CN118687582B

CN118687582B - A vehicle positioning method based on vehicle trajectory image matching

Info

Publication number: CN118687582B
Application number: CN202410565466.XA
Authority: CN
Inventors: 许舒晨; 孙永荣; 罗康
Original assignee: Nanjing University of Aeronautics and Astronautics
Current assignee: Nanjing University of Aeronautics and Astronautics
Priority date: 2024-05-09
Filing date: 2024-05-09
Publication date: 2025-07-25
Anticipated expiration: 2044-05-09
Also published as: CN118687582A

Abstract

The invention discloses a vehicle positioning method based on vehicle track imaging matching, which comprises the steps of loading a regional road network map according to the outline position of a vehicle at an initial moment, converting the regional road network map to a geographic coordinate system of northeast, calculating regional road network map boundaries to carry out whole-area adjustment, constructing a road network image pyramid substrate, imaging single road elements in the road network map, imaging all single road elements in the regional road network, fusing all level single road element images to construct a road network pyramid image, generating an initial vehicle track, converting the initial vehicle track to a geographic coordinate system, dividing the initial vehicle track into a long straight track and a curved track, imaging long straight road sections of the vehicle track until a long straight-clamped curved track is formed, and matching the regional road network image and the vehicle track image layer by layer to position the vehicle. The invention can position the vehicle under the condition that the initial positioning information of the vehicle is not clear.

Description

Vehicle positioning method based on vehicle track imaging matching

Technical Field

The invention belongs to the field of map matching and navigation positioning, and particularly relates to a vehicle positioning method based on vehicle track imaging matching.

Background

Among car navigation systems, satellite and inertial integrated navigation systems are widely used by virtue of their excellent positioning performance and high cost performance. However, in urban environments, satellite signals are easily shielded by high-rise buildings to cause positioning failure, and when the signals reach the ground after rebounding from the building elevation, multiple paths are formed to seriously influence positioning accuracy. The pure inertial positioning system can provide accurate positioning information in short time, but as the driving distance increases, the principle positioning errors of inertial navigation are continuously accumulated, and catastrophic positioning results can be seriously caused. Researchers can make up for the defects of the traditional integrated navigation system to a certain extent by introducing external positioning systems such as a wheel type odometer and the like for fusion positioning, but the building cost of the vehicle navigation system and the complexity of data processing are inevitably increased.

With the iterative upgrade of computer hardware and the rapid development of data processing technology, more and more researchers use cameras to realize vehicle positioning. The motion state of the vehicle is tracked by inputting a series of images into a visual odometer (Visual Odometry, VO for short). The visual odometer extracts characteristic elements in image frames, calculates a frame pose difference value through a characteristic matching relation between adjacent image frames, essentially belongs to one of dead reckoning methods, and needs to define the initial pose of the vehicle. With increasing time and distance travelled, small errors can accumulate gradually and in turn lead to a decrease in the accuracy of the visual odometer positioning. In the existing method, the vehicle pose is mostly recursively calculated by utilizing a visual odometer under the condition of determining the initial vehicle pose, and the accumulated error is restrained by introducing external constraint, so that the problem of vehicle positioning under the condition of undefined initial position is not considered.

Disclosure of Invention

The invention aims to provide a vehicle positioning method based on vehicle track imaging matching, which can convert the matching of a road network unordered point set and a vehicle track ordered point set into an image matching problem and can position a vehicle under the condition that the initial positioning information of the vehicle is not clear.

The vehicle positioning method based on the vehicle track imaging matching comprises the following steps:

(1) Loading an regional road network map according to the outline position of the vehicle at the initial moment;

(2) Converting the regional road network map to a geographic coordinate system of northeast days;

(3) Calculating regional road network map boundaries for whole-area adjustment, and constructing a road network image pyramid substrate;

(4) Imaging a single road element in the road network map;

(5) Imaging all single road elements in the regional road network, and fusing the single road element images of each level to construct a road network pyramid image;

(6) Generating an initial vehicle track by utilizing a visual odometer and converting the initial vehicle track into a geographic coordinate system;

(7) Dividing an initial vehicle track into a long straight track and a curved track according to the angular speed of the change of the inter-frame course angle;

(8) Repeatedly executing the step (6) and the step (7) until a long straight-clamping bending track is formed, and performing imaging treatment on a long straight road section of the vehicle track;

(9) The regional road network image and the vehicle track image are matched layer by layer to locate the vehicle.

Further, the regional road network map in the step (1) is a "point-line" topological map, and is composed of a single road element Way with an indefinite number, the Way is an ordered set composed of a series of nodes, roads are represented by broken line segments connected by nodes in the set, and the single Node element Node data comprises latitude lat, longitude lon and altitude alt where the single Node element Node data is located.

Further, the implementation process of the step (2) is as follows:

Converting LLA coordinates of internal nodes of all Way elements into corresponding geocentric fixed coordinates, and recording LLA coordinates (lat ₀,lon₀,alt₀) of first Node elements Node ₀ in a road network original data list, wherein ECEF coordinates (X ₀,Y₀,Z₀) are as follows:

Wherein R _N represents the principal radius of curvature of the reference ellipsoid, f is the reference ellipsoid flattening, and R _e is the equatorial plane long radius;

Converting ECEF coordinates of all nodes into northeast coordinates, selecting Node ₀ as a reference point, and calculating a conversion matrix from a geocentric fixed coordinate system to the northeast coordinates according to LLA coordinates

For the target point Node _i, the LLA coordinate is recorded as (lat _i,lon_i,alt_i), the ECEF coordinate is calculated as (X _i,Y_i,Z_i), and the ENU coordinate of the target point (E _i,N_i,U_i) is:

further, the implementation process of the step (3) is as follows:

Traversing the internal nodes of each road element Way in the regional road network, calculating the regional road network map boundary under the ENU coordinate system, and carrying out whole-area adjustment on the regional road network map boundary, wherein the regional road network map boundary comprises a North coordinate maximum value North _BMax, a North coordinate minimum value North _BMin, an East coordinate maximum value East _BMax and an East coordinate minimum value East _BMin;

Constructing a Num layer road network image pyramid substrate, and setting the single grid side length from 1 layer of the image pyramid substrate to the Num layer as The method comprises the steps of (1) increasing the number of grids proportionally, wherein pyramid substrate data are square node data of each example forming a substrate, the direction of increasing east coordinates is E-axis positive direction, the direction of increasing north coordinates is N-axis positive direction, the calculation of the square node coordinates of the ith row and the jth column of an nth layer is shown as a formula (7), and the value ranges of i and j are shown as a formula (8):

further, the implementation process of the step (4) is as follows:

Carrying out quadrilateral expansion on a single road element according to the road width, imaging the single road element according to the quadrilateral expansion result and the superposition condition of pyramid grids of all levels, and finally storing the single road element as a single-channel image, wherein the single grid in the pyramid grid is regarded as a pixel, the pixel value is between 0 and 255, wherein a value of 0 represents black, a value of 255 represents white, and the intermediate value is gray of different levels;

For the n-th layer pyramid substrate single square node and single quadrilateral expansion result, firstly judging the inclusion relation between four vertexes of the square node and the quadrilateral expansion result, for the upper left vertex P _LU, the four points of the quadrilateral are R _A、R_B、R_C and R _D respectively in anticlockwise order, if AndThe values of (a) are the same, namely four values are positive or negative, the determination point P _LU is positioned in the quadrangle, then the intersection point of each side line segment of the square node and each side line segment of the quadrangle is calculated, finally the polygonal area S _Pol formed by all intersection points and contained vertexes is calculated (if the sum of the intersection point and the contained vertexes is less than 3, S _Pol =0), the single square side length of the nth layer pyramid substrate is BD/2 ^n-1, the area is S _Rec＝BD²/2^2n-2, and the pixel value pixel of the square node is:

piexl=255-255*S_Pol/S_Rec (9)

and traversing square nodes in the pyramid substrate to calculate pixel values, and then imaging the level road elements.

Further, the implementation process of the step (5) is as follows:

The road network pyramid image fusion construction strategy is designed, for two cases of pixels to be fused at a single pixel position in the same hierarchy, if one pixel value is 255 in the two cases of pixels to be fused, which indicates that no road imaging result exists at the position, the other case of pixel value is directly used as a fused pixel value, if one pixel value is 0 in the two cases of pixels to be fused, which indicates that a complete road imaging result exists at the position, the two cases of pixels to be fused are directly used as the fused pixel value, if neither of the two cases of pixels to be fused is 0 or 255, which indicates that two cases of pixels to be fused have imaging results with different proportions at the position, the half fused pixel value is used as the fused pixel value, and the imaging results of each single road element in the regional road network map are traversed to form the whole regional road network pyramid image in a fusion mode.

Further, the implementation process of the step (7) is as follows:

According to the pose change of the frame output by the visual odometer, the angular speed dpsi of the change of the inter-frame heading angle is obtained, if dpsi < psi _sta, the frame is judged to belong to a long straight track, and if dpsi > psi _sta, the frame is judged to belong to a curved track, wherein psi _sta is a preset value.

The implementation process of the step (8) is as follows:

The imaging method of the vehicle track does not consider the problem of marginalization, and for the single square node and the single quadrilateral expansion result of the nth layer pyramid substrate, if the center point of the square node is contained by the quadrilateral expansion result, the pixel value of the square node is set to 0 to indicate that the vehicle passes through the area in the running process of the vehicle, and if the center point of the square node is not contained by the quadrilateral expansion result, the pixel value of the square node is set to 255 to indicate that the vehicle does not pass through the area in the running process of the vehicle.

Further, the implementation process of the step (9) is as follows:

For the road network image and the vehicle track image of the same level, traversing all positions from the upper left corner, and calculating the sum of numerical differences between the vehicle track image and pixels of the overlapping part of the vehicle track image and the vehicle track image in the road network image;

The traversal matching is preferably executed from the lower level of the pyramid, the image of the higher level shifts in a small range on the basis of the matching result of the lower level so as to accelerate the matching and positioning process, and the average value of the offset is taken as the final value of the matching and positioning result.

Compared with the prior art, the invention has the advantages that:

1. According to the invention, the initial vehicle running track provided by the visual odometer is used as a matching source, and the matching positioning of the vehicle can be completed under the condition that the approximate range of the initial position of the vehicle is known only by matching with the road network map;

2. The positioning method designed by the invention only depends on a 'point-line' road network map, does not need richer map information, the required data is road network node position information, and the required map data volume is small, so that the method is suitable for a large-scale scene;

3. According to the invention, the matching of the unordered point set of the road network and the ordered point set of the vehicle track is converted into the image matching problem, pyramid images of different levels and a matching positioning method are designed, and the real-time and accuracy requirements of positioning can be met simultaneously through the matching of the pyramid images of different levels;

4. the invention designs a regional road network map pyramid imaging method, which can effectively utilize the topological structure information of the 'point-line' road network map to construct road network images of different levels, and can represent the weight of the position in the road network by the size of a pixel value so as to provide a reference library for matching and positioning vehicles;

5. the invention designs the vehicle track segmentation method and the pyramid imaging method based on course change, which can simplify and efficiently utilize the original vehicle track data and provide a matching source for vehicle matching and positioning.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a map schematic diagram of OpenStreetMap areas;

FIG. 3 is a schematic diagram of a regional road network map;

FIG. 4 is a schematic diagram of a regional road network map in the northeast coordinate system;

FIG. 5 is a conceptual diagram of an image pyramid network;

FIG. 6 is a diagram of a road quadrilateral expansion result;

FIG. 7 is a schematic diagram of a road image pyramid;

FIG. 8 is a schematic diagram of a regional road network image pyramid;

FIG. 9 is a schematic diagram of an initial vehicle track segmentation result;

FIG. 10 is a schematic view of a long straight section vehicle track image pyramid;

Fig. 11 is a schematic diagram of a vehicle track imaging matching positioning result.

Detailed Description

The invention is described in further detail below with reference to the accompanying drawings.

As shown in FIG. 1, the invention provides a vehicle positioning method based on vehicle track imaging matching, which is used for positioning the initial position of a vehicle layer by layer according to regional road network and vehicle track pyramid imaging results, wherein the known conditions are the outline position and the initial attitude angle of the vehicle at the initial moment, and a longer straight road section is preferably selected for starting positioning in the initial stage. The specific process is as follows:

And step 1, loading the regional road network map according to the outline position of the vehicle at the initial moment.

Knowing the approximate location of the vehicle at the initial time, downloading the regional road network map from the OpenStreetMap website according to the location information. As shown in FIG. 2, the map of OpenStreetMap website area is composed of an indefinite number of single road elements Way, the Way is an ordered set composed of a series of nodes, the roads are represented by broken line segments connected by nodes in the set, and the single Node element Node data comprises latitude lat, longitude lon and altitude alt where the single Node element Node data are located. And selecting a region with higher density of a certain road network, wherein the loaded regional road network map is shown in fig. 3.

And 2, converting the regional road network map into a northeast geographic coordinate system.

The original data of OpenStreetMap website road network map nodes are node latitude and longitude height (LLA) coordinates, and in order to facilitate subsequent calculation, the method converts the original data into a regional northeast geographic coordinate system, and the conversion process is divided into two steps, which are all mature means in the industry and are necessary for complete explanation of contents. The first step, converting the LLA coordinates of all Way elements into corresponding earth centered earth anchor (ECEF) coordinates, taking the first Node element Node ₀ in the original data list of the road network as an example, recording the LLA coordinates as lat ₀,lon₀,alt₀, and the ECEF coordinates (X ₀,Y₀,Z₀) can be obtained by the formula (1), wherein R _N represents the principal radius of curvature of the reference ellipsoid, can be obtained by the formula (2), f= 0.0033528 represents the reference ellipsoid flattening rate, and R _e = 6378137 (unit: meters) represents the equatorial plane long radius:

The second step is to convert ECEF coordinates of all nodes into northeast (ENU) coordinates, select Node ₀ as reference point, calculate the transformation matrix from geocentric and geodetic coordinate system to northeast (northeast) coordinate system according to LLA coordinates The method comprises the following steps:

For the target Node _i, the LLA coordinate is denoted by lat _i,lon_i,alt_i, the ECEF coordinate is denoted by X _i,Y_i,Z_i, and the ENU coordinate (E _i,N_i,U_i) of the target Node is denoted by equation (4). Converting a regional road network map under the LLA coordinate system shown in FIG. 3 to an ENU coordinate system as shown in FIG. 4.

And 3, calculating regional road network map boundaries and carrying out whole-area adjustment.

The invention traverses the road element Way internal nodes in the regional road network and calculates the regional road network map boundary under the ENU coordinate system. Road network map boundaries are East _Min、East_Max、North_Min and North _Max, where East _Min and East _Max represent East boundary minimum and maximum values, and North _Min and North _Max represent North boundary minimum and maximum values. Taking the regional road network shown in fig. 4 as an example, eastern boundaries thereof are East _Min = -141.923 and East _Max = 553.884, respectively, and North _Min = -308.434 and North _Max = 664.127, respectively. In order to facilitate the construction of the road network image pyramid network in proportion, the whole area adjustment is carried out on the road network boundary. The new boundary range after the whole area adjustment should contain the original road network, and the boundary value should be an integer multiple of the preset value Boulen. for the boundary minimum, taking East _Min as an example, the boundary East _BMin after the whole area adjustment is calculated as shown in a formula (5), wherein floor () represents a downward rounding function, north _BMin is calculated as a rule, and for the boundary maximum, taking East _Max as an example, the boundary East _BMax after the whole area adjustment is calculated as shown in a formula (6), wherein ceil () represents an upward rounding function, north _BMax is calculated as a rule:

East_BMin＝floor(East_Min)*Boulen (5)

East_BMax＝ceil(East_Max)*Boulen (6)

Taking Boulen =40, after the whole area of each boundary of the regional road network shown in fig. 4 is adjusted, the eastern boundary is East _BMin = -160 and East _BMax =560, and the North _BMin = -320 and North _BMax =680.

And 4, constructing an image pyramid substrate.

The invention constructs a Num layer road network image pyramid substrate, and sets the single grid side length from 1 layer of the image pyramid substrate to the Num layer asThe pyramid grid concept formed is shown in fig. 5. The pyramid substrate data required by the invention is square node data of each example of the composition substrate, the left lower corner of the pyramid substrate is taken as an original point, the increasing direction of the east coordinate is E-axis positive direction, the increasing direction of the north coordinate is N-axis positive direction, the calculation of the square node coordinates of the ith row and the jth column of the nth layer is shown as a formula (7), and the value ranges of i and j are shown as a formula (8):

The number of pyramid base grids of the low level is small, the number of grids needed to be traversed in the matching and positioning process is small, the initial matching process is facilitated to be accelerated, and the positioning hot zone is fast determined, and the number of pyramid grids of the high level is large, although the matching and positioning time is long, the hot zone positioning result of the low level grid can be accelerated, so that a more accurate positioning result can be obtained.

And 5, imaging a single road element in the road network map.

The single road element in the original road network map is an ordered node set, and the road shape is approximately represented by a broken line segment formed by sequentially connecting all nodes in the set. The point-line topological road form has the advantages of less required storage amount, suitability for road network modeling in a large-scale scene, difficulty in determining the corresponding relation between road nodes and vehicle track points in the matching positioning process, and low positioning accuracy and even incapability of positioning.

The invention converts the matching of a road network unordered point set and a vehicle track ordered point set into an image matching problem, and firstly, pyramid images of all road elements are required to be constructed as matching references. For a single road element, it is first quadrilateral expanded to further simulate the real road shape. The quadrilateral expansion is a road simulation method provided by related researchers in the industry, and the rectangular expansion and the quadrilateral expansion are sequentially carried out on the node connecting lines according to the road width, so that the road shape can be effectively restored. Taking the road shown by the rectangular frame of fig. 4 as an example, the result of the square expansion of the road is shown in fig. 6.

According to the single road quadrilateral expansion result and the superposition condition of the pyramid substrate, single road elements are imaged and stored as single-channel images. The single channel image is composed of a two-dimensional matrix, the pixel value of each pixel point represents the color depth of the single channel image, the pixel value is between 0 and 255, wherein 0 represents black, 255 represents white, and the intermediate value is gray with different levels. For a single square node and a single quadrilateral expansion result of the nth layer pyramid substrate, firstly judging the inclusion relation between four vertexes of the square node and the quadrilateral expansion result. Taking the upper left vertex as an example, it is denoted as P _LU, the four points of the quadrangle are R _A、R_B、R_C and R _D respectively in the counterclockwise order, if AndThe values of (a) are the same, namely the four values are positive or negative, the determination point P _LU is positioned in the interior of the quadrangle, then the intersection points of the line segments of each side of the square node and the line segments of each side of the quadrangle are calculated, and finally the polygonal area S _Pol formed by all the intersection points and the contained vertexes is calculated (if the sum of the intersection points and the contained vertexes is less than 3, S _Pol =0). The single square side length of the nth layer pyramid substrate is BD/2 ^n-1, the area is S _Rec＝BD²/2^2n-2, and the pixel value pixel of the square node is calculated by:

piexl=255-255*S_Pol/S_Rec (9)

And traversing square nodes in the pyramid substrate to calculate pixel values, and then imaging the level road elements. The square nodes positioned in the quadrilateral expansion have large overlapping areas, the pixels are black, the possibility of representing the vehicle running in the area is high, the square nodes positioned at the edges of the quadrilateral expansion have small overlapping areas, the pixels are gray with different degrees, the smaller the overlapping areas are, the pixels are closer to white, the possibility of representing the vehicle running in the area is lower, and the actual vehicle running situation is met. The level 1 to level 4 imaging results of the roads shown by the rectangular frame in fig. 4 are shown in fig. 7, and the level (a) to (d) in fig. 7 correspond to the level 1 to level 4 bases of the pyramid, respectively.

And 6, fusing the imaging results of each single-path network element to form a regional road network pyramid image.

And traversing the imaging results of each single road element in the regional road network map, and fusing to form an integral regional road network pyramid image. For two cases of pixels to be fused at a single pixel position in the same hierarchy, the invention designs a fusion construction strategy as follows:

If one pixel value of the two pixel points to be fused is 255, the position of the pixel point is not provided with a road imaging result, and the other pixel value is directly used as the fused pixel point value.

If one pixel value is 0 in the two cases of pixels to be fused, the position is indicated to have a complete road imaging result, and 0 is directly used as the fused pixel value.

If the two pixel points to be fused are not 0 or 255, the two pixel points are represented to have different proportions of imaging results at the positions, and the half fused pixel value is used as the fused pixel point value.

The road elements in the regional road network shown in fig. 4 are subjected to imaging processing one by one and are fused, the fusion results of the layers 1 to 4 are shown in fig. 8, and the pyramid 1 to 4-level images respectively correspond to (a) to (d) in fig. 8.

And 7, generating an initial vehicle track by utilizing a visual odometer and converting the initial vehicle track into a geographic coordinate system.

The visual odometer is an industry mature means, generally, a single or a plurality of cameras are used for obtaining a frame pose difference value through the steps of image acquisition, feature extraction, frame pose recovery, pose optimization and the like, and the vehicle pose is recovered in a recursive mode to generate a vehicle track. The invention assumes that the initial position of a vehicle is the first road element head node position of a regional road network map road set, and obtains the track { T _wm,m＝0,1,…,N_f } of the vehicle under a World coordinate system (initial camera coordinate system) according to a visual odometer, wherein W represents a World coordinate system World system, and N _f represents the number of image frames processed by the visual odometer. Track { T _wm,m＝0,1,…,N_f } warpIs converted into a track { T _bm,m＝0,1,…,N_f } under the body system (b system), whereinRepresenting a transformation matrix from the world coordinate system to the body system, obtained by actual measurement, including rotation vectorsTranslation vectorTrack { T _bm,m＝0,1,…,N_f } warpConverting into a track { T _gm,m＝0,1,…,N_f } under a geographic system, whereinRepresenting a transformation matrix from the body system to the geographic system, obtained from known initial attitude angles of the vehicle, including a rotation matrixTranslation matrix

And 8, dividing the initial vehicle track.

The invention designs a vehicle track segmentation criterion, and segments a long straight section and a curved section of an initial vehicle track. The invention takes the north direction as the heading angle 0 degree reference, takes the clockwise direction as the heading angle increasing direction, and the heading angle value range is 0-360 degrees. The visual odometer obtains an initial vehicle track { T _gm,m＝0,1,…,N_f } under a geographic coordinate system, and records that coordinates of vehicles under the geographic coordinate system at the time T-1 and the time T are P _t-1＝(E_t-1,N_t-1,U_t-1) and P _t＝(E_t,N_t,U_t), wherein E _t-1 and E _t represent east coordinates, N _t-1 and N _t represent north coordinates, and U _t-1 and U _t represent sky coordinates. Defining the heading direction of a vehicle at the moment t as a vectorThe difference between the coordinates of the northeast and north directions is as follows:

(E_△,N_△)＝(E_t-E_t-1,N_t-N_t-1) (10)

Vector quantity Heading courseThe calculation can be performed by the formula (11), wherein dzero takes on a value of 0.00000001.

For a frame to be solved, calculating the heading direction difference dpsi between the frame and the previous moment, if dpsi < psi _sta, judging that the frame belongs to a long straight track, and if dpsi > psi _sta, judging that the frame belongs to a curved track, wherein psi _sta is a preset value. It should be noted that, during driving, there may be a short-time jerk of the vehicle to avoid an emergency, and the present invention only determines that a road section of a continuous multiframe belonging to a curved track is a curved road section. The initial vehicle track is divided, and the division result is shown in fig. 9.

And 9, forming a long straight clamp bending track, and imaging a long straight section of the vehicle track.

For a single long straight track, the road section to which the vehicle belongs can be matched according to the heading of the single long straight track, but the specific position of the single long straight track in the road section cannot be determined. After the curved track, the running direction of the vehicle is changed, and long straight tracks in front and behind the curved track can form constraint in multiple directions to match and position the specific position of the vehicle. The invention repeatedly executes the step 7 and the step 8 until a long straight clamp bending track is formed, wherein a typical long straight clamp bending track is shown in fig. 9, and the two long straight tracks before and after the bending track have obvious course differences.

Similar to the road network pyramid imaging method, the boundary of the vehicle track map is calculated according to the method of the step 3, the whole area is adjusted, the vehicle track image pyramid substrate is constructed according to the method of the step 4, and the vehicle width is used for replacing the road width to extend the quadrangle of the long straight track. The long straight track has stability, can provide stable information to be matched, and the curved track is greatly influenced by the operation habit of a driver and the actual road conditions, and cannot provide stable information to be matched, so that the information to be matched is abandoned. The imaging method of the vehicle track is different from the imaging method of a single road element, and the problem of marginalization is not required to be considered. For the single square node and the single quadrilateral expansion result of the nth layer pyramid substrate, if the central point of the square node is contained by the quadrilateral expansion result, setting the pixel value of the square node to be 0, and the pixel to be black, indicating that the area passes through in the vehicle driving process, and if the central point of the square node is not contained by the quadrilateral expansion result, setting the pixel value of the square node to be 255, and the pixel to be white, indicating that the area does not pass through in the vehicle driving process. The "long straight-clip curved" track shown in fig. 9 is imaged, and the imaging results of the levels 1 to 4 are shown in fig. 10, where (a) to (d) in fig. 10 correspond to the pyramid 1 to 4 level substrates, respectively.

And 10, matching the regional road network and the vehicle track pyramid image layer by layer to position the vehicle.

The method traverses regional road network pyramid images, and takes the corresponding-level vehicle track images to match and position the vehicles. For the road network image and the vehicle track image of the same level, traversing all positions from the upper left corner, and calculating the sum of numerical differences between the vehicle track image and pixels of the overlapping part of the vehicle track image and the vehicle track image in the road network image. The invention designs that traversal matching is performed from the low level of the pyramid, and the image of the high level shifts in a small range based on the matching result of the low level so as to accelerate the matching positioning process. Since the width of the vehicle is smaller than the road width, there are multiple sets of pixel value differences that sum the same during the matching process and are the minimum during traversal. The invention takes the average value of the deviation amount, namely the final value of the matching positioning result, and the center coordinate of the square node corresponding to the pixel point corresponding to the starting point of the vehicle track is the initial position of the vehicle. In road networks with high similarity, there may be multiple matching results for a simple "long straight-clip curved" trajectory. Taking the road network shown in fig. 4 and the vehicle track shown in fig. 9 as an example, there are a plurality of matching positioning results shown by circle marks in fig. 11. By increasing the number of subsequent long straight tracks, false matching items can be removed, as shown in fig. 11, three sections of long straight tracks are mutually constrained, the initial position of a vehicle can be effectively matched and positioned, a part with higher transparency in the figure is a road network, and a part with lower transparency is a matching and positioning result of a vehicle track on the road network.

Claims

1. The vehicle positioning method based on the vehicle track imaging matching is characterized by comprising the following steps of:

(4) Imaging a single road element in the road network map;

(9) Matching regional road network images and vehicle track images layer by layer to locate vehicles;

the implementation process of the step (4) is as follows:

For the n-th layer pyramid substrate single square node and single quadrilateral expansion result, firstly judging the inclusion relation between four vertexes of the square node and the quadrilateral expansion result, for the upper left vertex P _LU, the four points of the quadrilateral are R _A、R_B、R_C and R _D respectively in anticlockwise order, if AndThe method comprises the steps of determining that a point P _LU is positioned in a quadrangle when four values are positive or negative, calculating the intersection point of each side line segment of a square node and each side line segment of the quadrangle, finally calculating the polygonal area S _Pol formed by all intersection points and contained vertexes, setting the single square side length of an nth layer pyramid substrate as BD/2 ^n-1, setting the area as S _Rec＝BD²/2^2n-2, and setting the pixel value pixel of the square node as:

piexl =255-255×s _Pol/S_Rec (9) calculates the pixel value of the square node in the pyramid base, so as to image the level road element.

2. The vehicle positioning method based on vehicle track imaging matching of claim 1, wherein the regional road network map in the step (1) is a 'point-line' topology map, and is composed of a single road element Way with an indefinite number, the Way is an ordered set composed of a series of nodes, roads are represented by broken line segments connected by nodes in the set, and the single Node data comprises latitude lat, longitude lon and altitude alt where the single Node data are located.

3. The vehicle positioning method based on vehicle track imaging matching of claim 1, wherein the implementation process of the step (2) is as follows:

Converting LLA coordinates of all Way element internal nodes into corresponding geocentric fixed coordinates, and recording LLA coordinates (lat ₀,lon₀,alt₀) of first nodes ₀ in a road network original data list, wherein ECEF coordinates (X ₀,Y₀,Z₀) are as follows:

4. The vehicle positioning method based on vehicle track imaging matching of claim 1, wherein the implementation process of the step (3) is as follows:

5. the vehicle positioning method based on vehicle track imaging matching of claim 1, wherein the implementation process of the step (5) is as follows:

6. The vehicle positioning method based on vehicle track imaging matching of claim 1, wherein the implementation process of the step (7) is as follows:

7. The vehicle positioning method based on vehicle track imaging matching of claim 1, wherein the implementation process of the step (8) is as follows:

8. The vehicle positioning method based on vehicle track imaging matching of claim 1, wherein the implementation process of the step (9) is as follows: