CN112632202A - Dynamic map matching method based on high-order hidden Markov model - Google Patents

Dynamic map matching method based on high-order hidden Markov model Download PDF

Info

Publication number
CN112632202A
CN112632202A CN202011505866.XA CN202011505866A CN112632202A CN 112632202 A CN112632202 A CN 112632202A CN 202011505866 A CN202011505866 A CN 202011505866A CN 112632202 A CN112632202 A CN 112632202A
Authority
CN
China
Prior art keywords
point
candidate
gps
gps point
markov model
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
Application number
CN202011505866.XA
Other languages
Chinese (zh)
Other versions
CN112632202B (en
Inventor
付晓
张悦
杨晨
张嘉旭
刘志远
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southeast University
Original Assignee
Southeast University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Southeast University filed Critical Southeast University
Priority to CN202011505866.XA priority Critical patent/CN112632202B/en
Publication of CN112632202A publication Critical patent/CN112632202A/en
Application granted granted Critical
Publication of CN112632202B publication Critical patent/CN112632202B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Algebra (AREA)
  • Probability & Statistics with Applications (AREA)
  • Evolutionary Biology (AREA)
  • Operations Research (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Remote Sensing (AREA)
  • Navigation (AREA)
  • Position Fixing By Use Of Radio Waves (AREA)

Abstract

The invention discloses a dynamic map matching method based on a high-order hidden Markov model, which comprises the steps of obtaining candidate road sections of GPS points through GPS point data and road network information, further obtaining candidate points of the GPS points in each candidate road section, then calculating the high-order hidden Markov model transition probability and the state transition probability of a high-order hidden Markov model of each candidate point of the GPS points, and solving the high-order hidden Markov model by adopting a Viterbi algorithm to realize the matching of the GPS points and the road sections in a road network.

Description

Dynamic map matching method based on high-order hidden Markov model
Technical Field
The invention relates to the technical field of map matching, in particular to a dynamic map matching method based on a high-order hidden Markov model.
Background
In recent years, with the development of Intelligent Transportation Systems (ITS), the accumulation rate of traffic data has also increased exponentially. GPS trajectory data is a key data source for various traffic analysis techniques as well as location services. In fact, because the positioning error of the GPS data is unavoidable, there is a map matching precision problem when the GPS data is specifically applied. The map matching problem is to match the GPS data with errors to the road network, reduce the influence of the errors and maximize the effectiveness of the data.
Existing map matching algorithms can be broadly classified into four categories according to the nature of the method they employ: geometric methods, topological methods, probabilistic methods, and other advanced techniques. The geometric method performs map matching by using geometric information (such as distance, angle, shape, and the like) of the GPS points and the road network without considering the topological relation of the road network, and has high efficiency, but has a high error when matching low-precision GPS data with a complex road network. The topological method considers both geometric factors and road network topological relation, but is greatly influenced by low-frequency sampling intervals and sampling noise. The probability statistical method sets an elliptical or rectangular confidence region for each GPS point, so that the probability is obtained according to the distance between the GPS point and the corresponding position in the confidence region, and the optimal matching path is determined according to the probability value.
Disclosure of Invention
The purpose of the invention is as follows: the dynamic map matching method is high in matching precision and matching efficiency, and is simple and easy to implement.
The technical scheme is as follows: the dynamic map matching method based on the high-order hidden Markov model is used for realizing the matching of the GPS position of a target object at a target moment t and a corresponding road section in a road network; the method is characterized by comprising the following steps:
step 1: aiming at GPS point g of target object at each time n within preset time rangenRespectively and sequentially executing the step 1.1 to the step 1.2, and then entering the step 2; wherein n is t-w +1, t-w, t, w is a preset time number,
step 1.1: incorporating GPS points gnAnd the road network information,establishing an R tree index for the road network, acquiring a road section near the GPS point in a preset range as a candidate road section, and further acquiring a GPS point gnCandidate point positions in each candidate road section corresponding to the candidate point positions; entering step 1.2;
step 1.2: obtaining GPS point gnThe high-order hidden Markov model observation probability of each candidate point and the GPS point gnThe state transition probability of the high-order hidden Markov model of each candidate point;
step 2, according to each GPS point gnAnd each GPS point gnUsing Viterbi algorithm to solve GPS point g of target object at time ttIn combination with the GPS point g, is a high-order hidden Markov model of each candidate pointtObtaining a GPS point g at the candidate point position in each candidate road section corresponding to the GPS pointtImplementing GPS point g at the best candidate point in the road networktAnd matching with road sections in the road network.
As a preferred aspect of the present invention, before step 1, the method further includes preprocessing a GPS point of the target object within a preset time period, where the preprocessing method includes:
GPS point g at each time for target objectnThe following steps are executed in real time:
step A, calculating a GPS point g of a target object at a time nnAnd GPS point g at time n-1n-1Great circle distance D betweenn-1，n：
Step B, judging the great circle distance Dn-1，nIf the distance is less than the preset minimum distance threshold, the GPS point g is droppednOtherwise, matching the data;
determining the great circle distance Dn-1，nIf the distance is greater than the preset maximum distance threshold, according to a formula:
Figure BDA0002844926830000021
GPS point g for time nnAnd GPS point g at time n-1n-1Interpolation processing is performed, and the interpolation point is set as a GPS point adjacent to the time n.
As a preferred embodiment of the invention, in step 1.1, GPS points g are acquirednThe method for the candidate point position in each candidate road section corresponding to the candidate point position comprises the following steps:
for GPS point gnEach candidate link of (a): from GPS point gnRespectively making vertical lines on each candidate road section, and if the drop foot falls on the candidate road section, defining the drop foot as a GPS point gnA candidate point in the candidate segment; if the drop falls on the extension line of the candidate road section, defining the GPS point g in the candidate road sectionnThe point with the shortest connecting line is a GPS point gnA candidate point in the candidate segment.
As a preferred embodiment of the present invention, in step 1.2, according to the formula:
Figure BDA0002844926830000022
obtaining GPS point gnThe ith candidate point
Figure BDA0002844926830000023
And GPS point gn-1The jth candidate point
Figure BDA0002844926830000024
High order hidden Markov model observation probability
Figure BDA0002844926830000025
I ═ 1,2.. In，InIs GPS point gnThe total number of candidate points of (a);
Figure BDA0002844926830000026
is GPS point gn-1J, where J is 1,2n-1，Jn-1Is GPS point gn-1The total number of candidate points of (a);
Figure BDA0002844926830000031
as candidate points
Figure BDA0002844926830000032
To the candidate point
Figure BDA0002844926830000033
The first order hidden markov model state transition probability,
Figure BDA0002844926830000034
as candidate points
Figure BDA0002844926830000035
The probability of observation of the first-order hidden markov model of (1),
Figure BDA0002844926830000036
as candidate points
Figure BDA0002844926830000037
First order hidden markov model observation probability.
As a preferred aspect of the present invention, according to the formula:
Figure BDA0002844926830000038
obtaining candidate points
Figure BDA0002844926830000039
To the candidate point
Figure BDA00028449268300000310
First order hidden Markov model state transition probability
Figure BDA00028449268300000311
Wherein p issameAs a weight parameter, snIs GPS point gn-1And GPS point gnDistance of great circle and candidate point
Figure BDA00028449268300000312
To
Figure BDA00028449268300000313
S is taken as the statistic valuenThe median of (3).
As a preferred aspect of the present invention, according to the formula:
Figure BDA00028449268300000314
obtaining candidate points
Figure BDA00028449268300000315
First order hidden Markov model observation probability
Figure BDA00028449268300000316
Wherein the content of the first and second substances,
Figure BDA00028449268300000317
is GPS point gnTo the candidate point
Figure BDA00028449268300000318
Great circle distance ofnFor a preset statistic, σnGet GPS point gnThe median of the distances to all candidate points, ρ is the preset road weight related to the road grade rlevel and the driver's preference level plevel for the road segment, θ is the angle α with the road directionroadAnd track direction angle alphaGPSThe associated heading weight.
As a preferred embodiment of the present invention, in step 1.2, according to the formula:
Figure BDA00028449268300000319
obtaining GPS point gnN candidate point of
Figure BDA00028449268300000320
State transition probability of high-order hidden Markov model
Figure BDA00028449268300000321
I ═ 1,2.. In，InIs GPS point gnThe total number of candidate points of (a);
Figure BDA00028449268300000322
is GPS point gn-1J, where J is 1,2n-1，Jn-1Is GPS point gn-1The total number of candidate points of (a);
Figure BDA00028449268300000323
is the GPS point gn of the target object at the time n-2-2K-th candidate point of (1, 2.. K'n；K'nFor GPS point gn-2The total number of candidate points of (a); k is a radical ofnFor GPS point gn-2And GPS point gnGreat circle distance and candidate points
Figure BDA00028449268300000324
And candidate point
Figure BDA00028449268300000325
The path distance difference between, λ is knAverage value of (a).
As a preferred aspect of the present invention, according to the following formula:
Figure BDA0002844926830000041
obtaining GPS points gn-2And GPS point gnGreat circle distance and candidate points
Figure BDA0002844926830000042
And candidate point
Figure BDA0002844926830000043
The difference k of the path distances betweenn。
In a preferred embodiment of the present invention, in step 3, the Viterbi algorithm is used to solve the GPS point g of the target object at time ttThe high-order hidden markov model of each candidate point is as follows:
Figure BDA0002844926830000044
and solving an objective function for the high-order hidden Markov model.
As a preferred scheme of the present invention, in step 1, the GPS points from step 1.1 to step 1.2 are respectively executed by selecting an adaptive sliding window; the GPS points that respectively perform step 1.1 to step 1.2 include: the GPS point of the target object at the target time t, and the GPS points of the target object at w-1 continuous time before and adjacent to the target time t.
Has the advantages that: compared with the prior art, the method provided by the invention combines the GPS point data and the road network information to obtain the high-order Markov model observation probability and the high-order Markov model state transition probability of the GPS point, and further solves the high-order hidden Markov model through the expanded Viterbi algorithm to realize the matching of the GPS point and the road section in the road network; the matching precision and the matching efficiency of the matching by the method are high, and the method is simple and easy to realize.
Drawings
FIG. 1 is a flow chart of a map matching method provided by an embodiment of the invention;
FIG. 2 is a diagram illustrating the possible influence of road weight parameter ρ on the matching result according to an embodiment of the present invention;
FIG. 3 is a diagram of the possible impact of vehicle heading weight θ on the matching results provided by an embodiment of the present invention;
FIG. 4 shows a data point g provided by an embodiment of the present inventiontThe candidate matching position of (2);
FIG. 5 is a mismatch case for a first order hidden Markov model provided by an embodiment of the invention;
FIG. 6 is a second order hidden Markov model dimension reduction process provided by an embodiment of the invention;
fig. 7 illustrates the steps of an extended Viterbi algorithm according to an embodiment of the invention;
FIG. 8 is a test example provided by an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
It should be clear that the hidden markov model is a mainstream paradigm of network-based dynamic modeling, and it well adapts to the process of finding a suitable matching point (i.e. hidden state) for each GPS point (i.e. observation state) in the map matching problem, and the hidden markov model combines factors such as geometry, topology, probability, etc. to effectively improve the map matching accuracy and has a better convergence effect; hidden markov models can be used in advanced map matching techniques.
The dynamic map matching method based on the high-order hidden Markov model is used for realizing the matching of the GPS position of a target object at a target moment t and a corresponding road section in a road network, as shown in figure 1.
When matching, preprocessing the GPS point data: for the currently received data point gtCalculate gtAnd gt-1Great circle distance Dt-1,tIf D ist-1，tIf less than threshold thmin, then current point g is droppedtIt is not matched. If D ist-1,tAbove the threshold value thmax, it is linearly interpolated. The interpolation formula is as follows:
Figure BDA0002844926830000051
through the data preprocessing process, the static GPS data points can be effectively removed. Meanwhile, linear interpolation is carried out on the GPS points which are sparsely distributed, and the spatial information contained in the GPS data is increased. According to the preprocessing method, only the GPS points at the target time t may be preprocessed, or the GPS points at the time before the time t may be preprocessed by referring to the method. GPS point g at time ttFor example, the method for obtaining the observation probability and the state transition probability of the high-order hidden markov model is specifically as follows, wherein the high-order hidden markov model is specifically a second-order hidden markov model.
And establishing an R tree index for the road network, and searching a candidate point position corresponding to the GPS point. The pre-processed GPS data is obtained by searching for a candidate point of its real location in a road network using a two-stage algorithm, and the specific implementation process is as follows:
the process 11: establishing an R tree index for the middle points of all road segments in the road network, and searching n road segments near the point in the error radius range of the GPS by using the R tree index as candidate road segments, wherein as shown in FIGS. 2 and 3, black lines in the graph represent the candidate road segments;
and (4) process 12: g is prepared fromtVertically projected on the candidate road section, projected point
Figure BDA0002844926830000052
A candidate point for that point. If the foot falls outside the road section, the distance g of the road section is takentOne end point nearer is used as
Figure BDA0002844926830000053
As shown in FIG. 4, gtAre respectively the candidate points of
Figure BDA0002844926830000054
gtDistances to the candidate points are respectively
Figure BDA0002844926830000055
Referring to FIG. 5, existing dynamic map matching algorithms typically only start with a single GPS point, taking into account its local geometry and road topologyThe flutter relationship causes the precision of the dynamic map matching algorithm to be far behind that of the global map matching algorithm. However, in application scenarios such as real-time navigation and road section travel time prediction, dynamic map matching is indispensable, and the real-time performance and the high efficiency of the dynamic map matching cannot be met by a global map matching algorithm. Considering the scenario shown in fig. 5, the conventional dynamic map matching algorithm cannot achieve correct matching. From gtPoint to gt+2At this point, the vehicle is not turning, but is traveling straight. The correct matching path should be
Figure BDA0002844926830000056
But in the process of dynamic first-order hidden Markov model map matching, it is wrongly matched as
Figure BDA0002844926830000057
The reason for this error is that the first-order hidden markov model map matching algorithm only considers the observation probability of a single point and the transition probability between two points, but lacks a measure of state transition on a larger scale.
In the scheme of the invention, the second-order hidden Markov model observation probability is calculated, the first-order hidden Markov model is improved during calculation, and then the observation probability in the second-order hidden Markov model is calculated, and the specific calculation process comprises the following steps:
the process 21: setting a sliding window: setting the size of the sliding window equal to the preset time number w, and if the current point gtAfter adding the sliding window, if the window overflows, deleting the first point g in the windowt-w+1At this time, gt-w+1The matching result of the points is finally determined. Points within the window may have their matching points changed as new points are added. The size of the sliding window has an influence on the matching precision and the matching efficiency of the algorithm. In general, increasing the window size allows the algorithm to consider more GPS points in one matching, thereby improving the matching accuracy. If the size of the sliding window contains all the GPS points, the algorithm is changed into global matching; however, too large a sliding window will also affect the matching efficiency, and the algorithm will lose the real-time matching characteristic. To balance the algorithmPrecision and computational efficiency, a self-adaptive sliding window is provided.
In the scheme provided by the invention, the size of the adaptive sliding window is defined into three levels: a small window, a middle window and a large window. By averaging the positioning errors of the current GPS points, our algorithm will automatically select the appropriate window size. The calculation formula of the average value of the GPS positioning error is defined as follows:
Figure BDA0002844926830000061
wherein c isnIs GPS point gnThe candidate points of (1).
The process 22: and calculating second-order observation probability. Second-order hidden Markov model observation probability matrix
Figure BDA0002844926830000062
The formula can be derived from the first order observation probability and the state transition probability as follows:
Figure BDA0002844926830000063
as can be seen from the above equation, the second-order hidden markov model observation probability is the product of the first-order observation probability and the first-order state transition probability of two consecutive GPS points.
Calculating a second-order state transition probability matrix: and improving the first-order hidden Markov model, and calculating the state transition probability in the second-order hidden Markov model. Defining second order state transition probabilities
Figure BDA0002844926830000064
The calculation formula of (a) is as follows:
Figure BDA0002844926830000065
wherein λ is ktAverage value of (a), ktIs GPS point gt-2And GPS point gtGreat circle distance and candidate points
Figure BDA0002844926830000066
And candidate point
Figure BDA0002844926830000067
The difference between the path distances, λ is ktAverage value of (a).
According to the following formula:
Figure BDA0002844926830000068
obtaining GPS point gt-2And GPS point gtGreat circle distance and candidate points
Figure BDA0002844926830000069
And candidate point
Figure BDA00028449268300000610
Difference k of path distance betweent。
And 5: the invention solves Markov model by second-order Viterbi algorithm, introduces second-order hidden Markov model to solve GPS track matching problem, the objective function solved by second-order hidden Markov model can be expressed as:
Figure BDA0002844926830000071
the Viterbi algorithm is a high-efficiency dynamic programming algorithm, can effectively avoid repeated search of paths, and can quickly obtain an optimal solution. It is widely used to solve first order hidden markov models. In order to solve the second-order hidden Markov model with complexity, the traditional Viterbi algorithm is expanded. The specific process is as follows:
the process 31: and (5) reducing the dimension of the model. In the case of a second-order hidden markov model,
Figure BDA0002844926830000072
considered as the observation probability, is equivalent to the observation probability of a single candidate point in the first-order hidden markov model. The observation probability of the second-order hidden markov model is the product of the observation probability and the state transition probability of two consecutive candidate points in the first-order hidden markov model. Thus, as shown in fig. 6, the order of the second order hidden markov model can be reduced according to the aforementioned method.
The process 32: and (5) carrying out iterative search. After completing the process 1, referring to fig. 7, iterative computation is performed using a conventional Viterbi algorithm to solve the second-order hidden markov model, and the process is as follows:
(1) and starting from the first layer of nodes, calculating the observation probability of each layer of nodes after dimensionality reduction and the state transition probability between two adjacent layers of nodes.
(2) The maximum total probability of each node from the second layer to the last layer is calculated. The maximum total probability for each node and the predecessor nodes are saved.
(3) And selecting the node with the highest total probability in the last layer, and backtracking the precursor nodes until reaching the first layer.
Through the steps, the optimal matching path of the GPS track data can be found along with the sliding of the sliding window. Fig. 8 is a visualization of a GPS track data matching result, where black dots represent GPS track data points, and black broken lines are paths obtained after the original GPS data points are matched to a road network by the method of the present invention.
The dynamic map matching method based on the high-order hidden Markov model is used for matching GPS track data in a complex urban road network environment. Based on the hidden Markov model, the traditional first-order hidden Markov model is expanded to a high order, and the space-time relation between different GPS points can be processed more effectively. Secondly, when the observation probability and the state transition probability of the high-order hidden Markov model are calculated, the trip preference of a driver to a road section, the road grade, the vehicle course and the road network topological relation are comprehensively considered, and the model is more in line with the actual situation. Meanwhile, compared with the traditional method for fixing the sliding window, the method adopts an expanded Viterbi algorithm to solve the high-order hidden Markov model, and uses a self-adaptive sliding window method to adjust the size of the sliding window in real time, so that the accuracy and the calculation efficiency of the algorithm can be balanced, and the map matching problem can be efficiently and dynamically solved.
The above description is only a preferred embodiment of the present invention, and it will be apparent to those skilled in the art that various modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be considered as the protection scope of the present invention.

Claims (10)

1. A dynamic map matching method based on a high-order hidden Markov model is used for realizing the matching of a GPS point of a target object at a target moment t and a corresponding road section in a road network; the method is characterized by comprising the following steps:
step 1: aiming at GPS point g of target object at each time n within preset time rangenRespectively and sequentially executing the step 1.1 to the step 1.2, and then entering the step 2; wherein n is t-w +1, t-w, t, w is a preset time number;
step 1.1: incorporating GPS points gnAnd road network information, wherein an R tree index is established for the road network, a road section near the GPS point in a preset range is obtained as a candidate road section, and then a GPS point g is obtainednCandidate point positions in each candidate road section corresponding to the candidate point positions; entering step 1.2;
step 1.2: obtaining GPS point gnThe high-order hidden Markov model observation probability of each candidate point and the GPS point gnThe state transition probability of the high-order hidden Markov model of each candidate point;
step 2: according to each GPS point gnAnd each GPS point gnUsing Viterbi algorithm to solve GPS point g of target object at time ttIn combination with the GPS point g, is a high-order hidden Markov model of each candidate pointtObtaining a GPS point g at the candidate point position in each candidate road section corresponding to the GPS pointtImplementing GPS point g at the best candidate point in the road networktAnd matching with road sections in the road network.
2. The dynamic map matching method based on the high-order hidden markov model according to claim 1, wherein before step 1, the method further comprises preprocessing the GPS points of the target object within a preset time period, and the preprocessing method comprises:
GPS point g at each time for target objectnThe following steps are executed in real time:
step A, calculating a GPS point g of a target object at a time nnAnd GPS point g at time n-1n-1Great circle distance D betweenn-1,n；
Step B, judging the great circle distance Dn-1，nIf the distance is less than the preset minimum distance threshold, the GPS point g is droppednOtherwise, matching the data;
determining the great circle distance Dn-1，nIf the distance is greater than the preset maximum distance threshold, according to a formula:
Figure FDA0002844926820000011
GPS point g for time nnAnd GPS point g at time n-1n-1Interpolation processing is performed, and the interpolation point is set as a GPS point adjacent to the time n.
3. The dynamic map matching method based on high-order hidden markov model according to claim 1, wherein in step 1.1, a GPS point g is obtainednThe method for the candidate point position in each candidate road section corresponding to the candidate point position comprises the following steps:
for GPS point gnEach candidate link of (a): from GPS point gnRespectively making vertical lines on each candidate road section, and if the drop foot falls on the candidate road section, defining the drop foot as a GPS point gnA candidate point in the candidate segment; if the drop falls on the extension line of the candidate road section, defining the GPS point g in the candidate road sectionnThe point with the shortest connecting line is a GPS point gnA candidate point in the candidate segment.
4. A dynamic map matching method based on a high order hidden markov model according to claim 1, wherein in step 1.2, according to the formula:
Figure FDA0002844926820000021
obtaining GPS point gnThe ith candidate point
Figure FDA0002844926820000022
And GPS point gn-1The jth candidate point
Figure FDA0002844926820000023
High order hidden Markov model observation probability
Figure FDA0002844926820000024
I ═ 1,2.. In，InIs GPS point gnThe total number of candidate points of (a);
Figure FDA0002844926820000025
is GPS point gn-1J, where J is 1,2n，JnIs GPS point gn-1The total number of candidate points of (a);
Figure FDA0002844926820000026
as candidate points
Figure FDA0002844926820000027
To the candidate point
Figure FDA0002844926820000028
The first order hidden markov model state transition probability,
Figure FDA0002844926820000029
as candidate points
Figure FDA00028449268200000210
The probability of observation of the first-order hidden markov model of (1),
Figure FDA00028449268200000211
as candidate points
Figure FDA00028449268200000212
First order hidden markov model observation probability.
5. The method of claim 4, wherein the method comprises, in accordance with the formula:
Figure FDA00028449268200000213
obtaining candidate points
Figure FDA00028449268200000214
To the candidate point
Figure FDA00028449268200000215
First order hidden Markov model state transition probability
Figure FDA00028449268200000216
Wherein p issameAs a weight parameter, snIs GPS point gn-1And GPS point gnDistance of great circle and candidate point
Figure FDA00028449268200000217
To
Figure FDA00028449268200000218
S is taken as the statistic valuenThe median of (3).
6. The method of claim 4, wherein the method comprises, in accordance with the formula:
Figure FDA00028449268200000219
obtaining candidate points
Figure FDA00028449268200000220
First order hidden Markov model observation probability
Figure FDA00028449268200000221
Wherein the content of the first and second substances,
Figure FDA00028449268200000222
is GPS point gnTo the candidate point
Figure FDA00028449268200000223
Great circle distance ofnFor a preset statistic, σnGet GPS point gnThe median of the distances to all candidate points, ρ is the preset road weight related to the road grade rlevel and the driver's preference level plevel for the road segment, θ is the angle α with the road directionroadAnd track direction angle alphaGPSThe associated heading weight.
7. A dynamic map matching method based on a high order hidden markov model according to claim 1, wherein in step 1.2, according to the formula:
Figure FDA0002844926820000031
obtaining GPS point gnN candidate point of
Figure FDA0002844926820000032
State transition probability of high-order hidden Markov model
Figure FDA0002844926820000033
I ═ 1,2.. In，InIs GPS point gnThe total number of candidate points of (a);
wherein the content of the first and second substances,
Figure FDA0002844926820000034
is GPS point gn-1J, where J is 1,2n-1，Jn-1Is GPS point gn-1The total number of candidate points of (a);
Figure FDA0002844926820000035
is the GPS point gn of the target object at the time n-2-2K-th candidate point of (1, 2.. K'n；K'nFor GPS point gn-2The total number of candidate points of (a); k is a radical ofnFor GPS point gn-2And GPS point gnGreat circle distance and candidate points
Figure FDA0002844926820000036
And candidate point
Figure FDA0002844926820000037
The path distance difference between, λ is knAverage value of (a).
8. The method of claim 7, wherein the method comprises the following steps:
Figure FDA0002844926820000038
obtaining GPS points gn-2And GPS point gnGreat circle distance and candidate points
Figure FDA0002844926820000039
And candidate point
Figure FDA00028449268200000310
The difference k of the path distances betweenn。
9. The method of claim 1, wherein in step 3, a Viterbi algorithm is used to solve the GPS point g of the target object at time ttThe high-order hidden markov model of each candidate point is as follows:
Figure FDA00028449268200000311
and solving an objective function for the high-order hidden Markov model.
10. The dynamic map matching method based on the high-order hidden markov model according to claim 1, wherein in step 1, the GPS points, each of which performs step 1.1 to step 1.2, are selected by an adaptive sliding window; the GPS points that respectively perform step 1.1 to step 1.2 include: the GPS point of the target object at the target time t, and the GPS points of the target object at w-1 continuous time before and adjacent to the target time t.
CN202011505866.XA 2020-12-18 2020-12-18 Dynamic map matching method based on high-order hidden Markov model Active CN112632202B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011505866.XA CN112632202B (en) 2020-12-18 2020-12-18 Dynamic map matching method based on high-order hidden Markov model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011505866.XA CN112632202B (en) 2020-12-18 2020-12-18 Dynamic map matching method based on high-order hidden Markov model

Publications (2)

Publication Number Publication Date
CN112632202A true CN112632202A (en) 2021-04-09
CN112632202B CN112632202B (en) 2022-11-18

Family

ID=75317250

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011505866.XA Active CN112632202B (en) 2020-12-18 2020-12-18 Dynamic map matching method based on high-order hidden Markov model

Country Status (1)

Country Link
CN (1) CN112632202B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113282699A (en) * 2021-07-22 2021-08-20 北京交通大学 Road network matching method for noisy and unidentified parameter bicycle track data
CN114088101A (en) * 2022-01-14 2022-02-25 华清科盛(北京)信息技术有限公司 Method for correcting matched positioning point based on hidden Markov model
CN115265555A (en) * 2022-07-25 2022-11-01 上海交通大学 Hidden Markov based multi-noise perception map matching correction method and system
WO2024021632A1 (en) * 2022-07-29 2024-02-01 京东城市(北京)数字科技有限公司 Real-time trajectory data processing method, apparatus and system, and electronic device

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106595680A (en) * 2016-12-15 2017-04-26 福州大学 Vehicle GPS data map matching method based on hidden markov model
CN109710714A (en) * 2018-12-21 2019-05-03 西安交通大学 A kind of improved road network method of hidden Markov model
CN110095127A (en) * 2019-04-08 2019-08-06 西北大学 A kind of hidden Markov model map-matching method based on segmentation

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106595680A (en) * 2016-12-15 2017-04-26 福州大学 Vehicle GPS data map matching method based on hidden markov model
CN109710714A (en) * 2018-12-21 2019-05-03 西安交通大学 A kind of improved road network method of hidden Markov model
CN110095127A (en) * 2019-04-08 2019-08-06 西北大学 A kind of hidden Markov model map-matching method based on segmentation

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113282699A (en) * 2021-07-22 2021-08-20 北京交通大学 Road network matching method for noisy and unidentified parameter bicycle track data
CN113282699B (en) * 2021-07-22 2022-01-28 北京交通大学 Road network matching method for noisy and unidentified parameter bicycle track data
CN114088101A (en) * 2022-01-14 2022-02-25 华清科盛(北京)信息技术有限公司 Method for correcting matched positioning point based on hidden Markov model
CN115265555A (en) * 2022-07-25 2022-11-01 上海交通大学 Hidden Markov based multi-noise perception map matching correction method and system
CN115265555B (en) * 2022-07-25 2024-05-07 上海交通大学 Map matching correction method and system based on hidden Markov multi-noise perception
WO2024021632A1 (en) * 2022-07-29 2024-02-01 京东城市(北京)数字科技有限公司 Real-time trajectory data processing method, apparatus and system, and electronic device

Also Published As

Publication number Publication date
CN112632202B (en) 2022-11-18

Similar Documents

Publication Publication Date Title
CN112632202B (en) Dynamic map matching method based on high-order hidden Markov model
CN110260870B (en) Map matching method, device, equipment and medium based on hidden Markov model
Hug et al. Particle-based pedestrian path prediction using LSTM-MDL models
Koller et al. Fast hidden Markov model map-matching for sparse and noisy trajectories
CN113065594B (en) Road network extraction method and device based on Beidou data and remote sensing image fusion
CN110736471B (en) Low-frequency floating vehicle track data road network matching method based on hidden Markov model
CN112882073B (en) Time-threshold-based hidden Markov model human-vehicle integration algorithm
CN112052802A (en) Front vehicle behavior identification method based on machine vision
CN109522385B (en) Method for judging M-N matching mode of multi-scale road network
CN110986965A (en) Low-frequency floating vehicle track data road network matching method based on hidden Markov model
CN115265555B (en) Map matching correction method and system based on hidden Markov multi-noise perception
Wang et al. Fusing heterogeneous traffic data by Kalman filters and Gaussian mixture models
CN116047498A (en) Maneuvering target tracking method based on maximum correlation entropy extended Kalman filtering
CN110011847B (en) Data source quality evaluation method under sensing cloud environment
Zhao et al. Adaptive non-linear joint probabilistic data association for vehicle target tracking
CN113932821A (en) Track map matching method based on continuous window average direction features
CN113645280A (en) Vehicle position data uploading method and device, electronic equipment and storage medium
CN117392268A (en) Laser scanning mapping method and system based on self-adaption combined CPD and ICP algorithm
CN110779526B (en) Path planning method, device and storage medium
Piyadi Gamage et al. Adjusted empirical likelihood for time series models
CN116304969A (en) Vehicle track multi-mode prediction method considering road information based on LSTM-GNN
CN106934124B (en) Adaptive variable window method based on measurement change detection
CN110411450A (en) It is a kind of for compressing the map-matching method of track
EP4266004A1 (en) User trajectory positioning method, electronic device and computer storage medium
CN109917794B (en) Global path planning method and device

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