CN105957342B - Track grade road plotting method and system based on crowdsourcing space-time big data - Google Patents
Track grade road plotting method and system based on crowdsourcing space-time big data Download PDFInfo
- Publication number
- CN105957342B CN105957342B CN201610370700.9A CN201610370700A CN105957342B CN 105957342 B CN105957342 B CN 105957342B CN 201610370700 A CN201610370700 A CN 201610370700A CN 105957342 B CN105957342 B CN 105957342B
- Authority
- CN
- China
- Prior art keywords
- lanes
- lane
- track
- rectangular window
- nlane
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 239000013598 vector Substances 0.000 claims abstract description 68
- 238000001514 detection method Methods 0.000 claims abstract description 43
- 238000010276 construction Methods 0.000 claims abstract description 13
- 230000004927 fusion Effects 0.000 claims abstract description 6
- 238000005457 optimization Methods 0.000 claims description 33
- 238000012937 correction Methods 0.000 claims description 32
- 238000013210 evaluation model Methods 0.000 claims description 23
- 230000003993 interaction Effects 0.000 claims description 21
- 238000013519 translation Methods 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 15
- 230000001360 synchronised effect Effects 0.000 claims description 15
- 238000013507 mapping Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 13
- 238000005259 measurement Methods 0.000 description 10
- 239000000203 mixture Substances 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 239000000523 sample Substances 0.000 description 9
- 238000011160 research Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000008384 inner phase Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
Landscapes
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention provides a kind of track grade road plotting method based on crowdsourcing space-time big data, similarity evaluating model including establishing track vector, growth clustering method progress track based on fusion Heuristics is preferred, and structure Gauss constrains mixed model, and uses EM Algorithm for Solving model parameters；Lane information is detected, obtains the first result of detection of road section track quantity；Based on road construction rule, first result of detection is modified；According to revised track quantity, lane center is modified according to adjacencies.Present invention reduces the cost for obtaining the fine road information in city, and detection method is simple, easy to implement.
Description
Technical Field
The invention relates to high-precision lane-level road mapping of crowdsourcing space-time big data, and belongs to the field of geographic information systems and intelligent traffic research.
Background
The high-precision road map is the blood and soul of future automatic driving, and the lane-level fine road information is a key component for constructing the high-precision road map. At present, researchers propose to extract lane lines from high-resolution remote sensing images based on a visual method, or to obtain road detail information by using high-precision laser point cloud data, and to extract road pavement information (road boundary lines, lane number, lane center lines) from a large amount of high-precision GPS track data collected by a measuring vehicle. Rogerset al, (1999) was one of the first researchers trying to extract road center lines and lane boundary lines using space-time DGPS (differential global positioning system) trajectory data. Then, on the basis of research of Rogers et al, the method for acquiring road information by using space-time GPS track data gradually develops into an end-to-end mode. This end-to-end mode of road information acquisition can be summarized as the following processes: the method comprises the steps of optimizing DGPS track data, matching the DGPS track data with existing map data, fitting a spline curve to a road center line, and finally extracting lane information and refining the geometric structure of an intersection through a clustering method. John Krumm proposes a road information acquisition mode which is separated from an original map, and the mode firstly adopts a track classification and fusion method to extract road level information from a large amount of DGPS track data, and then utilizes a Gaussian mixture model to extract lane information from a large amount of track data belonging to each road section. However, these approaches for acquiring lane-level fine road information all have the disadvantages of high data acquisition cost, long acquisition time, slow update speed, complex data processing, and the like.
With the rapid development of sensor technology, wireless communication and network technology, people are sensors, and people can generate a large amount of large data of space-time trajectories and contain rich fine road information and human behavior and activity information when going out. The acquisition of the trajectory data is gradually developed from acquisition of a measuring vehicle or a professional by a professional department into a form of recording the travel trajectory of the non-professional, and the acquisition of the data starts to be converted into a crowdsourcing mode. The vehicle-mounted trajectory data (crowd-sourced data) in the crowd-sourced mode is undoubtedly the best data source that can provide lane-level road information extraction at present. Compared with the existing taxi data, the vehicle-mounted track data acquired by the crowdsourcing mode belongs to big data (the big data refers to a data set which cannot be captured, managed and processed by a conventional software tool within a bearable time range, and is information assets which are high in mass, high in growth rate and diversified by needing a new processing mode and have stronger decision-making power, insight discovery power and flow optimization capability). At present, national scholars Tang furing et al (2015, 2016) propose that urban lane-level road information including lane number, lane steering and lane center line is extracted by using low-precision GPS track data, however, how to use crowd-sourced data to develop lane-level fine road mapping is a difficult problem for scientists all over the world.
Disclosure of Invention
On the basis of the research, the invention provides a novel technical scheme of high-precision lane-level road mapping (high-quality track data filtering and high-precision road information extraction) based on crowdsourcing spatio-temporal big data.
The technical scheme of the invention provides a lane level road mapping method based on crowdsourcing space-time big data, which comprises the following steps,
step 1, establishing a similarity evaluation model of a track vector, and setting v a And v b Are two different trajectory vectors, the similarity evaluation model is as follows,
wherein,representing the similarity value between vectors, e being a natural base number, omega 1 And ω 2 Respectively represent distance factors diff Hd And angle factor diff θab And ω is 1 +ω 2 =1; distance factor diff Hd And angle factor diff θab Respectively represent vectors v a And v b The distance difference and the angle difference of (a);
step 2, carrying out track optimization based on a growth clustering method of fusion empirical knowledge, wherein the track optimization comprises the step of determining a weight value omega of a similarity evaluation model according to the existing high-precision GPS track data and the synchronous low-precision GPS track data 1 And ω 2 Extracting prior knowledge of track optimization, and optimizing data by adopting a growth clustering mode based on the similarity between crowdsourcing track data;
step 3, constructing a Gaussian constraint mixed model, and solving model parameters by using an EM (effective electromagnetic) algorithm; the gaussian-constrained hybrid model is defined as follows,
wherein p (x) is represented as the comprehensive probability value of the Gaussian constraint mixed model, x represents a sample value to be calculated, and x represents the vertical coordinate value of the vertical projection of the track point in the judgment window on the longitudinal section when the lane calculation is carried out; k is the number of Gaussian components, each corresponding to a lane; omega j Is the weight of the jth Gaussian component and corresponds to the traffic flow of the lane; parameter mu 1 …μ k Is the average of the trajectories in each Gaussian component, equal to the centerline, μ, of each lane j Represents mu 1 …μ k Any value within the parameters, j =1,2, \8230;, k; σ is the standard deviation of the trajectories in each Gaussian component;
the number k of Gaussian components in the Gaussian constraint mixed model is obtained by calculating the value of the structural risk model and determining k according to the principle that the value of the structural risk model is minimum;
step 4, detecting lane information according to the result obtained in the step 3 to obtain a primary detection result of the number of lanes of the road section; the implementation mode is as follows,
setting a given group of Intersection intersections by taking all tracks on the same road section as an extraction unit 1 Intersection to Intersection interaction 2 Track set A of T From the track set A T Starting from one end of the window, constructing a moving rectangular window, wherein the long side of the moving rectangular window is parallel to the central lines covering all tracks at present, the wide side of the moving rectangular window is perpendicular to the central lines covering all tracks at present, the central line of the long side of the rectangular window is perpendicular to the central lines of all track data covered by the rectangular window, and the central line of the wide side of the rectangular window is superposed with the central line of the track data covered by the rectangular window;
moving rectangular window from trajectory set A T The method comprises the steps of starting translation according to the length of the long edge of a rectangular window, sequentially utilizing a Gaussian constraint hybrid model to detect the number of lanes and the center line of the lanes of a road section covered in each rectangular window, projecting all track points in the moving rectangular window onto the center line of the long edge of the rectangular window according to the moving rectangular window to obtain a projected track data set X = (X =) 1 ,x 2 ,…,x N ) T =1,2,3, \ 8230;, N, wherein x t Expressing the longitudinal coordinate value of the t-th track point after projection, wherein N is the number of the track points participating in projection; substituting the track data set X into a Gaussian constraint hybrid model, and extracting the number of lanes and the center line of the lanes of a road section in a rectangular window; suppose that Intersection interaction is driven from 1 Intersection to Intersection interaction 2 Track set A of T The rectangular window is translated for one time, and the number of lanes determined by each translation is recorded as Nlane f L, as a primary detection result of the number of lanes of the road section;
step 5, correcting the primary detection result based on the road construction rule according to the primary detection result of the number of lanes of the road section acquired in the step 4;
and 6, correcting the lane center line according to the adjacent condition according to the corrected lane number obtained in the step 5.
And, the primary detection result is corrected in step 5, which is implemented as follows,
first, the number of lanes Nlane determined for a certain translation in step 4 f Comparison of Nlane f+1 And Nlane f 、Nlane f+2 If Nlane f And Nlane f+1 If different, use Nlane f Replacement of Nlane f+1 ，f＝1,2,…,l-2；
Second step, according to Nlane f The result of the first step is classified by the value and distribution of (C), and s classes are provided and are marked as C g ＝<Nl g ,nc g >,Nl g Is a cluster of class C g Number of lanes of nc g Is Nlane g In (C) g G =1,2, \8230;, total number of s;
third, compare C g+1 And C g If Nl is g+1 Is different from Nl g And nc g+1 <, cv, order C g Nl of g Alternative C g+1 Nl of g+1 G =1,2, \8230s, s, the final optimization of the lane number results is done, where cv is a preset threshold.
And step 6 corrects the lane central line, which is realized as follows,
if the number of lanes of a certain section of road La is corrected, if the number of lanes of adjacent road sections Lb and Lc of La simultaneously meets the requirement that the number of lanes is the same as that of the lanes of La, and the number of lanes before and after correction does not change, then connecting the lane center lines of Lb and Lc to obtain the final lane center line of La; if the number of the lanes of the La adjacent road sections Lb or Lc is changed before and after the correction, calculating the road center line of the La road section calculated based on the position of the lane center line before the La correction according to the position of the lane center line extracted before the La correction, and re-determining the position of the lane center line after the La correction according to the width and the number of the lanes after the La correction.
The invention provides a lane-level road mapping system based on crowdsourcing space-time big data, which comprises the following modules,
a first module for establishing a similarity evaluation model of the track vector, and setting v a And v b Are two different trajectory vectors, the similarity evaluation model is as follows,
wherein,representing the similarity value between vectors, e being a natural base number, omega 1 And ω 2 Respectively represent distance factors diff Hd And angle factor diff θab And a weight value of ω 1 +ω 2 =1; distance factor diff Hd And angle factor diff θab Respectively represent vectors v a And v b Distance difference and angle difference of (a);
the second module is used for carrying out track optimization based on a growth clustering method of fusion empirical knowledge, and comprises the step of determining a weight value omega of a similarity evaluation model according to existing high-precision GPS track data and synchronous low-precision GPS track data 1 And ω 2 Extracting prior knowledge of track optimization, and performing data optimization by adopting a growing clustering mode based on the similarity between crowdsourcing track data;
the third module is used for constructing a Gaussian constraint mixed model and solving model parameters by using an EM (effective vector) algorithm; the gaussian constrained hybrid model is defined as follows,
wherein, p (x) represents the comprehensive probability value of the Gaussian constraint mixed model, x represents the sample value to be calculated, and x represents the value of the sample to be calculated when lane calculation is carried outJudging a vertical coordinate value of a track point in the window, which is vertically projected on a longitudinal section of the track point; k is the number of gaussian components, each corresponding to a lane; omega j The weight of the jth Gaussian component is the weight of the traffic flow of the corresponding lane; parameter mu 1 …μ k Is the average of the trajectories in each Gaussian component, equal to the centerline, μ, of each lane j Represents μ 1 …μ k Any value within the parameter, j =1,2, \ 8230;, k; σ is the standard deviation of the trajectories in each Gaussian component;
the number k of Gaussian components in the Gaussian constraint mixed model is obtained by calculating the value of the structural risk model and determining k according to the principle that the value of the structural risk model is minimum;
the fourth module is used for detecting lane information according to the result obtained by the third module to obtain a primary detection result of the number of lanes of the road section; the implementation is as follows, and the method,
all tracks on the same road section are taken as an extraction unit, and a given group of Intersection interaction is set 1 Intersection to Intersection interaction 2 Track set A of T From the track set A T Starting from one end of the window, constructing a moving rectangular window, wherein the long side of the moving rectangular window is parallel to the central lines covering all tracks at present, the wide side of the moving rectangular window is perpendicular to the central lines covering all tracks at present, the central line of the long side of the rectangular window is perpendicular to the central lines of all track data covered by the rectangular window, and the central line of the wide side of the rectangular window is superposed with the central line of the track data covered by the rectangular window;
moving rectangular window from trajectory set A T The method comprises the steps of starting translation according to the length of the long edge of a rectangular window, sequentially utilizing a Gaussian constraint hybrid model to detect the number of lanes and the center line of the lanes of a road section covered in each rectangular window, projecting all track points in the moving rectangular window onto the center line of the long edge of the rectangular window according to the moving rectangular window to obtain a projected track data set X = (X =) 1 ,x 2 ,…,x N ) T =1,2,3, \ 8230;, N, wherein x t The longitudinal coordinate value of the t-th track point after projection is represented,n is the number of the trace points participating in the projection; substituting the track data set X into a Gaussian constraint hybrid model, and extracting the number of lanes and the center line of the lanes of a road section in a rectangular window; suppose Intersection interaction 1 Intersection to Intersection interaction 2 Track set A of T The rectangular window is translated for one time, and the number of lanes determined by each translation is recorded as Nlane f L, as a primary detection result of the number of lanes of the road section;
the fifth module is used for correcting the primary detection result based on the road construction rule according to the primary detection result of the number of lanes of the road section acquired by the fourth module;
and the sixth module is used for correcting the lane center line according to the adjacent condition and the corrected lane number obtained by the fifth module.
And the fifth module corrects the primary detection result, which is realized as follows,
first, the number of lanes Nlane determined for a certain translation in the fourth module f Comparison of Nlane f+1 And Nlane f 、Nlane f+2 If Nlane f And Nlane f+1 If different, use Nlane f Replacement of Nlane f+1 ，f＝1,2,…,l-2；
Second step, according to Nlane f The result of the first step is classified by the value and distribution of (C), and s classes are provided and are marked as C g ＝<Nl g ,nc g >,Nl g Is a cluster class C g Number of lanes of nc g Is Nlane g In (C) g G =1,2, \ 8230;, total number of s;
third, compare C g+1 And C g If Nl is g+1 Is different from Nl g And nc g+1 <, cv, order C g Nl of g Alternative C g+1 Nl of g+1 G =1,2, \8230s, s, where cv is a preset threshold, completes the final optimization of the lane number result.
And the sixth module corrects the lane center line in the following way,
if the number of lanes in a certain section of road section La is corrected, if the number of lanes in the adjacent road sections Lb and Lc of La simultaneously meets the requirement that the number of lanes is the same as that of the lanes in La, and the number of lanes before and after correction is not changed, then connecting the lane center lines of Lb and Lc to obtain the final lane center line in La; if the number of the lanes of the La adjacent road sections Lb or Lc is changed before and after the correction, calculating the road center line of the La road section calculated based on the position of the lane center line before the La correction according to the position of the lane center line extracted before the La correction, and re-determining the position of the lane center line after the La correction according to the width and the number of the lanes after the La correction.
The invention constructs a high-precision lane-level road mapping technical scheme of crowdsourcing space-time big data, reduces the cost of acquiring urban fine road information, and has simple detection method and easy realization. The technical scheme provided by the invention comprises the following steps: firstly, selecting data with relatively high positioning precision from crowd-sourced track data by adopting a growth clustering method based on empirical knowledge according to the spatial similarity between a high-precision vehicle-mounted GPS track and synchronous low-precision GPS track data thereof; secondly, constructing a moving window vertical to the track data; then, longitudinally detecting all tracks of the road section by adopting an optimized Gaussian mixture model method to obtain the number of lanes in a detection window; further, by utilizing a road construction rule, namely, the condition that the lanes are additionally arranged only at the position close to the intersection on the same road section, and the number of the lanes in the middle part of the road section is usually kept unchanged, a lane number optimization strategy is provided, and the lane number information is corrected; and finally, correcting the extracted lane central line by using the corrected lane number information to finish the extraction of the lane level road information of the corresponding road section. The lane number judgment accuracy rate obtained by the method is 85%, the positioning accuracy of the lane center line is about 0.35m, the cost for obtaining urban fine road information is reduced, and the detection method is simple and easy to realize.
Description of the drawings:
FIG. 1 is a flow chart of a method of an embodiment of the present invention;
FIG. 2 is a schematic diagram of track vector similarity according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of empirical knowledge based growth clustering in accordance with an embodiment of the present invention;
fig. 4 is a schematic diagram of a preferred result of a trajectory of a growth clustering method based on empirical knowledge according to an embodiment of the present invention, where fig. 4a is a schematic diagram of an experimental area of crowdsourcing trajectory data, and fig. 4b is a schematic diagram of a preferred result of a trajectory;
FIG. 5 is a schematic diagram of a Gaussian mixture model and a detection of a lane centerline according to an embodiment of the present invention, where FIG. 5a is a result of the optimization of the Gaussian mixture model and FIG. 5b is a result of the detection of the lane centerline;
FIG. 6 is a schematic diagram of constructing rectangular window probe lane information according to an embodiment of the present invention;
FIG. 7 is a schematic view of lane number optimization for an embodiment of the present invention;
FIG. 8 is a diagram illustrating lane number detection results according to an embodiment of the present invention;
fig. 9 is a schematic diagram of the lane number optimization result according to the embodiment of the present invention.
Detailed Description
The technical scheme of the invention is explained in detail in the following by combining the drawings and the embodiment.
The invention provides a high-precision lane-level road mapping method for crowdsourcing of space-time big data, and the embodiment comprises the following steps of:
step 1, establishing a similarity evaluation model of the track vector. According to the driving characteristics of the vehicle, a general driver can drive along the center line of the lane according to the driving rule. Therefore, under the condition of not considering transient lane change behaviors, high-precision track data capable of truly depicting the driving track of a driver is generally concentrated near the center line of the lane, the distance between adjacent tracks is smaller than the width of the lane, and the course angle included angle between the tracks approaches to about 0 degree. In order to evaluate the similarity between such tracks belonging to the periphery of the center line of the same lane, the invention providesA new track vector similarity evaluation model is established, similarity measurement is carried out on the similarity model from two aspects of vertical distance and included angle between track vectors, wherein the track vectors refer to unit track vectors formed by each track point and course angle thereof in one track, namely, the moduli of the track vectors are the same and can be defined at will. The difference of the two trajectory vectors is measured in terms of both direction and distance. As shown in FIG. 2, N represents the north direction, v a <(x a ,y a ),(x a+1 ,y a+1 )>, and v b <(x b ,y b ),(x b+1 ,y b+1 )> is two different trajectory vectors, vector v a And v b Are respectively theta a And theta b The similarity evaluation model is as follows:
wherein,represents a similarity value between the vectors, anWhen the similarity value is 1, the two track vectors are completely the same, and when the similarity value is 0, the two track vectors are completely dissimilar; e is a natural base number; omega 1 And omega 2 Respectively represent distance factors (diff) Hd ) And angle factor (diff) θab ) And ω is 1 +ω 2 ＝1；diff Hd And diff θab Respectively represent vectors v a And v b The distance difference and the angle difference of (a). Equation 8 defines the distance difference diff between two trajectory vectors Hd Equation 9 defines the angular difference diff between the two trajectory vectors θab ：
Dis in equation 8 confine The lane width is a constant, which is used to constrain the similarity of the GPS tracks on the same lane near the center line of each lane. Hd ab Vector v a Starting point to v b The vertical distance of the starting point, the calculation formula is shown in formula 11; hd ba Is a vector v b Starting point to vector v a The vertical distance of the starting point is calculated according to the formula 12; delta theta is the vector v a And vector v b The difference of course angle of (1), wherein the vector v a And v b Are respectively theta a And theta b The calculation formulas are respectively as follows:
Δθ＝|θ a -θ b equation 10
And 2, carrying out track optimization based on a growth clustering method fusing empirical knowledge. The similarity value between the track vectors of the high-precision track points is relatively high. When the track optimization is carried out, firstly, the existing high-precision GPS track data and the synchronous low-precision GPS track data are needed, and the prior knowledge of the track optimization is extracted. In the embodiment, the precision of the high-precision DGPS track and the precision of the synchronous low-precision track are respectively 0.5m and 10-15m, and the sampling frequency is 1s. And (4) calculating the similarity between the low-precision track and the DGPS track according to the similarity estimation model established in the step one, and determining a weight value of the similarity evaluation model and acquiring empirical knowledge.
1) Similarity evaluation model weight value determination
Similarity is not only used for extracting prior knowledge in DGPS and low-precision GPS data, but also used for clustering and optimizing of multi-source data, so the invention further provides that the correlation of vertical distance, angle difference and measurement error is used for estimating omega 1 And ω 2 The value of (c).
For GPS track set T =<Trace 1 ,Trace 2 ,…,Trace s >, it contains s GPS tracks: trace 1 ,Trace 2 ,…,Trace s Synchronous high precision DGPS trajectories of a set of trajectories T are denoted by DT =<Dt 1 ,Dt 2 ,…,Dt s >，Dt 1 ,Dt 2 ,…,Dt s Is the trajectory data within the set DT. The accuracies of the track set DT and the track set T are 0.5m and 10-15m, respectively. Suppose the ith GPS Trace Trace i ＝<p 1 ,p 2 ,…,p n >，p 1 ,p 2 ,…,p n Respectively represent Trace Trace i The number of the track points is n; dt i ＝<rp 1 ,rp 2 ,…,rp n >，rp 1 ,rp 2 ,…,rp n Then is Trace Trace i Synchronous high-precision trajectory Dt i Tracing points of (1); trace i ∈T,Dt i E DT, i =1,2, \8230;, s. From Trace Trace i And its synchronous high-precision trajectory Dt i The track vector formed by the inner track points is expressed as follows: tv i ＝<v 1 ,v 2 ,…,v n-1 >，v 1 ,v 2 ,…,v n-1 Respectively represented by Trace i Point of track p 1 And p 2 ,…,p n-1 And p n A constructed trajectory vector; dv i ＝<rv 1 ,rv 2 ,…,rv n-1 >，rv 1 ,rv 2 ,…,rv n-1 Respectively represented by Trace i Synchronous high-precision trajectory Dt i Locus point rp 1 With rp 2 ,…,rp n-1 With rp n The constructed trajectory vector. Tv i And Dv i Are respectively expressed as: d i ＝<d 1 ,d 2 ,…，d n-1 >,d 1 ,d 2 ,…，d n-1 Representing a set of trajectory vectors Tv i And Dv i Inner phase corresponding trajectory vector v 1 ,v 2 ,…,v n-1 To rv 1 ,rv 2 ,…,rv n-1 The distance of (a); a. The i ＝<a 1 ,a 2 ,…，a n-1 >,a 1 ,a 2 ,…，a n-1 Representing a set of trajectory vectors Tv i And Dv i Inner phase corresponding trajectory vector v 1 ,v 2 ,…,v n-1 And rv 1 ,rv 2 ,…,rv n-1 I =1,2, \8230, s. The positioning error of all the trajectory data within the trajectory set T can be represented as E i ＝<ε 1 ,ε 2 ,…,ε n &In which E i Representing Trace Trace within a Trace set T i All track points and corresponding synchronous high-precision track Dt thereof i Of all points of the track, epsilon j Is a set E i An error value of any one of, wherein ∈ j ＝|p j -rp j |，p j Is Trace i Inner arbitrary one of the points of track, rp j Is p j Corresponding high precision trace points, i =1,2, \ 8230;, s, j =1,2, \ 8230;, n. In the similarity evaluation model, the weight value omega 1 And omega 2 Are respectively as follows (r) Dε And r Aε Respectively represent D i And E i Can estimate A based on the covariance matrix i And E i Value of (d):
ω 2 ＝1-ω 1 equation 14
2) Priori knowledge extraction
The first step is as follows: according to the determination of the vector similarity evaluation model and the weighted value provided in the step 1, calculating the similarity between the high-precision DGPS and the synchronous low-precision GPS thereof, and meanwhile, calculating the measurement error of the low-precision GPS data by comparing the position of the low-precision GPS track data with the position of the synchronous high-precision DGPS data;
and secondly, constructing attribute description pairs belonging to the same GPS track data according to the similarity obtained in the first step and the corresponding GPS measurement error, namely (the similarity value and the GPS measurement error) are attribute description pairs of a certain track.
And thirdly, selecting all the GPS data with the similarity values of more than 0.5, more than 0.6 and more than 0.7 of 8230from the attribute description pairs of the GPS data according to the similarity threshold value, namely from the similarity value of 0.5, counting the GPS measurement errors of all the GPS data in each similarity threshold value, and calculating the average value of the GPS measurement errors and the proportion of the number of the data meeting the threshold values in the total data.
And fourthly, defining the similarity threshold (for example, the similarity threshold is more than 0.5 and the similarity threshold is more than 0.6 \8230;) set in the third step as follows: ts h H =1,2,3,4,5; defining the average of the GPS measurement errors that satisfy these thresholds as Ts h H =1,2,3,4,5; the ratio of the GPS measurement error to the overall data meeting the threshold is defined as: per h H =1,2,3,4,5; and completing acquisition of prior knowledge.
Wherein, ST h Indicates that Ts is satisfied h Data Set of (ST) h e.T, T is the overall data set of experimental data for empirical extraction, ST h The percentage calculation formula is as follows:
Per h represents ST h Percentage of (A), N (ST) h ) And N (T) is each ST h And the number of T track points, ST h The calculation formula of (2) is as follows:
Tε h is Ts h Is the measurement error of h The sum of all track point errors, h =1,2, \ 8230;, 5,Ts h ,Tε h ，Per h Is recorded as RST h ＝<Ts h ,Tε h ,Per h >,RST h Is a priori knowledge set and serves as the priori knowledge of the growth clustering method.
According to the extracted priori knowledge, distance weight and angle weight in the similarity evaluation model are calculated, then data optimization is carried out by adopting a growing clustering method based on the similarity between crowdsourced trajectory data, wherein T epsilon in the priori knowledge is a clustering threshold value, per is used for selecting a proportion of high-precision data from the whole trajectory cluster, and the proportion is shown in figure 3 (v in figure 3) s Representing a seed vector, v sn Represents any one of the vectors, cluster, that performs similarity calculations with the seed vector 1 ，Cluster 2 ，Cluster 3 Representing a plurality of track vector clusters obtained after clustering; the 'Selected data' in part (d) in fig. 3 represents the finally Selected track data), and the main steps of the growing and clustering method are as follows:
the first step is as follows: initializing all track vectors and marking the track vectors as non-clustered; initializing the number of the current Class Cluster (CC) L ) Record CC L ＝1；
The second step: if there are unclustered trajectory vectors, randomly selecting one trajectory vector from the rest unclustered trajectory vectors as a seed trajectory vector v s The cluster mark of the seed track vector is CL (v) s ) And CL (v) s )＝CC L As shown in fig. 3 (a), the third step is performed; if all tracks have been clustered, as in part (c) of FIG. 3, go to the fifth step;
the third step: search v s Adjacent track vector, denoted v sn : if Sim (v) is satisfied s ,v sn )>, T epsilon, v s And v sn Merge into a track cluster, v sn Is denoted as CL (v) sn ) And CL (v) sn )＝CC L Entering the fourth step; if the current species can not be foundSubvector v s When the track vector of the similarity threshold value is satisfied, order CC L ＝CC L +1, returning to the second step; (Sim (v) s ,v sn ) Is v s And v sn Is the similarity threshold value of (D), T epsilon
The fourth step: let the track vector v sn As seed trajectory v s Returning to the third step, as shown in part (b) of fig. 3;
the fifth step: according to the operation results from the first step to the fourth step, namely the final clustering cluster, calculating the proportion of the number of track points in all clusters to the total number of track points participating in clustering, then arranging all the track clusters in a reverse order according to the ratio of the track points to the total track points, wherein Per represents the data optimization selectivity, starting from the track point proportion of the first cluster and accumulating and summing until Per is met, and selecting the clusters participating in accumulating and summing as high-precision data optimization, as shown in part (d) in FIG. 3, wherein the selected experiment area crowdsourcing track data is shown in FIG. 4a, and the track optimization result is shown in FIG. 4b.
And 3, constructing a Gaussian constraint mixed model, and solving model parameters by using an EM (effective electromagnetic) algorithm. On the basis of the existing algorithm for solving the lane information by the Gaussian constraint hybrid algorithm, the method is optimized. The Gaussian constrained mixture model is defined as follows:
wherein p (x) is represented as a comprehensive probability value of a Gaussian constraint hybrid model, and x represents a sample value to be calculated (when lane calculation is carried out, x represents a vertical coordinate value of a vertical projection of a track point in a judgment window on a longitudinal section of the track point); k is the number of Gaussian components, namely the number of Gaussian peaks in the constraint Gaussian mixture model, and represents the number of lanes, and each Gaussian component corresponds to one lane; omega 1 …ω k Is a weight per component, corresponding to the traffic flow per lane, where the weight values are positive and normalized, i.e., ω j Is the weight of the jth Gaussian component, ω j >0,j＝1,2,…,k,ω 1 +ω 2 +…ω k =1; parameter mu 1 …μ k Is the average of the trajectories in each Gaussian component, equal to the centerline of each lane, μ j Represents mu 1 …μ k Any value within the parameter, j =1,2, \ 8230;, k; σ is the standard deviation of the trajectory in each gaussian component, and since the width of each lane is generally the same as the width of the adjacent lane, σ is set to a constant, typically 1.75 (lane width is typically around 3.75m according to domestic road construction standards, which can be reset by those skilled in the art according to the selected regional road construction standards during implementation). Solving the model parameters using the EM method: theta j (m) (ω j (m) ,μ j (m) ,σ (m) ) J =1,2, \8230;, k, where m is the number of iterations. At present, there are many mature methods for solving each parameter of the gaussian mixture model by using the EM algorithm, and in the specific implementation process, technicians may refer to the existing methods, which is not described in detail herein.
The key of the Gaussian constraint hybrid model is to obtain the number of Gaussian components, that is, calculate the value of the structural risk model under each k value, and then select the k corresponding to the minimum value of the structural risk model as the number of lanes. The construction method of the structural risk model is as follows:
k＝min(R srm (p(x i |θ k ) ))) 19
L(x i ,p(x i |θ k ))＝-log(p(x i |θ k ) Equation 20)
R in formula 2 srm (p(x i |θ k ) Is a structural risk model, L (x) i ,p(x i |θ k ) Is an empirical risk model for assessing fitness, J (p (x) i |θ k ) Is a regularization term that is used to denote the model complexity, i.e., is expressed as: j. the design is a square TSW (p(x i |θ k ))，λ>, 0 is a regularization parameter, p (x) i |θ k ) Representing the sample value x i At model parameter θ k Gaussian probability value of the condition, wherein the model parameter θ k Can be expressed as: theta k (ω k ,μ k σ), n represents the number of samples, i =1,2, \ 8230;, n; the calculation formula is as follows:
wherein D w Is the optimized track laying width on the road surface, as shown in figure 5, and indicated by '1' in figure 5a with long dashed line st Gaussian component 'represents the first component of the Gaussian mixture model, short-dashed line labeled' 2 nd Gaussian component' represents the second component of the Gaussian mixture model; wherein the parameter μ is located in fig. 5b 1 、μ 2 The average value of the first gaussian component and the average value of the second gaussian component in fig. 5a correspond to the position of the center line of the first lane and the position of the center line of the second lane. (how to obtain the tiled width of the track on the road surface has been proposed by many methods at present, and in the specific implementation process, the skilled person can select by himself, which is not described in detail in the present invention), where k represents the number of possible lanes, and according to the current domestic road construction standard, the number of urban lanes generally includes two lanes, three lanes, four lanes, and five lanes, that is, k =2,3,4,5.Δ μ k Is the average value mu of two adjacent Gaussian peaks in the Gaussian component corresponding to k j J =1,2, \8230;, k, which also reflects the width of the detected lane. Δ μ k The calculation method of (A) is shown in formula 6, and k, η, and γ in formula 6 ij Is a hyperparameter of the EM algorithm based on maximum posterior estimation; x represents the sample value (when calculating the lane, x represents the track point in the judgment windowOrdinate values vertically projected on a longitudinal section thereof); n represents the number of sample data, ω j+1 A weight value representing the j +1 th Gaussian component, j =1,2, \8230, k-1, k is the Gaussian component fraction, k =2,3,4,5; very mature parameter recommendations exist in the existing research, and the calculation can be carried out by referring to the numerical values given in the existing method cases during specific implementation, which is not described in detail.
And 4, completing high-precision lane-level road mapping according to the optimized Gaussian mixture model method in the step 3, realizing detection of lane information and obtaining a primary detection result of the number of lanes in the road section. In the specific implementation process of high-precision lane-level road mapping, all tracks on the same road segment are taken as an extraction unit by using the existing method (how to classify all tracks on the same road segment into one extraction unit is known, and many mature methods exist at present, and those skilled in the art can refer to the existing method in the specific implementation process, which is not described in detail herein).
Suppose a given set of slave Intersection interactions 1 Intersection to Intersection interaction 2 Track set A of T From the track set A T Starting with one end, a moving rectangular window is constructed as shown in fig. 6. The length and width of the moving rectangular window are rh and rw respectively (the length and width of the rectangular window are preferably defined as 10m and 30m, and a person skilled in the art can preset values in specific implementation), wherein the long side of the moving rectangular window is parallel to the center lines currently covering all tracks, the wide side of the moving rectangular window is perpendicular to the center lines currently covering all tracks, the center line of the long side of the rectangular window is perpendicular to the center lines of all track data covered by the long side of the rectangular window, and the center line of the wide side of the rectangular window is overlapped with the center line of the track data covered by the wide side of the rectangular window, so how to obtain the center line of a section of track data.
The preferred implementation further provided by the embodiments is as follows:
will move the rectangular window from the railTrace set A T The translation is started according to the length of the long edge of the rectangular window, and the number of lanes and the center line of the lanes of the road section covered in each rectangular window are detected by sequentially utilizing a Gaussian constraint hybrid model. The specific method comprises the following steps: according to the constructed rectangular window, projecting all track points in the rectangular window onto the long-edge central line of the rectangular window to obtain a projected track data set X = (X) 1 ,x 2 ,…,x N ) T =1,2,3, \ 8230;, N, wherein x t And (4) expressing the longitudinal coordinate value of the t-th track point after projection, wherein N is the number of the track points participating in projection. And (3) substituting the track data set X into a corresponding calculation formula (namely a Gaussian constraint hybrid model shown in a formula 17) according to the step 3, and extracting the number of lanes and the center line of the lanes of the road section in the rectangular window. Suppose Intersection interaction 1 Intersection to Intersection interaction 2 Track set A of T The rectangular window is translated for one time, and the number of lanes determined by each translation is recorded as Nlane f F =1,2, \ 8230;, l, as a result of primary detection of the number of lanes of the road section.
Step 5, according to the primary detection result of the number of lanes of the road section obtained in the step 4, based on the road construction rules, providing a lane number optimization strategy, and correcting the primary detection result, as shown in fig. 7, where there is a section Part 1 、Part 2 、Part 3 Wherein Part 1 And Part 3 There is an extension lane area. In most cases, additional lanes always appear on the road section between two intersections near the intersection, and the number of lanes in the middle of the road section is usually kept unchanged, so the invention provides a method for optimizing the lane number extraction result, which comprises the following specific steps:
the first step is as follows: number of lanes Nlane determined for a certain translation in step 4 f Comparison of Nlane f+1 And Nlane f 、Nlane f+2 If Nlane f And Nlane f+1 If different, use Nlane f Replacement of Nlane f+1 ，f＝1,2,…,l-2。
The second step: according to Nlane f Value and distribution of (2) to the result of the first stepSorting, e.g. if Nlane e ,Nlane e+1 ,Nlane e+2 ,…,Nlane e+c Are classified into a class in which e is the same<l,e+c&l, and l. Suppose there are s classes, denoted C g ＝<Nl g ,nc g >,Nl g Is a cluster of class C g Number of lanes of nc g Is Nlane g In (C) g G =1,2, \ 8230;, total number of s.
The third step: comparison C g+1 And C g If Nl is g+1 Is different from Nl g And nc g+1 &Cv (cv is a threshold value depending on a road construction rule, and the modification rule is mainly embodied in that when an urban road approaches to an intersection, a buffer area range of an additional lane appears, for example, according to the current road design rule, the length of the additional lane is about 50m, namely, from a road section positioned at the intersection, the length of the additional lane is 50m, therefore, when the number of the divided sections judged by the number of the lanes is set to 10 meters, the cv is recommended to be 5, and a person skilled in the art can preset values during the specific implementation), so that C is enabled to be a value which is preset by the person skilled in the art g Nl of g Replacement C g+1 Nl of g+1 G =1,2, \8230s, s, the final optimization of the lane number results is done. The lane number detection result is shown in fig. 8, and the lane number detection optimization is shown in fig. 9 (wherein the abscissa of fig. 8 represents the movement number of the moving window during the sliding detection process, and the ordinate represents the lane number detection result for each sliding detection process).
And 6, correcting the center line of the corresponding lane according to the corrected number of lanes obtained in the step 5. When the number of lanes in a certain section of road is corrected, the center line of the corresponding lane is corrected by adopting the adjacent principle. The specific method comprises the following steps:
if the number of lanes of a certain section of road La is corrected, searching front and rear adjacent road sections La, and if the number of lanes of the road sections La adjacent to Lb and Lc simultaneously meet the requirement that the number of lanes is the same as that of the lanes of La, and the number of lanes before and after correction of the road sections is not changed, connecting Lb with the center line of the lanes of Lc to obtain the final center line of the lanes of La; if the number of lanes of the La adjacent road section Lb or Lc is changed before and after the correction, calculating the road center line of the La road section based on the position of the lane center line extracted before the La correction according to the position of the lane center line extracted before the La correction, (how to calculate the road center line of the La road section according to the position of the lane center line before the La correction, which is a mature method at present and is not described in detail), redefining the position of the lane center line after the La correction according to the width of the lane after the La correction and the number of lanes, that is, obtaining the position of the lane center line of each lane after the La correction by sequentially paralleling the road center line of the La road section to the road center line according to the number of lanes and the width of the lane at equal intervals.
Based on the invention, the lane information of the road to be urban can be conveniently acquired from the GPS track data, and basic road network data is provided for future intelligent navigation and unmanned driving.
In specific implementation, the method provided by the invention can realize automatic operation flow based on software technology, and can also realize a corresponding system in a modularized mode.
The invention provides a lane-level road mapping system based on crowdsourcing space-time big data, which comprises the following modules,
a first module for establishing a similarity evaluation model of the track vector, and setting v a And v b Are two different trajectory vectors, the similarity evaluation model is as follows,
wherein,representing the similarity value between vectors, e being a natural base number, omega 1 And omega 2 Respectively represent distance factors diff Hd And angle factor diff θab And ω is 1 +ω 2 =1; distance factor diff Hd And angle factor diff θab Respectively represent vectors v a And v b Distance difference and angle difference of (a);
the second module is used for carrying out track optimization based on a growth clustering method of fusion empirical knowledge, and comprises the step of determining a weight value omega of a similarity evaluation model according to existing high-precision GPS track data and synchronous low-precision GPS track data 1 And ω 2 Extracting prior knowledge of track optimization, and performing data optimization by adopting a growing clustering mode based on the similarity between crowdsourcing track data;
the third module is used for constructing a Gaussian constraint mixed model and solving model parameters by using an EM (effective vector) algorithm; the gaussian-constrained hybrid model is defined as follows,
wherein p (x) is represented as the comprehensive probability value of the Gaussian constraint mixed model, x represents a sample value to be calculated, and x represents the vertical coordinate value of the vertical projection of the track point in the judgment window on the longitudinal section when the lane calculation is carried out; k is the number of gaussian components, each corresponding to a lane; omega j The weight of the jth Gaussian component is the weight of the traffic flow of the corresponding lane; parameter mu 1 …μ k Is the average of the trajectories in each Gaussian component, equal to the centerline, μ, of each lane j Represents mu 1 …μ k Any value within the parameters, j =1,2, \8230;, k; σ is the standard deviation of the trajectory in each Gaussian component;
the number k of Gaussian components in the Gaussian constraint mixed model is obtained by calculating the value of the structural risk model and determining k according to the principle that the value of the structural risk model is minimum;
the fourth module is used for detecting lane information according to the result obtained by the third module to obtain a primary detection result of the number of lanes of the road section; the implementation is as follows, and the method,
all tracks on the same road section are taken as an extraction unit, and a given group of Intersection interaction is set 1 Intersection to Intersection interaction 2 Track set A of T From the track set A T Starting from one end of the window, constructing a moving rectangular window, wherein the long side of the moving rectangular window is parallel to the central lines covering all tracks at present, the wide side of the moving rectangular window is perpendicular to the central lines covering all tracks at present, the central line of the long side of the rectangular window is perpendicular to the central lines of all track data covered by the rectangular window, and the central line of the wide side of the rectangular window is superposed with the central line of the track data covered by the rectangular window;
moving rectangular window from trajectory set A T The method comprises the steps of starting translation according to the length of the long edge of a rectangular window, sequentially utilizing a Gaussian constraint hybrid model to detect the number of lanes and the center line of the lanes of a road section covered in each rectangular window, projecting all track points in the moving rectangular window onto the center line of the long edge of the rectangular window according to the moving rectangular window to obtain a projected track data set X = (X =) 1 ,x 2 ,…,x N ) T =1,2,3, \8230, N, where x t Expressing the longitudinal coordinate value of the t-th track point after projection, wherein N is the number of the track points participating in projection; substituting the track data set X into a Gaussian constraint hybrid model, and extracting the number of lanes and the center line of the lanes of a road section in a rectangular window; suppose Intersection interaction 1 Intersection to Intersection interaction 2 Track set A of T The rectangular window is translated for one time, and the number of lanes determined by each translation is recorded as Nlane f L, as a primary detection result of the number of lanes of the road section;
the fifth module is used for correcting the primary detection result according to the primary detection result of the number of lanes of the road section acquired by the fourth module and based on the road construction rule;
and the sixth module is used for correcting the lane center line according to the adjacent condition and the corrected lane number obtained by the fifth module.
The specific implementation of each module can refer to corresponding steps, which are not described in detail in the present invention.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments, or alternatives may be employed, by those skilled in the art, without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (6)
1. A lane-level road mapping method based on crowdsourcing space-time big data is characterized in that: comprises the following steps of (a) preparing a solution,
step 1, establishing a similarity evaluation model of a track vector, and setting v a And v b Are two different trajectory vectors, the similarity evaluation model is as follows,
wherein,representing the similarity value between vectors, e being a natural base number, omega 1 And ω 2 Respectively represent distance factors diff Hd And angle factor diff θab And ω is 1 +ω 2 =1; distance factor diff Hd And angle factor diff θab Respectively represent vectors v a And v b The distance difference and the angle difference of (a);
step 2, carrying out track optimization based on a growth clustering method of fusion empirical knowledge, wherein the track optimization comprises the step of determining a weight value omega of a similarity evaluation model according to the existing high-precision GPS track data and the synchronous low-precision GPS track data 1 And ω 2 Extracting prior knowledge of track optimization, and optimizing data by adopting a growth clustering mode based on the similarity between crowdsourcing track data;
step 3, constructing a Gaussian constraint mixed model, and solving model parameters by using an EM (effective electromagnetic) algorithm; the gaussian-constrained hybrid model is defined as follows,
wherein p (x) is represented as the comprehensive probability value of the Gaussian constraint mixed model, x represents a sample value to be calculated, and x represents the vertical coordinate value of the vertical projection of the track point in the judgment window on the longitudinal section when the lane calculation is carried out; k is the number of Gaussian components, each corresponding to a lane; omega j Is the weight of the jth Gaussian component and corresponds to the traffic flow of the lane; parameter mu 1 …μ k Is the average of the trajectories in each Gaussian component, equal to the centerline of each lane, μ j Represents mu 1 …μ k Any value within the parameters, j =1,2, \8230;, k; σ is the standard deviation of the trajectories in each Gaussian component;
the number k of Gaussian components in the Gaussian constrained hybrid model is obtained by calculating the value of the structural risk model and determining k on the basis of the minimum value of the structural risk model;
step 4, detecting lane information according to the result obtained in the step 3 to obtain a primary detection result of the number of lanes of the road section; the implementation mode is as follows,
setting a given group of Intersection intersections by taking all tracks on the same road section as an extraction unit 1 Intersection to Intersection interaction 2 Track set A of T From the set of trajectories A T Starting from one end of the window, constructing a moving rectangular window, wherein the long side of the moving rectangular window is parallel to the central lines covering all tracks at present, the wide side of the moving rectangular window is perpendicular to the central lines covering all tracks at present, the central line of the long side of the rectangular window is perpendicular to the central lines of all track data covered by the rectangular window, and the central line of the wide side of the rectangular window is superposed with the central line of the track data covered by the rectangular window;
moving rectangular window from trajectory set A T The method comprises the steps of starting translation according to the length of the long edge of the rectangular window, sequentially detecting the number of lanes and the center line of the lanes of a road section covered in each rectangular window by utilizing a Gaussian constraint hybrid model, and projecting all track points in the movable rectangular window to the long edge of the rectangular window according to the movable rectangular windowOn the central line, a projected trajectory data set X = (X) is obtained 1 ,x 2 ,…,x N ) Wherein x is 1 ,x 2 ,…,x N The vertical coordinate values of N track points are shown in 1,2,3, \ 8230after projection, and N is the number of the track points participating in projection; substituting the track data set X into a Gaussian constraint hybrid model, and extracting the number of lanes and the center line of the lanes of a road section in a rectangular window; suppose Intersection interaction 1 Intersection to Intersection interaction 2 Track set A of T The rectangular window is translated for one time, and the number of lanes determined by each translation is recorded as Nlane f L, as a primary detection result of the number of lanes of the road section;
step 5, correcting the primary detection result based on the road construction rule according to the primary detection result of the number of lanes of the road section acquired in the step 4;
and 6, correcting the lane center line according to the adjacent condition according to the corrected lane number obtained in the step 5.
2. The method of claim 1, wherein the method comprises: in step 5, the primary detection result is corrected in the following way,
first, the number of lanes Nlane determined for a certain translation in step 4 f Comparison of Nlane f+1 And Nlane f 、Nlane f+2 If Nlane f And Nlane f+1 If different, use Nlane f Replacement of Nlane f+1 ，f＝1,2,…,l-2；
Second step, according to Nlane f The result of the first step is classified by the value and distribution of (C), and s classes are provided and are marked as C g ＝<Nl g ,nc g >,Nl g Is the g-th cluster C g Number of lanes of nc g Number of lanes Nlane determined for each translation f In (C) g G =1,2, \ 8230;, total number of s;
thirdly, comparing the g +1 th cluster C g+1 And the g cluster C g If the g +1 th clusterC g+1 Number of lanes Nl g+1 Is different from Nl g And the number of lanes Nlane determined for each translation f Belong to C g+1 Total number of (nc) g+1 <, cv, order C g Nl of g Replacement C g+1 Nl of g+1 G =1,2, \8230s, s, the final optimization of the lane number results is done, where cv is a preset threshold.
3. The method of lane-level road mapping based on crowd-sourced spatio-temporal big data of claim 1 or 2, characterized in that: step 6, correcting the center line of the lane in the following way,
if the number of lanes of a certain section of road La is corrected, if the number of lanes of adjacent road sections Lb and Lc of La simultaneously meets the requirement that the number of lanes is the same as that of the lanes of La, and the number of lanes before and after correction does not change, then connecting the lane center lines of Lb and Lc to obtain the final lane center line of La; if the number of the lanes of the La adjacent road sections Lb or Lc is changed before and after the correction, calculating the road center line of the La road section estimated based on the lane center line position before the La correction according to the lane center line position extracted before the La correction, and re-determining the lane center line position after the La correction according to the lane width and the lane number after the La correction.
4. The utility model provides a lane level road mapping system based on crowd-sourced space-time big data which characterized in that: comprises the following modules which are used for realizing the functions of the system,
a first module for establishing a similarity evaluation model of the track vector, and setting v a And v b Are two different trajectory vectors, the similarity evaluation model is as follows,
wherein,representing the similarity value between vectors, e being a natural base，ω 1 And ω 2 Respectively represent distance factors diff Hd And angle factor diff θab And a weight value of ω 1 +ω 2 =1; distance factor diff Hd And angle factor diff θab Respectively represent vectors v a And v b Distance difference and angle difference of (a);
the second module is used for carrying out track optimization based on a growth clustering method of fusion empirical knowledge and comprises the steps of determining a weight value omega of a similarity evaluation model according to existing high-precision GPS track data and synchronous low-precision GPS track data 1 And ω 2 Extracting prior knowledge of track optimization, and optimizing data by adopting a growth clustering mode based on the similarity between crowdsourcing track data;
the third module is used for constructing a Gaussian constraint mixed model and solving model parameters by using an EM (effective electromagnetic) algorithm; the gaussian constrained hybrid model is defined as follows,
p (x) is expressed as a comprehensive probability value of the Gaussian constraint hybrid model, x is expressed as a sample value to be calculated, and x represents a vertical coordinate value of a track point in a judgment window vertically projected on a longitudinal section of the track point when lane calculation is carried out; k is the number of gaussian components, each corresponding to a lane; omega j Is the weight of the jth Gaussian component and corresponds to the traffic flow of the lane; parameter mu 1 …μ k Is the average of the trajectories in each Gaussian component, equal to the centerline of each lane, μ j Represents mu 1 …μ k Any value within the parameter, j =1,2, \ 8230;, k; σ is the standard deviation of the trajectory in each Gaussian component;
the number k of Gaussian components in the Gaussian constraint mixed model is obtained by calculating the value of the structural risk model and determining k according to the principle that the value of the structural risk model is minimum;
the fourth module is used for detecting lane information according to the result obtained by the third module to obtain a primary detection result of the number of lanes on the road section; the implementation is as follows, and the method,
setting a given group of Intersection intersections by taking all tracks on the same road section as an extraction unit 1 Intersection to Intersection interaction 2 Track set A of T From the set of trajectories A T Starting from one end of the window, constructing a moving rectangular window, wherein the long side of the moving rectangular window is parallel to the central lines covering all tracks at present, the wide side of the moving rectangular window is perpendicular to the central lines covering all tracks at present, the central line of the long side of the rectangular window is perpendicular to the central lines of all track data covered by the rectangular window, and the central line of the wide side of the rectangular window is superposed with the central line of the track data covered by the rectangular window;
moving rectangular window from trajectory set A T The method comprises the steps of sequentially detecting the number of lanes and the lane center line of a road section covered in each rectangular window by using a Gaussian constraint hybrid model according to a moving rectangular window, and projecting all track points in the moving rectangular window onto the long edge center line of the rectangular window to obtain a projected track data set X = (X is the length of the long edge of the rectangular window), wherein the translation is started according to the length of the long edge of the rectangular window 1 ,x 2 ,…,x N ) T =1,2,3, \8230, N, where x 1 ,x 2 ,…,x N The vertical coordinate values of N track points are shown in 1,2,3, \8230afterprojection, and N is the number of the track points participating in projection; substituting the track data set X into a Gaussian constraint hybrid model, and extracting the number of lanes and the center line of the lanes of a road section in a rectangular window; suppose that Intersection interaction is driven from 1 Intersection to Intersection interaction 2 Track set A of T The rectangular window is translated for one time, and the number of lanes determined by each translation is recorded as Nlane f L, as a primary detection result of the number of lanes of the road section;
the fifth module is used for correcting the primary detection result based on the road construction rule according to the primary detection result of the number of lanes of the road section acquired by the fourth module;
and the sixth module is used for correcting the lane center line according to the adjacent condition and the corrected lane number obtained by the fifth module.
5. The system of claim 4, wherein the system is configured to map roads at a lane level based on crowd-sourced spatiotemporal big data: the fifth module corrects the primary detection result, and the implementation manner is as follows,
first, the number of lanes Nlane determined for a certain translation in the fourth module f Comparison of Nlane f+1 And Nlane f 、Nlane f+2 If Nlane f And Nlane f+1 If different, use Nlane f Replacement of Nlane f+1 ，f＝1,2,…,l-2；
Second step, according to Nlane f The result of the first step is classified by the value and distribution of (C), and s classes are provided and are marked as C g ＝<Nl g ,nc g >,Nl g Is the g-th cluster C g Number of lanes, nc g Number of lanes Nlane determined for each translation f In the genus of C g G =1,2, \ 8230;, total number of s;
thirdly, comparing the g +1 th cluster C g+1 And the g cluster C g If the g +1 th cluster C g+1 Number of lanes Nl g+1 Is different from Nl g And the number of lanes Nlane determined for each translation f Belong to C g+1 Total number of (nc) g+1 <, cv, order C g Nl of g Replacement C g+1 Nl of g+1 G =1,2, \8230s, s, the final optimization of the lane number results is done, where cv is a preset threshold.
6. The system of claim 4 or 5, wherein the system is configured to map roads at lane level based on crowd-sourced spatiotemporal big data: the sixth module corrects the center line of the lane in the following way,
if the number of lanes in a certain section of road section La is corrected, if the number of lanes in the adjacent road sections Lb and Lc of La simultaneously meets the requirement that the number of lanes is the same as that of the lanes in La, and the number of lanes before and after correction is not changed, then connecting the lane center lines of Lb and Lc to obtain the final lane center line in La; if the number of the lanes of the La adjacent road sections Lb or Lc is changed before and after the correction, calculating the road center line of the La road section calculated based on the position of the lane center line before the La correction according to the position of the lane center line extracted before the La correction, and re-determining the position of the lane center line after the La correction according to the width and the number of the lanes after the La correction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610370700.9A CN105957342B (en) | 2016-05-30 | 2016-05-30 | Track grade road plotting method and system based on crowdsourcing space-time big data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610370700.9A CN105957342B (en) | 2016-05-30 | 2016-05-30 | Track grade road plotting method and system based on crowdsourcing space-time big data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105957342A CN105957342A (en) | 2016-09-21 |
| CN105957342B true CN105957342B (en) | 2018-05-29 |
Family
ID=56910078
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610370700.9A Active CN105957342B (en) | 2016-05-30 | 2016-05-30 | Track grade road plotting method and system based on crowdsourcing space-time big data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105957342B (en) |
Families Citing this family (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3358302B1 (en) * | 2015-09-30 | 2021-03-31 | Nissan Motor Co., Ltd. | Travel control method and travel control device |
| CN108664016B (en) | 2017-03-31 | 2020-09-15 | 腾讯科技（深圳）有限公司 | Method and device for determining lane center line |
| CN109579839B (en) * | 2017-09-29 | 2020-11-03 | 阿里巴巴（中国）有限公司 | Parallel path identification method, parallel path similarity determination method and device |
| CN107818577A (en) * | 2017-10-26 | 2018-03-20 | 滁州学院 | A kind of Parts Recognition and localization method based on mixed model |
| CN110443285A (en) * | 2019-07-16 | 2019-11-12 | 浙江大华技术股份有限公司 | The determination method, apparatus and computer storage medium of similar track |
| US11195027B2 (en) * | 2019-08-15 | 2021-12-07 | Toyota Motor Engineering And Manufacturing North America, Inc. | Automated crowd sourcing of road environment information |
| CN110634291B (en) * | 2019-09-17 | 2020-12-29 | 武汉中海庭数据技术有限公司 | High-precision map topology automatic construction method and system based on crowdsourcing data |
| CN110728735A (en) * | 2019-09-17 | 2020-01-24 | 武汉中海庭数据技术有限公司 | Road-level topological layer construction method and system |
| CN110749329B (en) * | 2019-10-26 | 2021-08-17 | 武汉中海庭数据技术有限公司 | Lane level topology construction method and device based on structured road |
| CN111288999B (en) * | 2020-02-19 | 2021-08-31 | 深圳大学 | Pedestrian road network attribute detection method, device and equipment based on mobile terminal |
| CN111341103B (en) * | 2020-03-03 | 2021-04-27 | 鹏城实验室 | Lane information extraction method, device, equipment and storage medium |
| CN111578964B (en) * | 2020-04-13 | 2022-01-07 | 河北德冠隆电子科技有限公司 | High-precision map road information rapid generation system and method based on space-time trajectory reconstruction |
| CN111858801B (en) * | 2020-06-30 | 2024-03-22 | 北京百度网讯科技有限公司 | Road information mining method and device, electronic equipment and storage medium |
| CN112150804B (en) * | 2020-08-31 | 2021-10-19 | 中国地质大学（武汉） | City multi-type intersection identification method based on MaskRCNN algorithm |
| CN112115817B (en) * | 2020-09-01 | 2024-06-07 | 国交空间信息技术（北京）有限公司 | Remote sensing image road track correctness checking method and device based on deep learning |
| CN112308961B (en) * | 2020-11-05 | 2022-06-10 | 湖南大学 | Robot rapid robust three-dimensional reconstruction method based on layered Gaussian mixture model |
| CN112488217B (en) * | 2020-12-05 | 2022-07-29 | 武汉中海庭数据技术有限公司 | Zebra crossing correction method, electronic device and storage medium |
| CN112733078B (en) * | 2020-12-29 | 2023-10-10 | 武汉中海庭数据技术有限公司 | Method and device for smooth connection among multiple paths of fragments of crowdsourcing data |
| CN112683286A (en) * | 2021-03-16 | 2021-04-20 | 季华实验室 | Method and system for establishing topological road network map, storage medium and electronic equipment |
| CN113453157B (en) * | 2021-08-31 | 2021-11-23 | 浙江宇视科技有限公司 | Time-space trajectory calibration method and device, storage medium and electronic equipment |
| CN115507874B (en) * | 2022-06-09 | 2024-03-01 | 广东省智能网联汽车创新中心有限公司 | Lane matching method and device based on V2X |
| CN115457767B (en) * | 2022-08-30 | 2023-08-29 | 同济大学 | Lane information extraction method based on crowdsourcing machine motor car track |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8116968B2 (en) * | 2008-12-23 | 2012-02-14 | National Chiao Tung University | Method for identification of traffic lane boundary |
| CN103903019B (en) * | 2014-04-11 | 2017-12-15 | 北京工业大学 | A kind of automatic generation method of multilane track of vehicle space-time diagram |
| US10013508B2 (en) * | 2014-10-07 | 2018-07-03 | Toyota Motor Engineering & Manufacturing North America, Inc. | Joint probabilistic modeling and inference of intersection structure |
| CN104700617B (en) * | 2015-04-02 | 2016-03-16 | 武汉大学 | Based on the high precision track information extracting method of low precision GPS track data |
| CN105138779B (en) * | 2015-08-31 | 2018-03-27 | 武汉大学 | Vehicle GPS space-time track big data method for optimizing and system |
| CN105444770B (en) * | 2015-12-18 | 2018-05-29 | 上海交通大学 | Track level map generation and localization method based on smart mobile phone |
-
2016
- 2016-05-30 CN CN201610370700.9A patent/CN105957342B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN105957342A (en) | 2016-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105957342B (en) | Track grade road plotting method and system based on crowdsourcing space-time big data | |
| US11681746B2 (en) | Structured prediction crosswalk generation | |
| EP3271748B1 (en) | Guided geometry extraction for localization of a device | |
| AU2019233779B2 (en) | Vehicle tracking | |
| AU2019201834B2 (en) | Geometric fingerprinting for localization of a device | |
| CN110264721B (en) | Urban intersection surrounding vehicle track prediction method | |
| AU2015370581B2 (en) | Localization of a device using multilateration | |
| Tang et al. | CLRIC: Collecting lane-based road information via crowdsourcing | |
| EP3238494B1 (en) | Selecting feature geometries for localization of a device | |
| KR20190082071A (en) | Method, apparatus, and computer readable storage medium for updating electronic map | |
| WO2016103026A2 (en) | Extracting feature geometries for localization of a device | |
| CN114509065B (en) | Map construction method, system, vehicle terminal, server and storage medium | |
| Konrad et al. | Localization in digital maps for road course estimation using grid maps | |
| US20230228593A1 (en) | Method, device and system for perceiving multi-site roadbed network and terminal | |
| CN106446785A (en) | Passable road detection method based on binocular vision | |
| CN112651991B (en) | Visual positioning method, device and computer system | |
| Dehbi et al. | Improving gps trajectories using 3d city models and kinematic point clouds | |
| WO2020118623A1 (en) | Method and system for generating an environment model for positioning | |
| CN116762094A (en) | Data processing method and device | |
| US20230025579A1 (en) | High-definition mapping | |
| CN117948963A (en) | Map construction method, device, equipment and storage medium | |
| IL281791B1 (en) | Method for improving orientation using a hand-held smartphone having an integral GPS device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |