-07-05 |
Family
ID=83288499
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210598479.8A Active CN115099308B (en) | 2022-05-30 | 2022-05-30 | Vehicle next position prediction method based on segmented track clustering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115099308B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115841765A (en) * | 2023-02-22 | 2023-03-24 | 交通运输部科学研究院 | Vehicle position blind area monitoring method and device, electronic equipment and readable storage medium |
CN116719068A (en) * | 2023-05-25 | 2023-09-08 | 浪潮智慧科技有限公司 | Water conservancy patrol monitoring method, device and medium based on fusion positioning |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20080084504A (en) * | 2007-03-16 | 2008-09-19 | 제주대학교 산학협력단 | Method for clustering similar trajectories of moving objects in road network databases |
CN113611115A (en) * | 2021-08-06 | 2021-11-05 | 安徽师范大学 | Vehicle track clustering method based on road network sensitive characteristics |
CN113902220A (en) * | 2021-11-10 | 2022-01-07 | 南京邮电大学 | Vehicle track prediction method based on adaptive density clustering algorithm |
-
2022
- 2022-05-30 CN CN202210598479.8A patent/CN115099308B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20080084504A (en) * | 2007-03-16 | 2008-09-19 | 제주대학교 산학협력단 | Method for clustering similar trajectories of moving objects in road network databases |
CN113611115A (en) * | 2021-08-06 | 2021-11-05 | 安徽师范大学 | Vehicle track clustering method based on road network sensitive characteristics |
CN113902220A (en) * | 2021-11-10 | 2022-01-07 | 南京邮电大学 | Vehicle track prediction method based on adaptive density clustering algorithm |
Non-Patent Citations (1)
Title |
---|
杨树亮;毕硕本;NKUNZIMANA A;黄铜;万蕾;: "一种出租车载客轨迹空间聚类方法", 计算机工程与应用, no. 14, 15 July 2018 (2018-07-15), pages 254 - 260 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115841765A (en) * | 2023-02-22 | 2023-03-24 | 交通运输部科学研究院 | Vehicle position blind area monitoring method and device, electronic equipment and readable storage medium |
CN116719068A (en) * | 2023-05-25 | 2023-09-08 | 浪潮智慧科技有限公司 | Water conservancy patrol monitoring method, device and medium based on fusion positioning |
CN116719068B (en) * | 2023-05-25 | 2024-05-28 | 浪潮智慧科技有限公司 | Water conservancy patrol monitoring method, device and medium based on fusion positioning |
Also Published As
Publication number | Publication date |
---|---|
CN115099308B (en) | 2024-07-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN115099308B (en) | Vehicle next position prediction method based on segmented track clustering | |
Hsueh et al. | Map matching for low-sampling-rate GPS trajectories by exploring real-time moving directions | |
CN106767873B (en) | A kind of map-matching method based on space-time | |
CN108362293B (en) | Vehicle track matching method based on key point technology | |
US10909470B2 (en) | Method and apparatus for providing semantic-free traffic prediction | |
CN107146446B (en) | A kind of paths chosen method based on RFID data and Dynamic Vehicle source | |
CN105091889A (en) | Hotspot path determination method and hotspot path determination equipment | |
CN102810118B (en) | A kind of change weighs net K nearest neighbor search method | |
CN105261212B (en) | A kind of trip space-time analysis method based on GPS data from taxi map match | |
CN113401143A (en) | Individualized self-adaptive trajectory prediction method based on driving style and intention | |
CN110174117A (en) | A kind of electric car charging route planning method | |
CN107747947B (en) | Collaborative travel route recommendation method based on historical GPS (global positioning system) track of user | |
CN109269516B (en) | Dynamic path induction method based on multi-target Sarsa learning | |
CN109275121A (en) | A kind of Vehicle tracing method based on adaptive extended kalman filtering | |
CN107917716B (en) | Fixed line navigation method, device, terminal and computer readable storage medium | |
CN110491158A (en) | A kind of bus arrival time prediction technique and system based on multivariate data fusion | |
CN109035787A (en) | It is a kind of to identify vehicles class method for distinguishing using mobile data | |
CN105551239A (en) | Travelling time prediction method and device | |
CN115265555B (en) | Map matching correction method and system based on hidden Markov multi-noise perception | |
CN102081840B (en) | Method for acquiring traffic information through mobile terminals | |
CN109190811A (en) | A kind of car speed tracking based on adaptive extended kalman filtering | |
CN109710714A (en) | A kind of improved road network method of hidden Markov model | |
CN110827540A (en) | Motor vehicle movement mode recognition method and system based on multi-mode data fusion | |
CN116975184A (en) | Vehicle track data processing method, device, equipment and readable storage medium | |
CN115795332A (en) | User travel mode identification method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |