CN110609881A - Vehicle trajectory deviation detection method, system and storage medium - Google Patents
Vehicle trajectory deviation detection method, system and storage medium Download PDFInfo
- Publication number
- CN110609881A CN110609881A CN201910803343.4A CN201910803343A CN110609881A CN 110609881 A CN110609881 A CN 110609881A CN 201910803343 A CN201910803343 A CN 201910803343A CN 110609881 A CN110609881 A CN 110609881A
- Authority
- CN
- China
- Prior art keywords
- track
- point
- historical
- data set
- track data
- 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.)
- Pending
Links
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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Abstract
The invention discloses a vehicle track deviation detection method, a system and a storage medium, wherein the method comprises the following steps: extracting a starting point, a terminal point and a turning point from a GPS data set of a driving route as key positions; according to the coordinates of the key positions, carrying out track matching in a historical track data set and/or a planning track data set acquired from an online map service; when historical track data or planning track data are matched, storing tracks described by a plurality of key positions into a historical track data set; and when the historical track data or the planning track data are not matched, judging that the driving route deviates from the track. The invention can compress the number of coordinate points for path matching by extracting key points from the GPS data set, thereby reducing the resources occupied by storage and operation; unnecessary details can be shielded, and the influence of a large amount of tiny errors on the matching precision is avoided. The invention can be widely applied to the field of big data.
Description
Technical Field
The invention relates to the field of big data, in particular to a vehicle track deviation detection method, a vehicle track deviation detection system and a storage medium.
Background
With the development of the internet, the online car reservation has become an option for many people to go out. However, the unlawful act associated with a cyber-contract occurs at all times. In order to restrict the driving behavior of the net car booking driver, a large number of net car booking driving monitoring methods are proposed. These detection methods include identity monitoring and route monitoring, among others.
Route monitoring generally utilizes GPS data for detection, and the principle of route monitoring is to compare the difference between a planned route and an actual driving route to detect a deviation route. However, in the prior art, a large number of GPS coordinate points are often collected to describe a route, which increases the data volume, occupies storage and calculation resources, and the like; on the other hand, because the GPS has errors and the vehicle may have lane changes, slight turns, and the like during driving, a large number of GPS coordinate points are used to describe the route, so that too many fine errors are introduced, and the matching accuracy is reduced.
Disclosure of Invention
To solve the above technical problems, the present invention aims to: a vehicle trajectory deviation detection method, system, and storage medium are provided to compress the data amount of GPS and improve the accuracy of matching.
A first aspect of an embodiment of the present invention provides:
a vehicle trajectory deviation detection method includes the steps of:
extracting a starting point, a terminal point and a turning point from a GPS data set of a driving route as key positions;
according to the coordinates of the key positions, carrying out track matching in a historical track data set and/or a planning track data set acquired from an online map service;
when historical track data or planning track data are matched, storing tracks described by a plurality of key positions into a historical track data set;
and when the historical track data or the planning track data are not matched, judging that the driving route deviates from the track.
Further, the starting point refers to the earliest collected coordinate point in the GPS data set, the ending point refers to the latest collected coordinate point in the GPS data set, and the turning point is obtained by:
taking any three coordinate points continuously acquired in time in the GPS data set as a first point, a second point and a third point;
taking the vectors of the first point and the second point as a first vector, and taking the vectors of the second point and the third point as a second vector;
and determining whether the second point is a turning point according to the size of an included angle between the first vector and the second vector.
Further, the track matching is performed in a historical track dataset and/or a planned track dataset obtained from an online map service according to the coordinates of the plurality of key positions, and specifically includes:
according to the coordinates of the starting point and the coordinates of the end point, track indexing is carried out in the historical track data set;
when the number of the historical tracks obtained by indexing in the historical track data set is larger than a first set threshold value, carrying out track matching through a clustering model according to a plurality of key positions;
and when the number of the historical tracks obtained by indexing in the historical track data set is less than or equal to a first set threshold value or the historical track data is not matched through a clustering model, carrying out track matching on a planning track data set obtained from the online map service according to a plurality of key positions.
Further, the number of the clustering models is at least two, and at least two clustering models are different types of models;
the track matching through the clustering model specifically includes:
if at least one clustering model is matched with the historical track data, taking the historical track data obtained by matching as an output result;
and if all the clustering models are not matched with the historical track data, judging that no result is matched.
Further, the method also comprises the following steps:
after the deviation of the driving route from the track is judged, accumulating the times of deviation of the driving route from the track;
and taking the point deviating from the track as a new starting point, and performing track indexing in the historical track data set according to the coordinates of the starting point and the coordinates of the end point again so as to perform new track matching.
Further, the method also comprises the following steps:
and when the number of times of the deviation of the driving route from the track is greater than a second set threshold value, generating alarm information.
Further, the method also comprises the following steps:
and performing de-diversification processing on the coordinates of the starting point and the coordinates of the end point.
Further, before the track matching, the method also comprises the following steps:
and aligning the two groups of track data which are matched with each other in sequence.
A second aspect of an embodiment of the present invention provides:
a vehicle trajectory deviation detection system comprising:
a memory for storing a program;
and the processor is used for loading the program to execute the track deviation detection method.
A third aspect of embodiments of the present invention provides:
a storage medium storing a program executed by a processor to implement the vehicle trajectory deviation detecting method.
The invention has the beneficial effects that: according to the invention, the number of coordinate points for path matching is reduced by extracting key points from the GPS data set, so that on one hand, resources occupied by storage and operation can be reduced; on the other hand, useless details can be shielded, and the influence of a large amount of tiny errors on the matching precision is avoided.
Drawings
FIG. 1 is a flow chart of a vehicle trajectory deviation detection method according to an embodiment of the present invention;
FIG. 2 is a flow chart of a vehicle trajectory deviation detection method according to another embodiment of the present invention;
FIG. 3 is a schematic diagram of GPS data acquisition points for a vehicle travel track in accordance with one embodiment of the present invention;
FIG. 4 is a schematic illustration of key points extracted from the trace of FIG. 3;
FIG. 5 is a diagram illustrating key points of a historical track in accordance with an embodiment of the present invention;
FIG. 6 is a schematic illustration of key points of a travel path of a first vehicle in accordance with an exemplary embodiment of the present invention;
FIG. 7 is a schematic illustration of key points of a travel path of a second vehicle in accordance with an embodiment of the present invention;
fig. 8 is a schematic diagram of key points of a travel track of a third vehicle according to an embodiment of the present invention.
Detailed Description
The invention is further described with reference to the drawings and the specific examples.
Referring to fig. 1, the present embodiment discloses a vehicle trajectory deviation detection method, which may be implemented by a terminal electronic device or a server, and includes step S101:
s101, extracting a starting point, an end point and a turning point from a GPS data set of a driving route as key positions.
In this step, the GPS data may be collected by the GPS module during the travel of the vehicle and stored in a data set. In this embodiment, these GPS data are collected in real time.
There are many GPS data points in the GPS dataset, however the number of these data points is very large. For urban roads, redundancy in data occurs due to the existence of limitations in road linearity, especially for relatively regular straight roads. For example, in a linear road, only two GPS data points are needed to express a motion trajectory, but the number of uploaded data points is often tens or even hundreds, which greatly increases the storage space, and when the inter-trajectory similarity measurement is performed, the longitude and latitude slightly differ due to behaviors such as vehicle lane change and the like, so that the inter-trajectory similarity is reduced, and the abnormal judgment on the trajectory is influenced. Therefore, compression processing is required for the GPS data set to reduce the storage space and enhance the post-algorithm processing efficiency.
The inflection point in this step is a point where a turn occurs, and an angle between the point of the criterion and a vector formed by the front point and the rear point is relatively large, for example, 30 ° or more, which can be adjusted according to the actual sensitivity requirement.
The end point and the start point are defined as the last point and the first point, respectively, in this embodiment.
And S102, according to the coordinates of the key positions, carrying out track matching in a historical track data set and/or a planning track data set acquired from an online map service.
In this step, the coordinates of the plurality of key positions may be compared with historical trajectory data, or the planned trajectory data obtained from the online map service may be compared, so as to match a suitable trajectory.
Specifically, in the present embodiment, the historical trajectory data is also described in terms of coordinates of a plurality of key positions. For example an L-shaped trajectory, which is described by the coordinates of only three key positions. Namely a starting point, a terminal point and an inflection point.
In this step, many comparison methods may be adopted, for example, a clustering model is established for matching, or matching is performed by comparing whether the key position falls into a route through which a comparison track passes, or the like.
Of course, in order to reduce the computation amount of matching, the start point and the end point may be indexed first, and the same trajectory as the end point and the start point may be found first for comparison.
S103, storing the tracks described by the key positions into a historical track data set when historical track data or planning track data are matched.
In this step, if the historical trajectory data or the planned trajectory data provided by the online map service is successfully matched, it is indicated that the vehicle does not deviate from the trajectory. In brief, the principle is that the track that a vehicle walks in the past or the track that a map plans according to the actual traffic condition is not abnormal.
However, the key points collected on the road where the vehicle travels each time may be different, and therefore, the track described by the key position where the vehicle travels is also stored in the historical track data set in the step, so that the data volume of the historical track is increased, and subsequent comparison and big data analysis are facilitated.
And S104, judging that the driving route deviates from the track when the historical track data or the planning track data are not matched.
In this step, if neither the historical trajectory data nor the planned trajectory data is matched, it indicates that the driving route has deviated from the trajectory, and the abnormal condition is present.
For the handling of the abnormal condition, various means can be adopted, for example, sending a short message to remind the passenger, reminding the management department, and the like. As for the trigger condition of the means, various ones may be set. For example, an alarm may be given for one abnormality or may be given for a plurality of abnormalities.
In many cases, if a sudden road condition, such as a car accident, occurs, causing the vehicle to detour, the driving track of the vehicle may be different from the historical track and the planned track of the map with a high probability. If an alarm is generated once, false alarm is easy to occur, so the embodiment adopts a mode of multiple accumulation. After each abnormality is determined, the trajectory is newly planned with the point where the abnormality occurs as a starting point. If the track of the vehicle in the subsequent running can be found out from the historical track data set or the planning track data set to be matched with the track, the running track of the vehicle is still normal. In this case, the individual abnormality does not trigger an alarm to prevent a false alarm from occurring in the case of an abrupt route change. Of course, in this embodiment, the planned trajectory data provided by the map service is generally the shortest route. The shortest route herein may be a route having the shortest distance or a route having the shortest time.
In summary, in the present embodiment, by extracting the key points from the GPS data set, the number of coordinate points used for path matching is reduced, and on one hand, resources occupied by storage and operation can be reduced; on the other hand, useless details can be shielded, and the influence of a large amount of tiny errors on the matching precision is avoided.
As a preferred embodiment, the starting point refers to the earliest collected coordinate point in the GPS data set, the ending point refers to the latest collected coordinate point in the GPS data set, and the turning point is obtained by:
taking any three coordinate points continuously acquired in time in the GPS data set as a first point, a second point and a third point;
taking the vectors of the first point and the second point as a first vector, and taking the vectors of the second point and the third point as a second vector;
and determining whether the second point is a turning point according to the size of an included angle between the first vector and the second vector.
The present embodiment provides definitions of a starting point, an end point, and a turning point, where whether a second point is a turning point is determined according to an included angle between a first vector and a second vector, which is implemented by specifically adopting the following method:
for three consecutive GPS data points (x) in a GPS data set in time first-to-last orderi，yi)、(xj，yj)、(xk，yk) Calculating a trajectory vectorAndwherein xp(p ═ i, j, k) denotes geographical longitude, yp(p ═ i, j, k) represents geographic latitude; then, converting the difference of the longitude and the latitude of the spherical surface into the difference of the length of the plane: latitude 1 ° translates to length (2 π R)/360, where R represents the radius of the earth; longitude 1 ° to length Wherein R represents the radius of the earth, lat represents the longitude value of the point, and Δ lon is the difference in latitude between the two points, namely 1 °; secondly, calculating the deflection angle of two track vectorsAndif it is(indicating not turning) and (indicating that the vehicle does not deviate from the current road and wR is the sum of all lane widths in one direction), the continuous change of the vehicle corner angle is small, the data of the three points can be approximately regarded as straight driving, and only the starting position point and the ending position point are reserved as key positions; if any one of the equations fails to be established, it represents that the vehicle position has largely changed, and all of the three points are regarded as key positions. Meanwhile, considering that the road may turn all the time but the turning angle is small during the road running, the method introduces the situationThe judgment of (1) is carried out,a velocity vector representing the current point k and the previous n-th point which has been extracted, and the establishment of the expression indicates that the continuous change of the vehicle turning angle is small, (x)-n，y-n) To (x)k，yk) The represented straight line vector can be used to approximate the path between two points, with acceptable error, and thus for (x)-n，y-n) To (x)k，yk) The dot data in between are compressed, and only the start and end coordinates are retained. All historical track points or all currently received real-time track points are processed through the modelJudging to obtain the key position
As a preferred embodiment, the performing track matching in the historical track dataset and/or the planned track dataset obtained from the online map service according to the coordinates of the plurality of key locations specifically includes:
according to the coordinates of the starting point and the coordinates of the end point, track indexing is carried out in the historical track data set;
when the number of the historical tracks obtained by indexing in the historical track data set is larger than a first set threshold value, carrying out track matching through a clustering model according to a plurality of key positions;
and when the number of the historical tracks obtained by indexing in the historical track data set is less than or equal to a first set threshold value or the historical track data is not matched through a clustering model, carrying out track matching on a planning track data set obtained from the online map service according to a plurality of key positions.
In this embodiment, in order to reduce the number of matching tracks and reduce the computation pressure, the index is first performed according to the starting point and the ending point, for example, from the middle wharf to the beijing road, all or part of the historical track data from the middle wharf to the beijing road is first extracted, and then matching is performed. Rather than taking the track data of the entire guangzhou city for comparison. Of course, in this embodiment, the history data may also be processed, so as to intercept a part of the trajectory as a matching object. For example, one historical track is from village to beijing road, the large dock in the route. In this embodiment, the track may be processed in advance, and the track from the wharf to the beijing road is extracted from the history track from the village to the beijing road as the starting point and the beijing road as the end point to be used as the matching target.
Specifically, in the matching process, since the acquisition time point and the position of the coordinates of the key position are different, the data needs to be aligned. The specific alignment is as follows:
for two paths, if the sampling time points are different, the distances of the obtained trajectory data points are greatly different, that is, the data are not aligned on the time line, so that the similarity between sequences cannot be effectively compared by using the conventional euclidean distance. Dynamic programming is used to align the two sequences. Given two sequences X ═ XiI ═ 1, 2.. multidot.m } and Y ═ YjJ 1, 2.. times.n.an alignment relationship w.w.is found between the two1，w2，...，wkWherein max (| X |, | Y |) is less than or equal to k and is less than or equal to | X | + | Y |. w is akIs of the form (i, j), where i denotes the i coordinate in X and j denotes the j coordinate in Y. The alignment path W must be from W1Starting with (1, 1) to wkEnd (m, n) to ensure that each coordinate in X and Y appears in W. In addition, i and j of W (i, j) in W must increase monotonically, i.e., for Wk(ii) and wk+1I ≦ i '≦ i +1 and j ≦ j' ≦ j + 1. And finally, calculating an alignment path sequence with the shortest distance through dynamic planning: d (i, j) ═ Dist (i, j) + min [ D (i-1, j), D (i, j-1), D (i-1, j-1)]. Thereby achieving alignment.
As a preferred embodiment, in order to increase the reliability of cluster matching, the number of the cluster models in this embodiment is at least two, and at least two cluster models are different types of models;
the track matching through the clustering model specifically includes:
if at least one clustering model is matched with the historical track data, taking the historical track data obtained by matching as an output result;
and if all the clustering models are not matched with the historical track data, judging that no result is matched.
In this embodiment, three optional models are described in detail, and the description content includes the process of establishing the model and selecting the parameters. The three models are: the system comprises a spectrum clustering sub-model, a Quick clusters sub-model and a DBSCAN clustering sub-model.
Constructing a spectral clustering sub-model: inputting a set Tr ═ { Tr) of all tracksi1, 2., N }, a trajectory distance matrix, and a cluster number k. Constructing a connection matrix W which is distance matrix and a degree matrix D, and initializing D to be an all-0 matrix with the same size as W. Sum rows i of W and assign D [ i, i]. Obtaining a Laplace matrix L, and enabling L to be D-W; obtaining a Laplace matrix D of the standard-0.5LD-0.5And calculating the eigenvector F of the first k smallest eigenvalues, constructing an eigenvector matrix F based on the eigenvector F, and obtaining a clustering result by using k-means.
Constructing a Quick Bundles clustering sub-model: inputting a set Tr ═ { Tr) of all tracksi1, 2,., N }, and an allowable distance difference δ. First, calculateAnd dflipped(s，m)＝d(s，mF)＝d(sFM), where s represents any one of the two streamlines being computed, m represents the mean of the Euclidean distances of the two streamlines, si、miRespectively representing the Euclidean distance mean value s of any corresponding point and the current point pair in the two tracksFAnd the sequence of the new track and the sequence of the original track are opposite. The minimum average direct inversion function MDF is then min (d)direct(s，m)，dflipped(s, m)). The sequence of trajectories used here is obtained by alignment warping. Secondly, extracting a historical track cluster Tr (Tr) to be matched from historical track data according to the actual positions of the starting point and the end point and by combining the coordinates distributed for the track starting point through the hot spot position of the geographic areai(i 1.., N), and assigning an allowable matching deviation distance δ for the journey according to the actual situation. First, cluster the history track { triTr in1Classified as class C1Then from tr2To trNEach time of calculationThe distance between the new tr and the MDF of the center of each current class of track, if the distance MDF is less than or equal to delta, the track is determined not to be classified into the new class, the class corresponding to the minimum MDF is selected, and the track is classified into the class until the historical track cluster { triAnd finishing all clustering to obtain a clustering result.
Constructing a DBSCAN clustering submodel: inputting a set Tr ═ { Tr) of all tracksi1, 2., N }, distance matrix, MinPts, r. First, all the tracks Tr ═ Tr are markediI 1, 2, N is not accessed and the sequence store class result C is initialized]. Traversing each track, and for each unaccessed track triFind its reachable track set Ri＝{trriWhere tr isriSatisfies dis [ tri，trri]< r. If len (Ri) ≧ MinPts, the temporary class sequence C _ tmp ═ is initialized]And r isiAdd C _ tmp. For each tr in RriIf trriIf not, marking the track as accessed and obtaining the reachable track set RrIf len (Rr) is not less than MinPts, then R is addedrAdding Ri(ii) a If trriNot in C, will trriAdd C _ tmp, otherwise mark trriFor noise, C _ tmp ═ trri]. And adding C _ tmp into C, judging whether len (C _ tmp) is 1 or not for each C _ tmp, and finally returning a class result.
Selecting spectral clustering parameters: the distance matrix D is input, initialized K ═ 1, 2.., m }, I, J is the empty set, P [0 · is input](0, 0.., 0). Find (i, j) ∈ arg max Dpq(P, q ∈ K), let P (1) ═ I, I ═ I }, J ═ K- { J }. For t 2 to t n, find (i, j) ∈ arg min Dpq(p ∈ I, q ∈ J), let p (t) ═ J, I ═ I + { J }, J ═ J- { J }. The cycle is over to get V ═ Vij|Vp(i)p(j)I is more than or equal to 1, j is less than or equal to n, and the number of the maximum non-0 subarrays of V is the parameter k of the spectral clustering.
And (3) selecting Quick Bundles clustering parameters: computingAndfurther obtainWherein, TjRepresenting the track, M representing the number of global clusters, nkDenotes the kth class CkNumber of tracks in, Cp(k)Is represented by CkCenter locus of (1), CpThe central track representing the overall track, i.e. the central track with the cluster number of 1, | Tj-Cp(k)||2Represents class CkDistance between each track and the center-like track, | Cp-Cp(k)||2Represents class CkDistance from the overall trajectory. And searching the maximum value of WB-INDEX by testing different central tracks to obtain the optimal clustering number M.
DBSCAN clustering parameter selection: there are two parameters of DBSCAN, MinPts and the clustering radius r. According to the guiding principle of the DBSCAN algorithm, combining the data types of the coordinates and the paths processed by the method, initializing MinPts to 4; sorting the values of each row of the distance matrix D from small to large to obtainAt this time, the value in i row and j column indicates the jth distance closest to the object i. Assuming that the rime distribution is obeyed, the rime distribution parameter for the jth distance is estimated using maximum likelihood estimation:the expected value of lambda is the r to be solved. After r is obtained, counting the number p of objects in each communication domain of each point iiThen expect itFor updating the value of MinPts.
In the embodiment, the real-time track data and the historical track data are respectively subjected to cluster analysis through three submodels to obtain three sub-results, if the current real-time track can be included in the historical tracks in all or part of the sub-results, the running route from the starting point to the current position of the vehicle is not deviated from the normal track, and the matching is judged to be successful; on the contrary, if the current real-time trajectory cannot be classified into the historical trajectories in the three types of sub-results, namely the current real-time trajectory is classified into one type separately, the matching is judged to be failed.
As a preferred embodiment, the method further comprises the following steps:
after the deviation of the driving route from the track is judged, accumulating the times of deviation of the driving route from the track;
and taking the point deviating from the track as a new starting point, and performing track indexing in the historical track data set according to the coordinates of the starting point and the coordinates of the end point again so as to perform new track matching.
The method also comprises the following steps:
and when the number of times of the deviation of the driving route from the track is greater than a second set threshold value, generating alarm information.
The implementation accumulates the times of the deviation of the driving route from the track, and the track is matched again according to the new starting point coordinate when the deviation of the track occurs each time. If the situation of deviating the track continues to appear, the vehicle running track is proved to be problematic. Otherwise, the situation of blocking the road and the like may only occur. The alarm accuracy can be improved.
As a preferred embodiment, the method further comprises the following steps:
and performing de-diversification processing on the coordinates of the starting point and the coordinates of the end point.
For an area hotspot with a large area, such as a large mall, a park and the like, the occupied area is large, so that the positions of the upper and lower vehicles are too dispersed, and are often uniformly distributed nearby rather than on a single coordinate, which adversely affects the judgment effect of the model. The scattered points are processed in a centralized mode, the mass center of the geographic area is used as the coordinate position of the area, the reasonability of the large area clustering effect can be enhanced, the reflection of the track carried by the track to the traffic hot Point is fully utilized, and the traffic hot Point accords with the position distribution of POI (Point of Interest) in the area, wherein the POI is the POI of Interest.
Using a geohash encoding process on the original geographic location coordinates in the de-diversification operation, the latitude and longitude can be converted into a string, and in most cases, the more the string prefixes match, the closer the distance. The method has the advantages that the method is high in de-diversification processing speed by using the geohash, and the integral clustering number does not need to be preset like the traditional clustering method. Furthermore, the method is very suitable for locally equalized data such as POI coordinates. After the coding conversion, the geographic area can be partitioned, and an index is assigned to each block area, so that the effects of coordinate concentration and de-decentralization of scattered coordinates near the large POI are achieved.
As a preferred embodiment, before performing the track matching, the method further comprises the following steps:
and aligning the two groups of track data which are matched with each other in sequence.
Aligning the sequences may reduce the difficulty of matching.
Referring to fig. 2, the present embodiment provides a preferred embodiment comprising the steps of: firstly, acquiring historical track point data and real-time track point data; then processing an OD coordinate cluster, wherein OD is a starting point and an end point; extracting key track points; dynamically adjusting the path index; when the historical tracks are few and the clustering track cluster matching fails, inquiring and acquiring a path set by an order OD shortest path through online map service; the method comprises the steps that clustering track cluster matching is carried out only when historical tracks are sufficient, and if the clustering track cluster matching is successful, the fact that a route is finished and a destination is reached is indicated; if the matching is successful in the set of paths obtained from the online mapping service, the completion of the formation is also declared to reach the predetermined destination. And if the historical track and the track set obtained from the map service do not obtain a matching result, indicating that the driving track is abnormal, and triggering an alarm. After the warning is triggered, if the warning is continuously triggered for many times, the driving track is judged to be abnormal, and the warning is reported. And after each abnormal driving track triggers warning, recording the current warning position as a new monitoring starting point, and re-executing track matching.
The embodiment discloses a vehicle track deviation detection method which comprises a first step, a second step and a ninth step.
The method comprises the following steps: and constructing a key position judgment model, extracting a key path and compressing track data. For any three consecutive GPS data points (x) in the datai，yi)、(xj，yj)、(xk，yk) Wherein x isp(p ═ i, j, k) denotes geographical longitude, yp(p ═ i, j, k) denotes the geographic latitude, and two trajectory vectors can be formedAndconverting the spherical longitude and latitude difference into a plane length difference: latitude 1 ° translates to length (2 π R)/360, where R represents the radius of the earth; longitude 1 ° to lengthWhere R represents the earth's radius, lat represents the longitude value of the point, and Δ lon is the difference in latitude between the two points, i.e., 1 °. Calculating the deflection angle of two track vectorsAndif it isAnd is(wRAll lane widths in one direction and) simultaneously(The velocity vector representing the current point k and the first n point that has been extracted) indicates that the continuous change of the vehicle angle is small, (x)-nY-n) to (x)k，yk) The straight line vector represented is used to describe the position variation error between two points is acceptable, and thus is for the current (x)-n，y-n) To (x)k，yk) The point data of (a) is compressed, only the coordinates (x) are retainedk，yk). All historical track points or all currently received real-time track points are processed through the modelJudging to obtain the key positionReferring to fig. 3 and 4, asterisks in fig. 3 are GPS points collected on the trajectory before the key position is not extracted, and fig. 4 is coordinate points left after the key position is extracted.
Step two: and (4) de-dispersing the coordinates of the starting point and the end point in the same area. And carrying out geohash distribution on the original geographic position coordinates by using a tree structure based on the geographic position, dividing the longitude and latitude values into two halves, distributing the next binary bit according to the left and right positions of the segmentation section and the like after each halving, and distributing 0 if the target longitude and latitude is in the left section and distributing 1 if the target longitude and latitude is in the right section. And combining the left and right longitude and latitude codes: the longitude code is arranged in an odd number position, and the latitude code is arranged in an even number position, so that the geographic position code is obtained.
Step three: the sequences are aligned. Considering that if the sampling time points of the same two tracks are different, the distances of the obtained track data points are greatly different, and therefore, the similarity between the sequences cannot be effectively compared by using the traditional Euclidean method. Dynamic programming is used to align the two sequences. Given two sequences X ═ XiI ═ 1, 2.. multidot.m } and Y ═ YjJ 1, 2.. times.n.an alignment relationship w.w.is found between the two1，w2，...，wkWherein max (| X |, | Y |) is less than or equal to k and is less than or equal to | X | + | Y |. w is akIs of the form (i, j), where i denotes the i coordinate in X and j denotes the j coordinate in Y. The alignment path W must be from W1Starting with (1, 1) to wkEnd-of-m, n to ensure that each coordinate in X and Y is atAnd is present in W. In addition, i and j of W (i, j) in W must increase monotonically, i.e., for Wk(ii) and wk+1I ≦ i '≦ i +1 and j ≦ j' ≦ j + 1. And finally, calculating an alignment path sequence with the shortest distance through dynamic planning: d (i, j) ═ Dist (i, j) + min [ D (i-1, j, Di, j-1, Di-1, j-1]。
Step four: and constructing an integrated clustering model, and clustering and matching the real-time track and the historical track clusters. The fourth step includes the following three steps.
Constructing a spectral clustering sub-model: inputting a set Tr ═ { Tr) of all tracksi1, 2., N }, a trajectory distance matrix, and a cluster number k. Constructing a connection matrix W which is distance matrix and a degree matrix D, and initializing D to be an all-0 matrix with the same size as W. Sum rows i of W and assign D [ i, i]. Obtaining a Laplace matrix L, and enabling L to be D-W; obtaining a Laplace matrix D of a standard-0.5LD-0.5Calculating a feature vector F of the first k minimum feature values, constructing a feature matrix F based on the F, and obtaining a clustering result C ═ C by using k-meansiI 1, 2,.., k }, and finally obtaining a clustering result a { a ═ a }i1, 2, k, wherein ai＝{pi，pi∈Ci}。
Constructing a Quick Bundles clustering sub-model: inputting a set Tr ═ { Tr) of all tracksiI 1, 2.., N }, allowing for the distance difference δ. First, calculateAnd dflipped(s，m)＝d(s，mF)＝d(sFM), where s represents any one of the two streamlines being computed, m represents the mean of the Euclidean distances of the two streamlines, si、miRespectively representing the Euclidean distance mean value s of any corresponding point and the current point pair in the two tracksFAnd the sequence of the new track and the sequence of the original track are opposite. The minimum average direct inversion function MDF is then min (d)direct(s，m)，dflipped(s, m)). Sequence of trajectories as used hereinObtained by aligning and regulating. Secondly, according to the actual positions of the starting point and the end point, combining the step two, by using the coordinates distributed for the starting point of the track through the hot spot position of the geographic area, extracting a historical track cluster Tr ═ Tr to be matched from the historical track datai(i 1.., N), and assigning an allowable matching deviation distance δ for the journey according to the actual situation. First, cluster the history track { triTr in1Classified as class C1Then from tr2To trNCalculating the MDF distance between each new tr and the center of each current class of track, if the MDF distance is less than or equal to 6, determining that the track can not be classified into a new class, selecting the class corresponding to the minimum MDF, and classifying the track into the class until the historical track cluster { triFinishing all clustering to obtain a clustering result C ═ Ci，i＝1，2，...，k}。
Constructing a DBSCAN clustering submodel: inputting a set Tr ═ { Tr) of all tracksi1, 2., N }, distance matrix, MinPts, r. First, all the tracks Tr ═ Tr are markediI 1, 2, N is not accessed and the sequence store class result C is initialized]. Traversing each track, and for each unaccessed track triFind its reachable track set Ri＝{trri{, where tr isriSatisfies dis [ tri，trri]< r. If len (Ri) ≧ MinPts, the temporary class sequence C _ tmp ═ is initialized]And r isiAdd C _ tmp. For each tr in RriIf trriIf not, marking the track as accessed and obtaining the reachable track set RrIf len (Rr) is not less than MinPts, then R is addedrAdding Ri(ii) a If trriNot in C, will trriAdd C _ tmp, otherwise mark trriFor noise, C _ tmp ═ trri]. Adding C _ tmp into C, judging whether len (C _ tmp) is 1 or not for each C _ tmp, and finally returning a class result C ═ Ci，i＝1，2，...，k}。
Step five: calculating the parameters of the three types of sub models, wherein the step five comprises the following three steps:
selecting spectral clustering parameters: inputting the distance matrix D, initializing K ═ 1,2, a., m }, I, J is an empty set, P [0 ]](0, 0.., 0). Find (i, j) ∈ arg max Dpq(P, q ∈ K), let P (1) ═ I, I ═ I }, J ═ K- { J }. For t 2 to t n, find (i, j) ∈ arg min Dpq(p ∈ I, q ∈ J), let p (t) ═ J, I ═ I + { J }, J ═ J- { J }. The cycle is over to get V ═ Vij|Vp(i)p(j)，1≤i，j≤n}。
And (3) selecting Quick Bundles clustering parameters: computingAndfurther obtainWherein, TjRepresenting the track, M representing the number of global clusters, nkDenotes the kth class CkNumber of tracks in, Cp(k)Is represented by CkCenter locus of (1), CpThe central track representing the overall track, i.e. the central track with the clustering number of 1, | Tj-Cp(k)||2Represents class CkDistance between each track and the center-like track, | Cp-Cp(k)||2Represents class CkDistance from the overall trajectory. And searching the maximum value of WB-INDEX by testing different central tracks to obtain the optimal clustering number M.
DBSCAN clustering parameter selection: let MinPts be 4, sort the values of each row of the distance matrix D from small to large to obtainAt this time, the value in i row and j column indicates the jth distance closest to the object i. Assuming adherence to the rime distribution, the estimate is made using the rime distribution parameter for the most entire jth distance with maximum likelihood estimate:the expected value of lambda is the r to be solved. After r is obtained, counting the object number in each communication domain of each point iNumber piThen expect itThe value of MinPts is obtained.
Step six: and respectively carrying out three clustering analyses on the real-time track data and the historical track data to obtain three sub-result classes, searching a pair of track bundles with the largest DTW distance in each result class, and calculating the deviation distance mean value of points in each class according to the number of track beam points.
Step seven: and for the cluster analysis result data obtained in the step six, if the MDF distance between the real-time track and any track set is larger than delta at a certain time t, namely the real-time track deviates from the conventional track set in the historical data, determining that the vehicle driving track deviates, triggering an alarm, updating the warning position to be a new starting point, and continuously judging the track abnormal condition between the new starting point and the initial end point.
Step eight: for the case that the algorithm matching path judgment fails due to the deviation of the seven paths in the step, acquiring a track set between a starting point and an end point in a shortest path mode by using an online map service for expanding a data set: if the matching state is good and the preset destination is finally reached, judging that the travel is normal; if the driving state is still deviated from the normal path in the matching process at the moment and is too large, a warning is triggered.
Step nine: recording the warning times in the seventh step and the eighth step in the driving process, if multiple warnings are triggered according to a certain threshold value, judging that the travel is abnormal, and immediately reporting abnormal driving information; and if the real-time track is not deviated from the conventional track class in the historical data or the alarm frequency is less than the threshold value when the vehicle reaches the terminal, judging that the travel state is normal.
Step ten: the driving state of the 3 network appointment vehicles is obtained by referring to fig. 5 to 8, wherein fig. 5 shows the history track, and fig. 6, 7 and 8 show the driving route of the test vehicle. Among them, fig. 6 and 7 show temporary abnormalities caused by replanning routes during driving, and fig. 8 shows serious trajectory deviation during driving, and triggers an alarm many times.
The embodiment discloses a vehicle track deviation detection system, which includes:
a memory for storing a program;
and the processor is used for loading the program to execute the track deviation detection method.
The present embodiment discloses a storage medium storing a program executed by a processor to implement the vehicle trajectory deviation detecting method.
The above-described system and storage medium embodiments can be used to implement method embodiments and achieve the same technical result.
The step numbers in the above method embodiments are set for convenience of illustration only, the order between the steps is not limited at all, and the execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A vehicle trajectory deviation detection method is characterized in that: the method comprises the following steps:
extracting a starting point, a terminal point and a turning point from a GPS data set of a driving route as key positions;
according to the coordinates of the key positions, carrying out track matching in a historical track data set and/or a planning track data set acquired from an online map service;
when historical track data or planning track data are matched, storing tracks described by a plurality of key positions into a historical track data set;
and when the historical track data or the planning track data are not matched, judging that the driving route deviates from the track.
2. The vehicle track deviation detection method according to claim 1, characterized in that: the starting point refers to the earliest collected coordinate point in the GPS data set, the ending point refers to the latest collected coordinate point in the GPS data set, and the turning point is obtained by:
taking any three coordinate points continuously acquired in time in the GPS data set as a first point, a second point and a third point;
taking the vectors of the first point and the second point as a first vector, and taking the vectors of the second point and the third point as a second vector;
and determining whether the second point is a turning point according to the size of an included angle between the first vector and the second vector.
3. The vehicle track deviation detection method according to claim 1, characterized in that: the track matching is performed in a historical track data set and/or a planned track data set acquired from an online map service according to the coordinates of the plurality of key positions, and the track matching specifically includes:
according to the coordinates of the starting point and the coordinates of the end point, track indexing is carried out in the historical track data set;
when the number of the historical tracks obtained by indexing in the historical track data set is larger than a first set threshold value, carrying out track matching through a clustering model according to a plurality of key positions;
and when the number of the historical tracks obtained by indexing in the historical track data set is less than or equal to a first set threshold value or the historical track data is not matched through a clustering model, carrying out track matching on a planning track data set obtained from the online map service according to a plurality of key positions.
4. A vehicle trajectory deviation detection method according to claim 3, characterized in that: the number of the clustering models is at least two, and at least two clustering models are different types of models;
the track matching through the clustering model specifically includes:
if at least one clustering model is matched with the historical track data, taking the historical track data obtained by matching as an output result;
and if all the clustering models are not matched with the historical track data, judging that no result is matched.
5. A vehicle trajectory deviation detection method according to claim 3, characterized in that: further comprising the steps of:
after the deviation of the driving route from the track is judged, accumulating the times of deviation of the driving route from the track;
and taking the point deviating from the track as a new starting point, and performing track indexing in the historical track data set according to the coordinates of the starting point and the coordinates of the end point again so as to perform new track matching.
6. The vehicle track deviation detection method according to claim 5, characterized in that: further comprising the steps of:
and when the number of times of the deviation of the driving route from the track is greater than a second set threshold value, generating alarm information.
7. The vehicle track deviation detection method according to any one of claims 1 to 6, characterized in that: further comprising the steps of:
and performing de-diversification processing on the coordinates of the starting point and the coordinates of the end point.
8. The vehicle track deviation detection method according to any one of claims 1 to 6, characterized in that: before the track matching is carried out, the method further comprises the following steps:
and aligning the two groups of track data which are matched with each other in sequence.
9. A vehicle trajectory deviation detection system characterized by: the method comprises the following steps:
a memory for storing a program;
a processor for loading the program to perform the trajectory deviation detection method according to any one of claims 1 to 8.
10. A storage medium storing a program, characterized in that: the program is executed by a processor to implement the vehicle trajectory deviation detection method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910803343.4A CN110609881A (en) | 2019-08-28 | 2019-08-28 | Vehicle trajectory deviation detection method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910803343.4A CN110609881A (en) | 2019-08-28 | 2019-08-28 | Vehicle trajectory deviation detection method, system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110609881A true CN110609881A (en) | 2019-12-24 |
Family
ID=68891007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910803343.4A Pending CN110609881A (en) | 2019-08-28 | 2019-08-28 | Vehicle trajectory deviation detection method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110609881A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111858790A (en) * | 2020-04-03 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | Detour reminding method and device, electronic equipment and medium |
| CN111882873A (en) * | 2020-07-22 | 2020-11-03 | 平安国际智慧城市科技股份有限公司 | Track anomaly detection method, device, equipment and medium |
| CN111913194A (en) * | 2020-08-07 | 2020-11-10 | 上海联适导航技术有限公司 | Pile position guiding method, device, equipment and computer readable storage medium |
| CN112348265A (en) * | 2020-11-10 | 2021-02-09 | 交控科技股份有限公司 | Feasible path mining method and device under monitoring scene |
| CN112418471A (en) * | 2021-01-25 | 2021-02-26 | 南京领行科技股份有限公司 | Network appointment vehicle driving monitoring method and device, electronic equipment and storage medium |
| CN112991749A (en) * | 2021-05-13 | 2021-06-18 | 腾讯科技（深圳）有限公司 | Vehicle violation identification method and device, medium and electronic equipment |
| CN112985440A (en) * | 2021-02-20 | 2021-06-18 | 北京嘀嘀无限科技发展有限公司 | Method, device, storage medium and program product for detecting deviation of driving track |
| CN112991806A (en) * | 2021-02-18 | 2021-06-18 | 安徽中科美络信息技术有限公司 | Vehicle track monitoring method and device |
| CN113256982A (en) * | 2021-06-18 | 2021-08-13 | 城云科技（中国）有限公司 | Vehicle illegal driving detection method and device, computer equipment and storage medium |
| CN113377845A (en) * | 2021-08-12 | 2021-09-10 | 成都天府市民云服务有限公司 | Intelligent old-age care data processing method and system based on big data |
| CN113607170A (en) * | 2021-07-31 | 2021-11-05 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Real-time detection method for track deviation behavior of air and sea target |
| CN113847925A (en) * | 2021-09-15 | 2021-12-28 | 北京中交兴路车联网科技有限公司 | Method, device, equipment and medium for detecting vehicle yaw based on track data |
| CN115480275A (en) * | 2022-09-15 | 2022-12-16 | 中华人民共和国广东海事局 | Motion state acquisition method and device, computer equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120296560A1 (en) * | 2011-05-19 | 2012-11-22 | Microsoft Corporation | Inferring a Behavioral State of a Vehicle |
| CN105739495A (en) * | 2016-01-29 | 2016-07-06 | 大连楼兰科技股份有限公司 | Driving path planning method and device and automatic steering system |
| CN107766808A (en) * | 2017-09-30 | 2018-03-06 | 北京泓达九通科技发展有限公司 | The method and system that Vehicle Object motion track clusters in road network space |
| CN108133611A (en) * | 2016-12-01 | 2018-06-08 | 中兴通讯股份有限公司 | Vehicle driving trace monitoring method and system |
| CN109784571A (en) * | 2019-01-24 | 2019-05-21 | 郑鹏 | A kind of wisdom travel system and terminal |
| CN109784371A (en) * | 2018-12-14 | 2019-05-21 | 北京三快在线科技有限公司 | Net about vehicle monitoring and managing method, device and storage medium |
| CN110134126A (en) * | 2019-05-15 | 2019-08-16 | 百度在线网络技术（北京）有限公司 | Path matching method, apparatus, equipment and medium |
-
2019
- 2019-08-28 CN CN201910803343.4A patent/CN110609881A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120296560A1 (en) * | 2011-05-19 | 2012-11-22 | Microsoft Corporation | Inferring a Behavioral State of a Vehicle |
| CN105739495A (en) * | 2016-01-29 | 2016-07-06 | 大连楼兰科技股份有限公司 | Driving path planning method and device and automatic steering system |
| CN108133611A (en) * | 2016-12-01 | 2018-06-08 | 中兴通讯股份有限公司 | Vehicle driving trace monitoring method and system |
| CN107766808A (en) * | 2017-09-30 | 2018-03-06 | 北京泓达九通科技发展有限公司 | The method and system that Vehicle Object motion track clusters in road network space |
| CN109784371A (en) * | 2018-12-14 | 2019-05-21 | 北京三快在线科技有限公司 | Net about vehicle monitoring and managing method, device and storage medium |
| CN109784571A (en) * | 2019-01-24 | 2019-05-21 | 郑鹏 | A kind of wisdom travel system and terminal |
| CN110134126A (en) * | 2019-05-15 | 2019-08-16 | 百度在线网络技术（北京）有限公司 | Path matching method, apparatus, equipment and medium |
Non-Patent Citations (2)
| Title |
|---|
| 杨洁: "基于移动大数据的轨迹匹配算法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》 * |
| 胡圆等: "基于密度聚类的出租车异常轨迹检测", 《计算机与现代化》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111858790A (en) * | 2020-04-03 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | Detour reminding method and device, electronic equipment and medium |
| CN111882873B (en) * | 2020-07-22 | 2022-01-28 | 平安国际智慧城市科技股份有限公司 | Track anomaly detection method, device, equipment and medium |
| CN111882873A (en) * | 2020-07-22 | 2020-11-03 | 平安国际智慧城市科技股份有限公司 | Track anomaly detection method, device, equipment and medium |
| CN111913194A (en) * | 2020-08-07 | 2020-11-10 | 上海联适导航技术有限公司 | Pile position guiding method, device, equipment and computer readable storage medium |
| CN112348265A (en) * | 2020-11-10 | 2021-02-09 | 交控科技股份有限公司 | Feasible path mining method and device under monitoring scene |
| CN112418471A (en) * | 2021-01-25 | 2021-02-26 | 南京领行科技股份有限公司 | Network appointment vehicle driving monitoring method and device, electronic equipment and storage medium |
| CN112991806A (en) * | 2021-02-18 | 2021-06-18 | 安徽中科美络信息技术有限公司 | Vehicle track monitoring method and device |
| CN112985440A (en) * | 2021-02-20 | 2021-06-18 | 北京嘀嘀无限科技发展有限公司 | Method, device, storage medium and program product for detecting deviation of driving track |
| CN112985440B (en) * | 2021-02-20 | 2024-02-23 | 北京嘀嘀无限科技发展有限公司 | Method, device, storage medium and program product for detecting travel track deviation |
| CN112991749A (en) * | 2021-05-13 | 2021-06-18 | 腾讯科技（深圳）有限公司 | Vehicle violation identification method and device, medium and electronic equipment |
| CN112991749B (en) * | 2021-05-13 | 2021-07-20 | 腾讯科技（深圳）有限公司 | Vehicle violation identification method and device, medium and electronic equipment |
| CN113256982B (en) * | 2021-06-18 | 2021-09-28 | 城云科技（中国）有限公司 | Vehicle illegal driving detection method and device, computer equipment and storage medium |
| CN113256982A (en) * | 2021-06-18 | 2021-08-13 | 城云科技（中国）有限公司 | Vehicle illegal driving detection method and device, computer equipment and storage medium |
| CN113607170A (en) * | 2021-07-31 | 2021-11-05 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Real-time detection method for track deviation behavior of air and sea target |
| CN113607170B (en) * | 2021-07-31 | 2023-12-12 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Real-time detection method for deviation behavior of navigation path of air-sea target |
| CN113377845A (en) * | 2021-08-12 | 2021-09-10 | 成都天府市民云服务有限公司 | Intelligent old-age care data processing method and system based on big data |
| CN113377845B (en) * | 2021-08-12 | 2021-10-29 | 成都天府市民云服务有限公司 | Intelligent old-age care data processing method and system based on big data |
| CN113847925A (en) * | 2021-09-15 | 2021-12-28 | 北京中交兴路车联网科技有限公司 | Method, device, equipment and medium for detecting vehicle yaw based on track data |
| CN115480275A (en) * | 2022-09-15 | 2022-12-16 | 中华人民共和国广东海事局 | Motion state acquisition method and device, computer equipment and storage medium |
| CN115480275B (en) * | 2022-09-15 | 2023-08-08 | 中华人民共和国广东海事局 | Motion state acquisition method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110609881A (en) | Vehicle trajectory deviation detection method, system and storage medium | |
| CN108519094B (en) | Local path planning method and cloud processing terminal | |
| CN109405839B (en) | Traffic network off-line map matching algorithm based on multiple paths | |
| US20200019815A1 (en) | Clustering for k-anonymity in location trajectory data | |
| CN111539454B (en) | Vehicle track clustering method and system based on meta-learning | |
| US20200019585A1 (en) | Triangulation for k-anonymity in location trajectory data | |
| CN114005280B (en) | Vehicle track prediction method based on uncertainty estimation | |
| CN109459047A (en) | Vehicle high-precision location data is accurately matched with navigation map and lane change behavior detection method | |
| CN109544914B (en) | Shared bicycle reverse behavior identification method based on historical GPS track | |
| CN111209457B (en) | Target typical activity pattern deviation warning method | |
| CN110516702B (en) | Discrete path planning method based on streaming data | |
| CN113587944B (en) | Quasi-real-time vehicle driving route generation method, system and equipment | |
| CN106441316A (en) | Single-point road network matching method based on historical data | |
| CN106940929B (en) | Traffic data prediction method and device | |
| CN116010838A (en) | Vehicle track clustering method integrating density value and K-means algorithm | |
| CN115164922A (en) | Path planning method, system, equipment and storage medium | |
| CN111123333B (en) | Vehicle track positioning method fusing bayonet and GPS data | |
| CN112633812B (en) | Track segmentation method, device, equipment and storage medium for freight vehicle | |
| CN113932821A (en) | Track map matching method based on continuous window average direction features | |
| CN116244391A (en) | Method for extracting typical array position of massive track targets | |
| CN116467615A (en) | Clustering method and device for vehicle tracks, storage medium and electronic device | |
| CN110793531A (en) | Road matching method and device and readable storage medium | |
| CN116484244A (en) | Automatic driving accident occurrence mechanism analysis method based on clustering model | |
| CN115662124A (en) | GPS track data road section flow matching method based on network coding | |
| CN115184951A (en) | Method, equipment and medium for detecting pedestrian and vehicle flow in real time based on laser radar |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191224 |
|
| RJ01 | Rejection of invention patent application after publication |