CN111539454A - Vehicle track clustering method and system based on meta-learning - Google Patents
Vehicle track clustering method and system based on meta-learning Download PDFInfo
- Publication number
- CN111539454A CN111539454A CN202010238469.4A CN202010238469A CN111539454A CN 111539454 A CN111539454 A CN 111539454A CN 202010238469 A CN202010238469 A CN 202010238469A CN 111539454 A CN111539454 A CN 111539454A
- Authority
- CN
- China
- Prior art keywords
- vehicle track
- data
- track data
- gps vehicle
- clustering
- 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 33
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 144
- 238000012549 training Methods 0.000 claims abstract description 64
- 238000012360 testing method Methods 0.000 claims description 49
- 238000011156 evaluation Methods 0.000 claims description 30
- 238000010276 construction Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 16
- 238000005259 measurement Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 9
- 238000004364 calculation method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/42—Determining position
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a vehicle track clustering method and system based on meta-learning, belongs to the technical field of vehicle track clustering, and solves the problem that in the prior art, the optimal clustering result cannot be obtained due to the fact that a single track clustering algorithm is adopted by track data of various types. A vehicle track clustering method based on meta-learning comprises the following steps: collecting different types of GPS vehicle track data, and acquiring an optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data; collecting GPS vehicle track data, and training to obtain a meta-learner for vehicle track type division; and acquiring a vehicle track type corresponding to the GPS vehicle track data by using the meta-learner, and clustering the GPS vehicle track data by using an optimal DBSCAN clustering algorithm corresponding to the vehicle track type to obtain a clustering result of the GPS vehicle track data. The optimal clustering result can be obtained for various different types of track data.
Description
Technical Field
The invention relates to the technical field of vehicle track clustering, in particular to a vehicle track clustering method and system based on meta-learning.
Background
In recent years, with the travel demand of people and the wide transportation of goods, more and more vehicles including automobiles and trucks of different types appear in the lives of people, and a large amount of GPS track data is generated every day during the driving process of the vehicles. The track data is a space-time data sequence which is left in the space by the moving object along with the time change and contains a large amount of information, so that the behavior of the moving object can be more intuitively known, and based on the GPS track data of a large number of vehicles, the track clustering can be carried out to carry out data mining, and the potential utilization value can be explored.
Because the track has the characteristics that the track is composed of a plurality of track sequences, if the clustering is performed based on track points or the clustering is performed on the whole track segment, historical driving routes and comprehensive information cannot be observed from the clustering result, so that the GPS track is clustered in a track segment mode, the existing track segment clustering method is different in track division modes, track similarity distance measurement and the like, track data types can be completely different tracks formed by multiple vehicles, and the optimal clustering result cannot be obtained if a single track clustering algorithm is adopted for the track data of multiple different types.
Disclosure of Invention
The invention aims to overcome at least one technical defect and provides a vehicle track clustering method and system based on meta-learning.
In one aspect, the invention provides a vehicle track clustering method based on meta-learning, which comprises the following steps of:
collecting different types of GPS vehicle track data, clustering the different types of GPS vehicle track data by using different DBSCAN clustering algorithms to obtain cluster evaluation indexes corresponding to the different types of GPS vehicle track data, and obtaining the optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data according to the cluster evaluation indexes of the different types of GPS vehicle track data;
collecting GPS vehicle track data, dividing the collected GPS vehicle track data into a training data set and a test data set, and training by using the training data set and the test data set to obtain a meta-learner for vehicle track type division;
and re-collecting GPS vehicle track data, acquiring a vehicle track type corresponding to the GPS vehicle track data by using the meta-learner, and clustering the GPS vehicle track data by using an optimal DBSCAN clustering algorithm corresponding to the vehicle track type to obtain a clustering result of the GPS vehicle track data.
Further, the different DBSCAN clustering algorithms specifically include:
dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, measuring the distance between sub track line segments by using DTW distance, and performing a corresponding DBSCAN clustering algorithm; dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, and performing a corresponding DBSCAN clustering algorithm according to the distance between Hausedorff and quantum track line segments; dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, measuring distances between sub track line segments by weighted distance, and carrying out corresponding DBSCAN clustering algorithm; dividing GPS vehicle track data into sub tracks by combining the MDL minimum description length and an angle threshold value, and measuring the distance between sub track line segments by using a DTW distance; dividing GPS vehicle track data into sub tracks in a mode of combining the MDL minimum description length and an angle threshold value to form a corresponding DBSCAN clustering algorithm according to the distance between Hausedorff distance quantum track line segments; dividing GPS vehicle track data into sub tracks in a dividing mode combining the MDL minimum description length and the angle threshold value, measuring the distance between sub track line segments by the weighted distance, and carrying out a corresponding DBSCAN clustering algorithm.
Further, the obtaining of the optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data according to the clustering evaluation indexes of the different types of GPS vehicle track data specifically comprises comparing DBI values or DI values obtained by clustering the different types of GPS vehicle track data under different DBSCAN clustering algorithms; if the DBI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is smaller than the DBI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data, or if the DI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is larger than the DI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data.
Further, training by using the training data set and the test data set to obtain the meta-learner for vehicle track type division, specifically including calculating Euclidean distances between the test data and each training data, selecting k training data points with the minimum Euclidean distance from the test data, determining frequencies of vehicle track types corresponding to the k training data points, returning the vehicle track type with the highest frequency as a prediction classification of the test data, wherein k is a positive integer.
On the other hand, the invention also provides a vehicle track clustering system based on meta-learning, which comprises a data acquisition module, a clustering algorithm matching module, a meta-learning device construction module and a track data clustering result acquisition module;
the data acquisition module is used for acquiring different types of GPS vehicle track data;
the clustering algorithm matching module is used for clustering different types of GPS vehicle track data by using different DBSCAN clustering algorithms to obtain clustering evaluation indexes corresponding to the different types of GPS vehicle track data, and obtaining the optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data according to the clustering evaluation indexes of the different types of GPS vehicle track data;
the meta-learner building module is used for dividing GPS vehicle track data acquired by the data acquisition module into a training data set and a test data set, and training by using the training data set and the test data set to obtain a meta-learner for vehicle track type division;
and the track data clustering result acquisition module is used for enabling the meta-learner to acquire the vehicle track type corresponding to the GPS vehicle track data, and enabling the optimal DBSCAN clustering algorithm corresponding to the vehicle track type to cluster the GPS vehicle track data to obtain a clustering result of the GPS vehicle track data.
Further, the clustering algorithm matching module acquires optimal DBSCAN clustering algorithms corresponding to different types of GPS vehicle track data according to clustering evaluation indexes of different types of GPS vehicle track data, and specifically comprises the steps of comparing DBI values or DI values of different types of GPS vehicle track data obtained under different DBSCAN clustering algorithms; if the DBI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is smaller than the DBI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data, or if the DI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is larger than the DI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data.
Further, the meta learner constructing module is used for training by using the training data set and the test data set to obtain the meta learner for vehicle track type division, and specifically includes calculating Euclidean distances between the test data and each training data, selecting k training data points with the minimum Euclidean distances from the test data, determining frequencies of vehicle track types corresponding to the k training data points, returning the vehicle track type with the highest frequency as a prediction classification of the test data, and k is a positive integer.
Compared with the prior art, the invention has the beneficial effects that: the method comprises the steps of collecting different types of GPS vehicle track data, clustering the different types of GPS vehicle track data by using different DBSCAN clustering algorithms respectively to obtain cluster evaluation indexes corresponding to the different types of GPS vehicle track data, and obtaining the optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data according to the cluster evaluation indexes of the different types of GPS vehicle track data; collecting GPS vehicle track data, dividing the collected GPS vehicle track data into a training data set and a test data set, and training by using the training data set and the test data set to obtain a meta-learner for vehicle track type division; re-collecting GPS vehicle track data, acquiring a vehicle track type corresponding to the GPS vehicle track data by using the meta-learner, and clustering the GPS vehicle track data by using an optimal DBSCAN clustering algorithm corresponding to the vehicle track type to obtain a clustering result of the GPS vehicle track data; the optimal clustering result can be obtained for various different types of track data.
Drawings
Fig. 1 is a schematic flowchart of a vehicle trajectory clustering method based on meta learning according to embodiment 1 of the present invention;
fig. 2 is a schematic diagram of a taxi driving track according to embodiment 2 of the present invention;
FIG. 3 is a schematic diagram of the transportation path between national nodes according to embodiment 2 of the present invention;
fig. 4 is a schematic diagram of a highway driving track according to embodiment 2 of the present invention;
fig. 5 is a schematic diagram of a truck transportation track between fixed nodes according to embodiment 2 of the present invention;
FIG. 6 is a schematic diagram of dividing track segments by stop points according to embodiment 2 of the present invention;
FIG. 7 is a schematic diagram of calculating a composite distance between two track segments according to embodiment 2 of the present invention;
FIG. 8 is a detailed frame diagram of the trajectory clustering method based on meta-learning according to embodiment 2 of the present invention;
FIGS. 9-12 are schematic diagrams illustrating the clustering results of 4 kinds of trajectory data according to embodiment 2 of the present invention;
FIG. 13 is a comparison chart of the evaluation of the clustering effect according to embodiment 2 of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Example 1
The embodiment of the invention provides a vehicle track clustering method based on meta-learning, which comprises the following steps:
collecting different types of GPS vehicle track data, clustering the different types of GPS vehicle track data by using different DBSCAN clustering algorithms to obtain cluster evaluation indexes corresponding to the different types of GPS vehicle track data, and obtaining the optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data according to the cluster evaluation indexes of the different types of GPS vehicle track data;
collecting GPS vehicle track data, dividing the collected GPS vehicle track data into a training data set and a test data set, and training by using the training data set and the test data set to obtain a meta-learner for vehicle track type division;
and re-collecting GPS vehicle track data, acquiring a vehicle track type corresponding to the GPS vehicle track data by using the meta-learner, and clustering the GPS vehicle track data by using an optimal DBSCAN clustering algorithm corresponding to the vehicle track type to obtain a clustering result of the GPS vehicle track data.
Specifically, when GPS track data of a vehicle is collected, historical tracks of the vehicle are collected through GPS equipment built in the vehicle; the GPS track data of the historical driving of the vehicle is formed by a plurality of complete tracks { TR _1, TR _2, TR _3, …, TR _ n } formed by track data collected by GPS equipment of a plurality of vehicles, and the data generally comprises information of recording license plate numbers, positioning time, longitude, latitude, speed, azimuth angles and the like of the vehicles.
Preferably, the different DBSCAN clustering algorithms specifically include:
dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, measuring the distance between sub track line segments by using DTW distance, and performing a corresponding DBSCAN clustering algorithm; dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, and performing a corresponding DBSCAN clustering algorithm according to the distance between Hausedorff and quantum track line segments; dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, measuring distances between sub track line segments by weighted distance, and carrying out corresponding DBSCAN clustering algorithm; dividing GPS vehicle track data into sub tracks by combining the MDL minimum description length and an angle threshold value, and measuring the distance between sub track line segments by using a DTW distance; dividing GPS vehicle track data into sub tracks in a mode of combining the MDL minimum description length and an angle threshold value to form a corresponding DBSCAN clustering algorithm according to the distance between Hausedorff distance quantum track line segments; dividing GPS vehicle track data into sub tracks in a dividing mode combining the MDL minimum description length and the angle threshold value, measuring the distance between sub track line segments by the weighted distance, and carrying out a corresponding DBSCAN clustering algorithm.
It should be noted that, the optimal clustering algorithm for determining each type of track data by sub-track clustering is performed, firstly, n complete track data of partial different types of track data are manually selected for division, and each track is divided into n sub-tracks { Seg _1, Seg _2, Seg _3, …, Seg _ n }; then defining a sub-track distance measurement function for measuring the similarity degree of two objects during clustering, wherein the similarity distance measurement is close to form a cluster, and the distance between each cluster is as far as possible to form different clusters; when the clustering algorithm of the sub-tracks adopts an improved DBSCAN density clustering algorithm, the similarity of track samples is measured by changing the original Euclidean distance measurement and adopting a defined sub-track distance function, then the sub-tracks are clustered, and the optimal clustering algorithm of each type of track data is determined according to a clustering evaluation index DBI or DI value;
preferably, according to the clustering evaluation indexes of different types of GPS vehicle track data, acquiring optimal DBSCAN clustering algorithms corresponding to the different types of GPS vehicle track data, specifically comprising comparing DBI values or DI values of the different types of GPS vehicle track data obtained under clustering of the different types of DBSCAN clustering algorithms; if the DBI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is smaller than the DBI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data, or if the DI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is larger than the DI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data.
Preferably, the meta learner for vehicle trajectory type classification is obtained by training with the training data set and the test data set, and specifically includes calculating euclidean distances between the test data and each training data, selecting k training data points with the smallest euclidean distances from the test data, determining frequencies of vehicle trajectory types corresponding to the k training data points, and returning the vehicle trajectory type with the highest frequency as the prediction classification of the test data.
It should be noted that, by adopting the idea of meta-learning, a KNN nearest neighbor learner is established, clustering algorithm recommendation is performed on all trajectory data, the types of the unknown trajectory data are classified, and the optimal clustering algorithm is selected for clustering according to different types of trajectory data, so that how to cluster is determined by the optimal clustering algorithm, and the optimal clustering effect is achieved.
Example 2
The embodiment of the invention provides a vehicle track clustering method based on meta-learning, which is used for acquiring different types of GPS vehicle track data, wherein the GPS track data is acquired from 726 vehicles; the data volume of each vehicle track comprises 6 ten thousand pieces of track data, after preprocessing, the average effective track data volume of each vehicle per day is 231, each vehicle at least comprises 1617 tracks in one week, and the GPS track data of the vehicle comprises information such as longitude, latitude, speed, positioning time and the like of positioning as shown in a table 1;
TABLE 1
For the track data of the vehicles, the types of the vehicles are diversified, for example, the data of taxies or trucks can be greatly different due to different types of the vehicles; or the GPS track data of the highway section which runs for a long distance and the short-distance transportation data in the city are also greatly different; the types of the GPS track data are diversified, the track training data are divided into a plurality of types of track data according to the acquisition way and the transport driving characteristics, 4 types of data track types are selected in the embodiment, and as shown in FIGS. 2 to 5, the data track types are respectively a taxi driving track schematic diagram, a national inter-node transport track schematic diagram, a highway driving track schematic diagram and a fixed inter-node truck transport track schematic diagram;
clustering different types of GPS vehicle track data by using different DBSCAN clustering algorithms to obtain cluster evaluation indexes corresponding to the different types of GPS vehicle track data, and obtaining optimal DBSCAN clustering algorithms corresponding to the different types of GPS vehicle track data according to the cluster evaluation indexes of the different types of GPS vehicle track data;
the method comprises the following steps that a track dividing mode and a distance function of similarity measurement in a track clustering process have great influence on a clustering result, in a sub-track dividing stage, a complete GPS track is divided into two dividing modes, the first MDL minimum description length is divided by combining with an angle threshold, namely, the angle between two sub-track line segments is calculated on the basis of calculating the minimum description length of the sub-track, and if the minimum description length exceeds the set angle threshold, the feature point is divided into the sub-tracks; secondly, sub-track division is carried out according to vehicle stopping points, namely, in an integral GPS track road section, in the actual driving process, the sub-track is stopped on a middle node for a period of time due to loading, unloading, carrying and the like, and time intervals exist in recorded track data; performing track division according to the stay points, taking the points of which the change of the longitude and the latitude exceeds a threshold value according to time or space as feature selection points, taking the stay points as feature selection points for dividing a long track, and dividing track sections into schematic diagrams by the stay points, wherein as shown in fig. 6, the standing points where the track stays are taken as feature points for vehicle driving GPS track segmentation, and are divided into a plurality of sub-tracks;
for the DBSCAN clustering algorithm, a track clustering distance function needs to be determined, the similarity of common euclidean distance-based measurement tracks cannot meet the distance measurement between sub-track segments, and the following 3 sub-track segment distance measurement functions are adopted in the embodiment;
(1) track similarity (distance between sub-track segments) is measured based on DTW distance
DTW (dynamic Time warping) dynamic Time warping is a method for measuring similarity of two Time sequences with different lengths; for the divided two sub-tracks Segi＝{p1,p2,...,pi}，Segj＝(p1,p2,...,pi) The calculation formula is as follows:
wherein, gamma (Seg)i,Segj) Representing two sub-trajectories Segi、SegjThe distance between the two DTWs is the DTW distance,
is to calculate two sub-trajectories Segi、SegjThe Euclidean distance between two points of the first point;
(2) measuring track similarity based on Hausdorff distance
The Hausdorff distance is a measure describing the degree of similarity between two sets of points, which is a defined form of the distance between two sets of points; for the divided two sub-tracks Segi＝{p1,p2,...,pi}，Segj＝(p1,p2,...,pi) Hausdorff distance H (Seg) therebetweeni,Segj) The calculation formula is as follows:
H(Segi,Segj)＝max{h(Segi,Segj),h(Segj,Segi)}
wherein,
h(Segi,Segj)＝max(pi∈Segi)min(pj∈Segj)||pi-pj||
h(Segj,Segi)＝max(pj∈Segj)min(pi∈Segi)||pj-pi||
h(Segi,Segj) Is to ask for SegiAt any point p iniThe maximum of the minimum distances to all points in the other trajectory; is a set of points SegiSum point set SegjThe Euclidean distance is used as the distance between the two adjacent pixels.
(3) Weighted average of vertical, parallel and angular distances measures inter-track similarity
As shown in FIG. 4, the distances between the trace segments of the defined quantity mentioned in the TRACLUS cluster are closely related to the parallel distance, the perpendicular distance and the angle between the two line segments, and therefore, a schematic diagram of calculating the combined distance between the two trace segments is obtained by combining the perpendicular, the parallel and the angle distances, and the two trace segments are L as shown in FIG. 7iAnd Lj，dist(Li,Lj) For the integrated weighted distance, the calculation is as follows;
dist(Li,Lj)＝d⊥(Li,Lj)+dP(Li,Lj)+dθ(Li,Lj)
wherein the vertical distance, parallel distance and angle calculation formulas are as follows.
dP＝min(L0,L1)
dθ＝||Lj||×sin(θ)
Wherein D is0、D1Is LjAt LiUpper vertical distance, L0、L1Is LjAt LiTheta is the included angle between two track line segments; based on 2 track division modes and 3 different distance metricsThe function can enable the DBSCAN clustering algorithm to have various combinations, and 6 different DBSCAN clustering algorithms can be formed to cluster different types of track data as shown in the table 2;
TABLE 2
Sequentially clustering the selected 4 types of vehicle track data under 6 clustering algorithms, selecting the optimal track clustering algorithm of each type of data according to the clustering evaluation index, and corresponding the DBSCAN algorithm of track division and distance measurement with the 4 types of vehicle track data; when the track data is encountered again, the corresponding DBSCAN algorithm can be directly applied to obtain the optimal clustering result;
the clustering evaluation indexes are also internal and external, because part of data is clustered, the internal indexes which do not use a reference model to directly investigate clustering results are adopted as the evaluation indexes, DBI, DI and the like are commonly used, the smaller the DBI value is, the larger the DI value is, the better the clustering effect is represented, and table 3 shows the track clustering results corresponding to different track types (taking DBI as an example);
TABLE 3
It can be seen from table 3 that the smaller the DBI value is, the better the clustering effect is, and the best clustering effect of the tracks of different data types is determined according to the smallest DBI value; the minimum DBI values of the track types 1 and 2 are the clustering results of DBSCAN _1, and the optimal clustering algorithms are DBSCAN _ 1; the optimal clustering algorithms of the track type 3 and the track type 4 are DBSCAN _2 and DBSCAN _3 respectively; the average clustering effect of the DBSCAN _3 is better; when new track data is clustered, judging whether the tracks corresponding to the optimal clustering algorithm are similar or not, and adopting the same clustering algorithm; a meta-learner for vehicle track type division is established, and track clustering can be performed on all track data;
in order to enable each kind of track data to be capable of adaptively selecting the best combination for DBSCAN clustering, Meta Learning (Meta Learning) or Learning of an academic society is adopted, and a Meta learner is applied to train an input Meta feature set X and a target feature set Y of different training samples to obtain a Meta learner model; common machine learning algorithms can be used as meta-learner models: decision trees (CART), multi-level perceptron (MLP), K-nearest neighbor (KNN) algorithms; in the embodiment, KNN is used as a meta-learner model; setting a k value, and then taking k data closest to the sample as a classification category through Euclidean distance measurement; a detailed frame diagram of a trajectory clustering method based on meta-learning, as shown in fig. 8;
collecting GPS vehicle track data, dividing the collected GPS vehicle track data into a training data set and a testing data set, and separating the track data in the initial clustering process according to 30% of the testing data and 70% of the training data; training by using the training data set and the test data set to obtain a meta learner for vehicle track type division; extracting meta feature set X from training data setnExtracting a target characteristic set Y from different types of track data in test datamAnd the characteristics of the continuous track data are extracted, as shown in table 4,
TABLE 4
In Table 4, m and n are characteristics of the trajectory data itself, and the logarithm is taken to normalize the range of values, log2Number of objects (GPS trace data volume), log2Number of attributes (number of vehicles in GPS track); rhosThe method is average absolute correlation (a Spireman correlation coefficient is adopted here), the Spireman correlation coefficient belongs to a non-parameter statistical method, and the distribution of original variables is not required; the calculation formula is as follows:
wherein d isiFor rank difference, the variables X, Y are sorted separatelyThe difference of the latter position change, e.g. sorted by longitude from small to large and by latitude from small to large, d of the trajectory dataiThe rank difference is the difference between the two sequencing positions; n is the number of data in the variable;
skewness (skewness), which is a measure for describing the skew direction and degree of statistical data distribution, is a digital feature of the asymmetric degree of statistical data distribution, and defines that skewness is the standard third-order center distance of a sample, and the calculation formula is as follows:
wherein σ is a standard deviation, x is a mean value, and n is the number of tracks;
kurtosis (Kurtosis), also called a Kurtosis coefficient, represents a characteristic number of a probability density distribution curve at the peak value height of an average value, namely a statistic for describing the steepness degree of all value distribution forms in the population; the formula for the peak is as follows:
average value of speed:
training according to KNN algorithm steps: 1) calculating Euclidean distances between the test data and each training data; 2) sorting according to the increasing relation of the distances; 3) selecting k test data points with the minimum distance; 4) determining the occurrence frequency of the track data category where the first k test data points are located; 5) and returning the category with the highest occurrence frequency in the first k points as a prediction classification result of the test data. A common method is to start with k being 1, estimate the error rate of the classifier by using a test set, repeat the KNN algorithm step process, and add 1 to k each time, allowing one neighbor to be added; selecting a k value corresponding to the minimum error rate;
re-collecting GPS vehicle track data, acquiring a vehicle track type corresponding to the GPS vehicle track data by using the meta-learner, and clustering the GPS vehicle track data by using an optimal DBSCAN clustering algorithm corresponding to the vehicle track type to obtain a clustering result of the GPS vehicle track data;
in a specific embodiment, after the meta learner is established, all the track data can be input into the meta learner, and the output is the type of the corresponding track data; clustering the GPS vehicle track data by using an optimal DBSCAN clustering algorithm corresponding to the vehicle track type to obtain a clustering result of the GPS vehicle track data; the clustering result diagrams of the 4 kinds of trajectory data (trajectory data 1-4) are respectively shown in fig. 9-12, and are clustering results of all trajectory data based on meta learning;
comparing the vehicle track clustering method based on meta-learning disclosed by the embodiment of the invention with a vehicle track clustering method without meta-learning, for example, after processing the same track data by DBSCAN _3 and TRACLUS clustering, using DBI index as an index for evaluating clustering effect to compare, wherein the clustering results of the vehicle track clustering method based on meta-learning have 4 types, so that the average value of measurement index DBI is compared with other methods, and a clustering effect evaluation comparison graph is shown in FIG. 13; the DBI value of the common DBSCAN density clustering algorithm is 3.3, the DBI value of the TRACLUS track clustering is 2.7, so that the TRACLUS track clustering algorithm is superior to the common DBSCAN track clustering, and the result obtained by adding the self-adaptive learning clustering of meta learning is 1.97; therefore, the result shows that the clustering effect of the vehicle track clustering method based on the meta learning disclosed by the invention is superior to that of other track clustering algorithms.
Example 3
The invention also provides a vehicle track clustering system based on meta-learning, which comprises a data acquisition module, a clustering algorithm matching module, a meta-learning device construction module and a track data clustering result acquisition module;
the data acquisition module is used for acquiring different types of GPS vehicle track data;
the clustering algorithm matching module is used for clustering different types of GPS vehicle track data by using different DBSCAN clustering algorithms to obtain clustering evaluation indexes corresponding to the different types of GPS vehicle track data, and obtaining the optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data according to the clustering evaluation indexes of the different types of GPS vehicle track data;
the meta-learner building module is used for dividing GPS vehicle track data acquired by the data acquisition module into a training data set and a test data set, and training by using the training data set and the test data set to obtain a meta-learner for vehicle track type division;
and the track data clustering result acquisition module is used for enabling the meta-learner to acquire the vehicle track type corresponding to the GPS vehicle track data, and enabling the optimal DBSCAN clustering algorithm corresponding to the vehicle track type to cluster the GPS vehicle track data to obtain a clustering result of the GPS vehicle track data.
Preferably, the clustering algorithm matching module obtains the optimal dbss clustering algorithm corresponding to different types of GPS vehicle trajectory data according to the clustering evaluation index of the different types of GPS vehicle trajectory data, and specifically includes comparing DBI values or DI values obtained by different types of GPS vehicle trajectory data under different dbss clustering algorithm clustering; if the DBI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is smaller than the DBI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data, or if the DI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is larger than the DI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data.
Preferably, the meta learner constructing module is configured to obtain the meta learner for vehicle trajectory type division by training using the training data set and the test data set, and specifically includes calculating euclidean distances between the test data and each of the training data, selecting k training data points with the smallest euclidean distances from the test data, determining frequencies of vehicle trajectory types corresponding to the k training data points, returning a vehicle trajectory type with the highest frequency as a prediction classification of the test data, where k is a positive integer.
The invention discloses a vehicle track clustering method and system based on meta-learning, which are characterized in that different types of GPS vehicle track data are collected, different DBSCAN clustering algorithms are used for clustering the different types of GPS vehicle track data respectively to obtain clustering evaluation indexes corresponding to the different types of GPS vehicle track data, and optimal DBSCAN clustering algorithms corresponding to the different types of GPS vehicle track data are obtained according to the clustering evaluation indexes of the different types of GPS vehicle track data; collecting GPS vehicle track data, dividing the collected GPS vehicle track data into a training data set and a test data set, and training by using the training data set and the test data set to obtain a meta-learner for vehicle track type division; re-collecting GPS vehicle track data, acquiring a vehicle track type corresponding to the GPS vehicle track data by using the meta-learner, and clustering the GPS vehicle track data by using an optimal DBSCAN clustering algorithm corresponding to the vehicle track type to obtain a clustering result of the GPS vehicle track data; the optimal clustering result can be obtained for various different types of track data.
The above-described embodiments of the present invention should not be construed as limiting the scope of the present invention. Any other corresponding changes and modifications made according to the technical idea of the present invention should be included in the protection scope of the claims of the present invention.
Claims (7)
1. A vehicle track clustering method based on meta-learning is characterized by comprising the following steps:
collecting different types of GPS vehicle track data, clustering the different types of GPS vehicle track data by using different DBSCAN clustering algorithms to obtain cluster evaluation indexes corresponding to the different types of GPS vehicle track data, and obtaining the optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data according to the cluster evaluation indexes of the different types of GPS vehicle track data;
collecting GPS vehicle track data, dividing the collected GPS vehicle track data into a training data set and a test data set, and training by using the training data set and the test data set to obtain a meta-learner for vehicle track type division;
and re-collecting GPS vehicle track data, acquiring a vehicle track type corresponding to the GPS vehicle track data by using the meta-learner, and clustering the GPS vehicle track data by using an optimal DBSCAN clustering algorithm corresponding to the vehicle track type to obtain a clustering result of the GPS vehicle track data.
2. The meta learning-based vehicle trajectory clustering method according to claim 1, wherein the different DBSCAN clustering algorithms specifically include:
dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, measuring the distance between sub track line segments by using DTW distance, and performing a corresponding DBSCAN clustering algorithm; dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, and performing a corresponding DBSCAN clustering algorithm according to the distance between Hausedorff and quantum track line segments; dividing GPS vehicle track data to form sub tracks in a stopping point dividing mode, measuring distances between sub track line segments by weighted distance, and carrying out corresponding DBSCAN clustering algorithm; dividing GPS vehicle track data into sub tracks by combining the MDL minimum description length and an angle threshold value, and measuring the distance between sub track line segments by using a DTW distance; dividing GPS vehicle track data into sub tracks in a mode of combining the MDL minimum description length and an angle threshold value to form a corresponding DBSCAN clustering algorithm according to the distance between Hausedorff distance quantum track line segments; dividing GPS vehicle track data into sub tracks in a dividing mode combining the MDL minimum description length and the angle threshold value, measuring the distance between sub track line segments by the weighted distance, and carrying out a corresponding DBSCAN clustering algorithm.
3. The vehicle track clustering method based on meta-learning according to claim 1, wherein the optimal DBSCAN clustering algorithm corresponding to different types of GPS vehicle track data is obtained according to a clustering evaluation index of different types of GPS vehicle track data, and specifically includes comparing DBI values or DI values obtained by different types of GPS vehicle track data under different DBSCAN clustering algorithms; if the DBI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is smaller than the DBI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data, or if the DI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is larger than the DI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data.
4. The meta learning-based vehicle track clustering method according to claim 1, wherein a meta learner for vehicle track type classification is obtained by training with the training data set and the test data set, and specifically comprises calculating euclidean distances between the test data and each training data, selecting k training data points with the smallest euclidean distances from the test data, determining frequencies of vehicle track types corresponding to the k training data points, and returning the vehicle track type with the highest frequency as a prediction classification of the test data, wherein k is a positive integer.
5. A vehicle track clustering system based on meta-learning is characterized by comprising a data acquisition module, a clustering algorithm matching module, a meta-learner construction module and a track data clustering result acquisition module;
the data acquisition module is used for acquiring different types of GPS vehicle track data;
the clustering algorithm matching module is used for clustering different types of GPS vehicle track data by using different DBSCAN clustering algorithms to obtain clustering evaluation indexes corresponding to the different types of GPS vehicle track data, and obtaining the optimal DBSCAN clustering algorithm corresponding to the different types of GPS vehicle track data according to the clustering evaluation indexes of the different types of GPS vehicle track data;
the meta-learner building module is used for dividing GPS vehicle track data acquired by the data acquisition module into a training data set and a test data set, and training by using the training data set and the test data set to obtain a meta-learner for vehicle track type division;
and the track data clustering result acquisition module is used for enabling the meta-learner to acquire the vehicle track type corresponding to the GPS vehicle track data, and enabling the optimal DBSCAN clustering algorithm corresponding to the vehicle track type to cluster the GPS vehicle track data to obtain a clustering result of the GPS vehicle track data.
6. The meta-learning based vehicle track clustering system according to claim 5, wherein the clustering algorithm matching module obtains optimal DBSCAN clustering algorithms corresponding to different types of GPS vehicle track data according to clustering evaluation indexes of different types of GPS vehicle track data, and specifically comprises comparing DBI values or DI values obtained by different types of GPS vehicle track data under different DBSCAN clustering algorithms; if the DBI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is smaller than the DBI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data, or if the DI value of certain type of GPS vehicle track data clustered by certain type of DBSCAN clustering algorithm is larger than the DI value of the certain type of GPS vehicle track data clustered by other types of DBSCAN clustering algorithms, the certain type of DBSCAN clustering algorithm is the best DBSCAN clustering algorithm corresponding to the certain type of GPS vehicle track data.
7. The meta learning based vehicle track clustering system according to claim 5, wherein the meta learner construction module is configured to obtain the meta learner for vehicle track type classification by training using the training data set and the test data set, and specifically includes calculating Euclidean distances between the test data and each training data, selecting k training data points with the smallest Euclidean distance from the test data, determining frequencies of vehicle track types corresponding to the k training data points, and returning a vehicle track type with a highest frequency as a prediction classification of the test data, where k is a positive integer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010238469.4A CN111539454B (en) | 2020-03-30 | 2020-03-30 | Vehicle track clustering method and system based on meta-learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010238469.4A CN111539454B (en) | 2020-03-30 | 2020-03-30 | Vehicle track clustering method and system based on meta-learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111539454A true CN111539454A (en) | 2020-08-14 |
| CN111539454B CN111539454B (en) | 2023-04-18 |
Family
ID=71980143
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010238469.4A Active CN111539454B (en) | 2020-03-30 | 2020-03-30 | Vehicle track clustering method and system based on meta-learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111539454B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112050820A (en) * | 2020-09-02 | 2020-12-08 | 平安科技（深圳）有限公司 | Road matching method and device, electronic equipment and readable storage medium |
| CN112215261A (en) * | 2020-09-18 | 2021-01-12 | 武汉理工大学 | Vehicle OD point clustering method, system, device and storage medium based on meta-learning |
| CN112269844A (en) * | 2020-09-24 | 2021-01-26 | 桂林电子科技大学 | General adjoint mode distributed mining method based on large-scale track data |
| CN112345261A (en) * | 2020-10-29 | 2021-02-09 | 南京航空航天大学 | Aero-engine pumping system abnormity detection method based on improved DBSCAN algorithm |
| CN112394354A (en) * | 2020-12-02 | 2021-02-23 | 中国人民解放军国防科技大学 | Method for identifying HRRP fusion target small samples based on meta-learning in different polarization modes |
| CN112633592A (en) * | 2020-12-30 | 2021-04-09 | 鱼快创领智能科技(南京)有限公司 | Vehicle frequent-running route calculation method and system based on machine learning clustering algorithm |
| CN112906812A (en) * | 2021-03-09 | 2021-06-04 | 西安电子科技大学 | Vehicle track clustering method based on outlier removal |
| CN113052265A (en) * | 2021-04-26 | 2021-06-29 | 上海海事大学 | Moving object track simplification algorithm based on feature selection |
| CN113902220A (en) * | 2021-11-10 | 2022-01-07 | 南京邮电大学 | Vehicle track prediction method based on adaptive density clustering algorithm |
| CN113963541A (en) * | 2021-10-21 | 2022-01-21 | 宜信普惠信息咨询(北京)有限公司 | Method and device for identifying vehicle aggregation event based on GPS signal data |
| CN114353810A (en) * | 2022-01-10 | 2022-04-15 | 河海大学 | HMM efficient map matching method based on R tree and track segmentation |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103605362A (en) * | 2013-09-11 | 2014-02-26 | 天津工业大学 | Learning and anomaly detection method based on multi-feature motion modes of vehicle traces |
| CN105117737A (en) * | 2015-10-08 | 2015-12-02 | 北京邮电大学 | Method and apparatus for determining real direction of vehicle on basis of locus vector of vehicle |
| US20160265918A1 (en) * | 2013-10-30 | 2016-09-15 | Hewlett Packard Enterprise Development Lp | Signal space based navigation |
| US20180359651A1 (en) * | 2017-06-12 | 2018-12-13 | Cisco Technology, Inc. | Automatic characterization of ap behaviors |
| CN110309383A (en) * | 2019-06-17 | 2019-10-08 | 武汉科技大学 | Ship trajectory clustering analysis method based on improved DBSCAN algorithm |
| CN110570658A (en) * | 2019-10-23 | 2019-12-13 | 江苏智通交通科技有限公司 | Method for identifying and analyzing abnormal vehicle track at intersection based on hierarchical clustering |
-
2020
- 2020-03-30 CN CN202010238469.4A patent/CN111539454B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103605362A (en) * | 2013-09-11 | 2014-02-26 | 天津工业大学 | Learning and anomaly detection method based on multi-feature motion modes of vehicle traces |
| US20160265918A1 (en) * | 2013-10-30 | 2016-09-15 | Hewlett Packard Enterprise Development Lp | Signal space based navigation |
| CN105117737A (en) * | 2015-10-08 | 2015-12-02 | 北京邮电大学 | Method and apparatus for determining real direction of vehicle on basis of locus vector of vehicle |
| US20180359651A1 (en) * | 2017-06-12 | 2018-12-13 | Cisco Technology, Inc. | Automatic characterization of ap behaviors |
| CN110309383A (en) * | 2019-06-17 | 2019-10-08 | 武汉科技大学 | Ship trajectory clustering analysis method based on improved DBSCAN algorithm |
| CN110570658A (en) * | 2019-10-23 | 2019-12-13 | 江苏智通交通科技有限公司 | Method for identifying and analyzing abnormal vehicle track at intersection based on hierarchical clustering |
Non-Patent Citations (1)
| Title |
|---|
| 江玉玲; 熊振南; 唐基宏: "基于轨迹段DBSCAN的船舶轨迹聚类算法" * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112050820B (en) * | 2020-09-02 | 2024-05-07 | 平安科技（深圳）有限公司 | Road matching method, device, electronic equipment and readable storage medium |
| CN112050820A (en) * | 2020-09-02 | 2020-12-08 | 平安科技（深圳）有限公司 | Road matching method and device, electronic equipment and readable storage medium |
| CN112215261A (en) * | 2020-09-18 | 2021-01-12 | 武汉理工大学 | Vehicle OD point clustering method, system, device and storage medium based on meta-learning |
| CN112269844B (en) * | 2020-09-24 | 2021-08-06 | 桂林电子科技大学 | General adjoint mode distributed mining method based on large-scale track data |
| CN112269844A (en) * | 2020-09-24 | 2021-01-26 | 桂林电子科技大学 | General adjoint mode distributed mining method based on large-scale track data |
| CN112345261B (en) * | 2020-10-29 | 2022-05-03 | 南京航空航天大学 | Aero-engine pumping system abnormity detection method based on improved DBSCAN algorithm |
| CN112345261A (en) * | 2020-10-29 | 2021-02-09 | 南京航空航天大学 | Aero-engine pumping system abnormity detection method based on improved DBSCAN algorithm |
| CN112394354A (en) * | 2020-12-02 | 2021-02-23 | 中国人民解放军国防科技大学 | Method for identifying HRRP fusion target small samples based on meta-learning in different polarization modes |
| CN112633592A (en) * | 2020-12-30 | 2021-04-09 | 鱼快创领智能科技(南京)有限公司 | Vehicle frequent-running route calculation method and system based on machine learning clustering algorithm |
| CN112633592B (en) * | 2020-12-30 | 2023-07-18 | 鱼快创领智能科技(南京)有限公司 | Vehicle constant running route calculation method and system based on machine learning clustering algorithm |
| CN112906812A (en) * | 2021-03-09 | 2021-06-04 | 西安电子科技大学 | Vehicle track clustering method based on outlier removal |
| CN113052265A (en) * | 2021-04-26 | 2021-06-29 | 上海海事大学 | Moving object track simplification algorithm based on feature selection |
| CN113052265B (en) * | 2021-04-26 | 2024-04-05 | 上海海事大学 | Moving object track simplifying method based on feature selection |
| CN113963541A (en) * | 2021-10-21 | 2022-01-21 | 宜信普惠信息咨询(北京)有限公司 | Method and device for identifying vehicle aggregation event based on GPS signal data |
| CN113902220A (en) * | 2021-11-10 | 2022-01-07 | 南京邮电大学 | Vehicle track prediction method based on adaptive density clustering algorithm |
| CN113902220B (en) * | 2021-11-10 | 2024-04-02 | 南京邮电大学 | Vehicle track prediction method based on self-adaptive density clustering algorithm |
| CN114353810A (en) * | 2022-01-10 | 2022-04-15 | 河海大学 | HMM efficient map matching method based on R tree and track segmentation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111539454B (en) | 2023-04-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111539454B (en) | Vehicle track clustering method and system based on meta-learning | |
| US20160125307A1 (en) | Air quality inference using multiple data sources | |
| CN111832225A (en) | Method for constructing driving condition of automobile | |
| CN113159364A (en) | Passenger flow prediction method and system for large-scale traffic station | |
| CN107845260B (en) | Method for identifying public transport trip mode of user | |
| CN111949896B (en) | Track abnormality detection method based on track big data | |
| CN112101132B (en) | Traffic condition prediction method based on graph embedding model and metric learning | |
| Zhang et al. | Vehicle re-identification for lane-level travel time estimations on congested urban road networks using video images | |
| CN111144446A (en) | Driver identity recognition method and system based on space-time grid | |
| CN115600833A (en) | Smart city restriction scheme determination method, internet of things system, device and medium | |
| CN106940929B (en) | Traffic data prediction method and device | |
| Wang et al. | Vehicle reidentification with self-adaptive time windows for real-time travel time estimation | |
| CN116071722A (en) | Lane geometric information extraction method, system, equipment and medium based on road section track | |
| CN116010838A (en) | Vehicle track clustering method integrating density value and K-means algorithm | |
| Gkountouna et al. | Traffic flow estimation using probe vehicle data | |
| CN116578661A (en) | Vehicle track time-space reconstruction method and system based on attention mechanism | |
| CN117542195A (en) | Traffic running state identification method, device, equipment and medium | |
| CN117409569A (en) | Intelligent traffic system and method based on big data | |
| CN116989801A (en) | Map matching method and device for low-frequency long tracks of complex road network | |
| Widhalm et al. | Robust road link speed estimates for sparse or missing probe vehicle data | |
| CN112651455B (en) | Traffic flow missing value filling method based on non-negative matrix factorization and dynamic time warping algorithm | |
| CN113313008B (en) | Target and identification tracking method based on YOLOv3 network and mean shift | |
| CN114492544A (en) | Model training method and device and traffic incident occurrence probability evaluation method and device | |
| Yuan et al. | Adaptive Gaussian mixture model for identifying outliers in historical route travel times | |
| Gong et al. | Heterogeneous Traffic Flow Detection Using CAV-Based Sensor With I-GAIN |
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 |