A fast crowd-sourced map alignment method based on spatial coordinate decomposition

Abstract

The application relates to a fast crowd-sourced map alignment method based on spatial coordinate decomposition, which comprises the following steps: classifying each marker in a road scene according to a distribution rarity degree, constructing a marker feature set M1 and M2 of a map one and a map two to be aligned based on a classification result; constructing a translation cost function based on the similarity between each marker contained in the feature set M1 and the feature set M2; and obtaining the best matching between the map one and the map two when the corresponding translation transformation of the minimum value of the translation cost function is calculated. The 1500m length crowd-sourced feature map can be aligned within 50ms, and meanwhile, the result can be guaranteed to be globally optimal.

Description

Technical Field

[0001] This invention relates to the field of crowdsourced mapping technology, and in particular to a fast crowdsourced map alignment method based on spatial coordinate decomposition. Background Technology

[0002] Crowdsourced mapping offers advantages such as low cost and high data freshness. However, the fusion and updating of crowdsourced data depends on the identification, matching, and alignment of the collected map fragments. Map alignment is a complex pattern recognition problem, and its high computational cost, low accuracy, and low recall have always been pain points hindering the large-scale application of crowdsourced mapping.

[0003] Feature maps using sparse representation have small data volumes and low bandwidth requirements for data upload, making them widely used in large-scale crowdsourced mapping. However, sparse representation suffers from detail loss, and the involvement of image segmentation and semantic recognition algorithms introduces errors in shape, existence, and type.

[0004] Crowdsourcing models are limited by the cost of data collection, often deploying vehicles with mediocre performance and low computing power for sensing and processing. This leads to data generated by these vehicles commonly suffering from missing features, incorrect data types, and geometric distortions, further increasing the difficulty of map alignment.

[0005] How to extract effective patterns from such high-error data and complete pattern matching in an efficient manner is a key problem that crowdsourced map matching algorithms need to solve. Summary of the Invention

[0006] This invention addresses the technical problems existing in the prior art by providing a fast crowdsourced map alignment method based on spatial coordinate decomposition, which can align a 1500-meter-long crowdsourced feature map within 50 milliseconds, while ensuring that the result is globally optimal.

[0007] According to a first aspect of the present invention, a fast crowdsourced map alignment method based on spatial coordinate decomposition is provided, comprising:

[0008] Step 1: Classify the various landmarks in the road scene according to their distribution sparseness, and construct the alignment target based on the classification results. Figure One peacefully Figure Two The sets of marker features M1 and M2;

[0009] Step 2: Construct a translation cost function based on the similarity between the various markers contained in the feature set M1 and feature set M2;

[0010] Step 3, calculate the translation transformation corresponding to the minimum value of the translation cost function as the ground Figure One peacefullyFigure Two between the best match.

[0011] On the basis of the above technical solutions, the application can be further improved as follows.

[0012] Optionally, the process of constructing the feature sets M1 and M2 of the two maps to be aligned in step 1 comprises:

[0013] Step 101, classifying each feature in the road scene into a first type of feature, a second type of feature and a third type of feature according to the distribution rarity from high to low;

[0014] Step 102, uniformly representing each feature in the first type of feature, and respectively representing each feature in the second type of feature according to the type;

[0015] Step 103, constructing the feature set M1 and the feature set M2 based on the representation method in step 102, with the geographical position coordinates of each feature.

[0016] Optionally, the first type of feature comprises traffic lights and signboards; and the second type of feature comprises stop lines, zebra crossings and road arrow groups.

[0017] The feature set M1 comprises a1 stop lines, b1 zebra crossings, c1 road arrows and d1 traffic lights or signboards.

[0018] The feature set M2 comprises a2 stop lines, b2 zebra crossings, c2 road arrows and d2 traffic lights or signboards.

[0019] wherein, and respectively represent the a1th stop line, the b1th zebra crossing, the c1th road arrow and the d1th traffic light or signboard in the first map. Figure One and respectively represent the a2th stop line, the b2th zebra crossing, the c2th road arrow and the d2th traffic light or signboard in the second map. Figure Two

[0020] Optionally, the step 1 further comprises:

[0021] calculating the minimum circumscribed rectangle B1 containing each feature in the feature set M1;

[0022] calculating the angle θ between the long side of the rectangle B1 and the x-axis of the Euclidean space;

[0023] rotating the first map clockwise by θ degrees. Figure One Figure Two rotating the second map clockwise by θ degrees.

[0024] Optionally, the association cost function is a one-to-one association cost function. ​​​

[0025] The land Figure One The marker f in 1,i Corresponding to the land Figure Two The marker f in 1,j The one-to-one association cost function is:

[0026] c(i,j)=(x i -x j ) 2 +(y i -y j ) 2 ;

[0027] Among them, (x i ,y i ) is f 1,i The coordinates, (x j ,y j ) is f 1,j The coordinates.

[0028] Optionally, the association cost function is a many-to-many association cost function;

[0029] Based on the one-to-one association cost function, the Hungarian algorithm is used to obtain the many-to-many association with the lowest cost, resulting in the following many-to-many association cost function:

[0030]

[0031] Optionally, the association cost function is a one-to-one association cost function of coordinate decomposition or a many-to-many association cost function of coordinate decomposition.

[0032] The one-to-one association cost function for x-axis coordinate decomposition is:

[0033]

[0034] The many-to-many association cost function for x-axis coordinate decomposition is:

[0035]

[0036] The one-to-one correlation cost function for y-axis coordinate decomposition is:

[0037] c y (i,j)=(y i -y j ) 2 ;

[0038] The many-to-many association cost function for y-axis coordinate decomposition is:

[0039]

[0040] Optionally, the correlation cost function is a translation cost function; the translation cost function comprises: an x-axis translation cost function and a y-axis translation cost function.

[0041] The x-axis translation cost function is:

[0042] CT x (T dx,dy ) = C x (M1, T dx,0 (M2));

[0043] The y-axis translation cost function is:

[0044] CT y (T dx,dy ) = C y (M1, T 0,dy (M2))

[0045] Wherein, T dx,dy (M2) represents a translation operation T dx,dy acting on the ground Figure Two , so that the coordinate values of all features in the ground Figure Two change.

[0046] Optionally, the correlation cost function is a weighted translation cost function; the calculation process of the weighted translation cost function comprises:

[0047] According to four features, the ground Figure One and the ground Figure Two are divided into four layers, and the x-axis offset cost CT x,St (T dx,dy ), CT y,St (T dx,dy ), CT x,Cr (T dx,dy ), CT x,Si (T dx,dy ) and the y-axis offset cost CT y,St (T dx,dy ), CT y,Cr (T dx,dy ), CT y,Ar (T dx,dy ), CT y,Si (T dx,dy ) are respectively calculated in the four layers.

[0048] The weighted translation cost function is obtained as follows:

[0049]

[0050]

[0051] wherein, η x,St , η x,Cr , η x,Ar , η x,Si are translation cost weights of the four features in x-axis respectively; η y,St , η y,Cr , η y,Ar , η y,Si are translation cost weights of the four features in y-axis respectively.

[0052] Optionally, the step 3 comprises:

[0053] Step 301, according to the value of WCT x (T dx,dy ), search the optimal x-axis translation dx1 between [-40, 40] with a step of 0.1 meter; search the optimal x-axis translation dx2 between [-4, 4] near dx1 with a step of 0.01 meter;

[0054] Step 302, according to the value of WCT y (T dx,dy ), search the optimal y-axis translation dy1 between [-40, 40] with a step of 0.1 meter; search the optimal y-axis translation dy2 between [-4, 4] near dy1 with a step of 0.01 meter;

[0055] Step 303, repeat the steps 301-302 for a certain number of times to obtain the translation transformation T dx2,dy2 ;

[0056] Step 304, apply the translation transformation T dx2,dy2 to the ground Figure Two , that is, obtain the optimal matching between the ground Figure One and the ground Figure Two .

[0057] The application provides a fast crowd-sourced map alignment method based on spatial coordinate decomposition, constructs three kinds of combination features with adjustable parameters, and uses the combination features for map feature point matching; compared with general features, the combination features are more unique, and can effectively resist the periodicity of road scenes; the method reduces the probability of false matching, and increases the robustness of the algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 The flow chart of the fast crowd-sourced map alignment method based on spatial coordinate decomposition provided by the application;

[0059] Figure 2 The schematic diagram for determining the minimum circumscribed rectangle and the included angle provided by the embodiment of the application;

[0060] Figure 3 The rotated map provided by the embodiment of the present application Figure One and the map Figure Two are shown in the schematic diagram. DETAILED DESCRIPTION

[0061] The principles and characteristics of the present application are described below in conjunction with the drawings, and the examples are used only to explain the present application and are not intended to limit the scope of the present application.

[0062] Figure 1 The flowchart of an embodiment of a fast crowd-sourced map alignment method based on spatial coordinate decomposition provided by the present application is shown in Figure 1 The alignment method comprises:

[0063] Step 1: classify each landmark in the road scene according to the distribution rarity, and construct the landmark feature set M1 and M2 of the maps to be aligned based on the classification results. Figure One and the map Figure Two .

[0064] Step 2: construct a translation cost function based on the similarity between each landmark contained in the feature set M1 and the feature set M2.

[0065] Step 3: calculate the minimum value of the translation cost function, and the corresponding translation transformation is the best match between the map Figure One and the map Figure Two .

[0066] When there are large errors in the shape and type of the feature map, and there are repeated patterns, the conventional matching algorithm is easy to be insufficient in error tolerance, or cannot find an effective pattern, or fall into a local optimum. Some methods that try to search for a global optimal solution may have high computational requirements and cannot meet the fast mapping demand of a large amount of data.

[0067] To solve this problem, the present application provides a fast crowd-sourced map alignment method based on spatial coordinate decomposition, which can align a crowd-sourced feature map of 1500 meters in length within 50 milliseconds, while ensuring that the result is globally optimal.

[0068] Embodiment 1

[0069] The embodiment 1 provided by the present application is an embodiment of a fast crowd-sourced map alignment method based on spatial coordinate decomposition provided by the present application, and in combination with Figure 2 It can be known that the embodiment of the alignment method comprises:

[0070] Step 1: classify each landmark in the road scene according to the distribution rarity, and construct the landmark feature set M1 and M2 of the maps to be aligned based on the classification results. Figure One and the map Figure Two .

[0071] In one possible embodiment, the process of constructing feature sets M1 and M2 of the two maps to be aligned in step 1 includes:

[0072] Step 101: Classify the various landmarks in the road scene into three categories according to their distribution rarity, from high to low: the first category, the second category, and the third category.

[0073] Step 102: Represent each feature in the first category in a unified manner, and represent each feature in the second category separately according to its type.

[0074] Step 103: Based on the representation method in step 102, construct feature set M1 and feature set M2 using the geographic coordinates of each feature.

[0075] In practice, the road crowdsourcing data matching algorithm processes two maps (denoted as maps 1, 2, and 3). Figure One peacefully Figure Two These data are generated by different data collection vehicles collecting data on the same road scene at different times. The road scene consists of road signs distributed in a narrow space (a typical road scene is 1500 meters long and 15 meters wide).

[0076] Based on their sparseness, road landmarks in this algorithm are divided into the following three categories:

[0077] Lane lines, curbs, and lampposts are mundane landmarks that appear periodically and repeatedly. Registration of these landmarks is prone to errors, so this algorithm ignores these features.

[0078] Stop lines, zebra crossings, and road arrows are relatively rare landmarks. Although they may repeat, they are usually spaced further apart (stop lines and zebra crossings typically repeat every 200 meters, while road arrows repeat every 50 meters).

[0079] Traffic lights and signs are the rarest types of signs, and they generally do not appear periodically, making them the least likely to be mismatched among all types of signs.

[0080] Stop lines are denoted as St, zebra crossings as Cr, road arrows as Ar, and traffic lights and signs are indiscriminately denoted as Si. Generally, features are denoted as F.

[0081] F = {St,Cr,Ar,Si}.

[0082] Features St, Cr, Ar, and Si are all represented as points in two-dimensional Euclidean space. The coordinates of these points are the x and y coordinates of the geographical location of the feature as perceived by the data acquisition vehicle. i ,y i).

[0083] Therefore Figure One peacefully Figure Two It can be represented by the following two sets.

[0084]

[0085]

[0086] in, and Respectively represent the land Figure One The a1th stop line, the b1st zebra crossing, the c1st road arrow, and the d1st traffic light or sign. and Respectively represent the land Figure Two The a2nd stop line, b2nd zebra crossing, c2nd road arrow, and d2nd traffic light or sign are included. M1 consists of a1 X features, b1 C features, c1 A features, and d1 S features; M2 consists of a2 X features, b2 C features, c2 A features, and d2 S features.

[0087] In one possible embodiment, step 1 is followed by:

[0088] The minimum bounding rectangle B1 of each feature contained in the feature set M1 is calculated.

[0089] The road scene is a long and narrow section, so the minimum bounding rectangle algorithm (a publicly available algorithm in most set operation libraries) can be used to find the minimum bounding rectangle B1 of M1.

[0090] Calculate the angle θ between the longer side of rectangle B1 and the x-axis of Euclidean space; for example... Figure 2 The diagram shown is a schematic diagram of determining the minimum circumscribed rectangle and included angle according to an embodiment of the present invention.

[0091] land Figure One peacefully Figure Two Rotate both clockwise by θ degrees to obtain M1' and M2', whose major axes are approximately parallel to the x-axis. For example... Figure 3 The image shown is a rotated ground provided in an embodiment of the present invention. Figure One peacefully Figure Two The diagram shows that the two maps are not aligned due to measurement errors during the data collection process.

[0092] Step 2: Construct a translation cost function based on the similarity between the various markers contained in feature set M1 and feature set M2.

[0093] In a possible implementation mode, the association cost function can be a one-to-one association cost function; this function is used to describe the similarity between two features associated together, and the higher the association cost, the less similar the two features are, and vice versa, the lower the association cost, the more similar the two features are.

[0094] The marker f Figure One in the ground 1,i corresponds to the marker f Figure Two in the ground 1,j , and the one-to-one association cost function is:

[0095] c(i,j)=(x i -x j ) 2 +(y i -y j ) 2 .

[0096] Where (x i , y i ) is the coordinate of f 1,i , and (x j , y j ) is the coordinate of f 1,j .

[0097] In a possible implementation mode, the association cost function can be a many-to-many association cost function; when there are multiple one-to-one association relationships, the appropriate distribution of corresponding matches is selected so that the total cost of all corresponding is the lowest, which is a typical bipartite matching problem, and the Hungarian algorithm can be used to achieve it. The Hungarian algorithm is a public algorithm. According to the one-to-one association cost function, the Hungarian algorithm is called to obtain the many-to-many association with the lowest cost.

[0098] According to the one-to-one association cost function, the Hungarian algorithm is called to obtain the many-to-many association with the lowest cost, and the many-to-many association cost function is:

[0099]

[0100] In a possible implementation mode, the association cost function can be a one-to-one association cost function or a many-to-many association cost function. The cost function defined above depends on both the x coordinate and the y coordinate, which brings higher algorithm complexity to the search for the optimal solution. Therefore, the algorithm decomposes the cost function into x coordinate and y coordinate, and can sample and search the x coordinate and y coordinate respectively to find the global optimal solution.

[0101] Specifically, the one-to-one association cost function of the x-axis coordinate decomposition is:

[0102]

[0103] The cost function of the coordinate-decomposed many-to-many association cost function is:

[0104]

[0105] The cost function of the coordinate-decomposed one-to-one association cost function is:

[0106] c y (i,j) = (y i -y j ) 2 .

[0107] The cost function of the coordinate-decomposed many-to-many association cost function is:

[0108]

[0109] The basis for this is that the distribution of most feature markers does not have periodicity in at least one coordinate axis, so that the extreme value search on each axis alone can find the appropriate point.

[0110] In a possible implementation mode, the association cost function can be a translation cost function; the translation cost function includes: an x-axis translation cost function and a y-axis translation cost function.

[0111] Suppose there is a translation operation T dx,dy acting on the ground Figure Two , so that the coordinate values of all features in the ground Figure Two are changed, denoted as T dx,dy (M2).

[0112] The coordinate-decomposed many-to-many association cost function between M1 and T dx,dy (M2) is defined as the translation cost function.

[0113] The x-axis translation cost function is:

[0114] CT x (T dx,dy ) = C x (M1, T dx,0 (M2)).

[0115] The y-axis translation cost function is:

[0116] CT y (T dx,dy ) = C y (M1, T 0,dy (M2)).

[0117] In a possible implementation mode,

[0118] The correlation cost function is a weighted translation cost function; the calculation process of the weighted translation cost function comprises:

[0119] According to the four features, the ground Figure One and the ground Figure Two are divided into four layers, and the x-axis offset cost CT x,St (T dx,dy ), CT y,St (T dx,dy ), CT x,Cr (T dx,dy ), CT x,Si (T dx,dy ) and the y-axis offset cost CT y,St (T dx,dy ), CT y,Cr (T dx,dy ), CT y,Ar (T dx,dy ), CT y,Si (T dx,dy ) are respectively calculated in the four layers.

[0120] CT x,St (T dx,dy ) = C x ({St i |St∈M1},{St j |St∈T dx,0 (M2)})

[0121] CT y,St (T dx,dy ) = C y ({St i |St∈M1},{St j |St∈T dx,0 (M2)})

[0122] CT x,Cr (T dx,dy ) = C x ({Cr i |Cr∈M1},{Cr j |Cr∈T dx,0 (M2)})

[0123] …

[0124] The cost functions are weighted and summed according to the weight of each feature, and the weighted translation cost function is:

[0125]

[0126]

[0127] Wherein, η x,St , ηx,Cr , η x,Ar , η x,Si are respectively translation cost weight of four features in x axis; η y,St , η y,Cr , η y,Ar , η y,Si are respectively translation cost weight of four features in y axis.

[0128] The following table shows the weight value table provided by the embodiment of the application:

[0129]

[0130] Step 3, the translation transformation corresponding to the minimum value of the translation cost function is calculated as the best matching between the ground Figure One and the ground Figure Two .

[0131] In a possible embodiment, step 3 includes:

[0132] Step 301, according to the value of WCT x (T dx,dy ), the best x axis translation dx1 between [-40, 40] is searched with a step distance of 0.1 meters; the best x axis translation dx2 between [-4, 4] near dx1 is searched with a step distance of 0.01 meters;

[0133] Step 302, according to the value of WCT y (T dx,dy ), the best y axis translation dy1 between [-40, 40] is searched with a step distance of 0.1 meters; the best y axis translation dy2 between [-4, 4] near dy1 is searched with a step distance of 0.01 meters;

[0134] Step 303, steps 301-302 are repeatedly executed for a certain number of times to obtain the translation transformation T dx2,dy2 .

[0135] In a specific implementation, steps 301-302 can be repeatedly executed for three times.

[0136] Step 304, the translation transformation T dx2,dy2 is applied to the ground Figure Two , that is, the best matching between the ground Figure One and the ground Figure Two is obtained.

[0137] The embodiment of the application provides a fast crowd-sourced map alignment method based on space coordinate decomposition, constructs three kinds of combination features with adjustable parameters, and uses the combination features for map feature point matching; compared with general features, the combination features are more unique, and can effectively resist the periodicity of a road scene; the method reduces the probability of false matching, and increases the robustness of an algorithm.

[0138] It should be noted that the description of the various embodiments has been presented for purposes of clarity and that it is not necessary to describe each and every embodiment separately or enumerate all its possible variations. It will be apparent to those skilled in the art that additional embodiments can be practiced which depart from the specific details of the described embodiments.

[0139] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system, or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0140] The present application is described in reference to the flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure One one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure One means for performing the function of one or more of the steps in the flowchart illustrations and / or block diagrams.

[0141] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart illustrations and / or block diagrams block or blocks. Figure One one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure One means for performing the function of one or more of the steps in the flowchart illustrations and / or block diagrams.

[0142] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart illustrations and / or block diagrams block or blocks. Figure One one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure One means for performing the function of one or more of the steps in the flowchart illustrations and / or block diagrams.

[0143] While the preferred embodiments of the application have been described, additional variations and modifications can be made to these embodiments by those skilled in the art once they have the benefit of the present disclosure without departing from the spirit and scope of the application. Accordingly, it is intended that the appended claims include all such modifications and variations as fall within the scope of the present application.

[0144] It is apparent that those skilled in the art can make various changes and modifications to the application without departing from the spirit and scope of the application. It is therefore intended that the present application cover all such changes and modifications that are within its scope.

Claims

1. A fast crowdsourced map alignment method based on spatial coordinate decomposition, characterized in that, The alignment method includes: Step 1: Classify the various landmarks in the road scene according to their distribution sparseness, and construct the landmark feature sets of Map 1 and Map 2 to be aligned based on the classification results. and ; Step 2, based on the feature set and feature set The similarity between the included markers is used to construct the translation cost function; Step 3: Calculate the translation transformation corresponding to the minimum value of the translation cost function as the best match between Map 1 and Map 2; In step 1, feature sets are constructed for the two maps to be aligned. and The process includes: Step 101: Classify the various landmarks in the road scene into three categories according to their distribution rarity, from high to low: the first category, the second category, and the third category. Step 102: Represent each feature in the first category uniformly, and represent each feature in the second category separately according to type; Step 103: Based on the representation method in step 102, construct the feature set using the geographic location coordinates of each feature. and feature set ; The first category of features includes traffic lights and signs; the second category of features includes stop lines, zebra crossings, and road arrow groups. The feature set ; The feature set ; in, , , and They represent the first in Map 1 The first stop line, the first The first zebra crossing, the first The road arrow and the first A traffic light or sign , , and They represent the numbers in map two respectively. The first stop line, the first The zebra crossing, the first The road arrow and the first A traffic light or sign; The third category of features includes: lane lines, curbs, and lampposts, constructing the feature set of the landmarks. and The third type of feature is ignored at that time.

2. The alignment method according to claim 1, characterized in that, Step 1 is followed by: The feature set is calculated. The minimum bounding rectangle of each feature contained therein ; Calculate the rectangle The angle between the longer side and the x-axis of Euclidean space ; Rotate both Map 1 and Map 2 clockwise. Spend.

3. The alignment method according to claim 1, characterized in that, The translation cost function is a one-to-one translation cost function; The landmarks in Map 1 Corresponding to the landmarks in Map 2 The one-to-one translation cost function is: ； in, for coordinates for The coordinates.

4. The alignment method according to claim 3, characterized in that, The translation cost function is a many-to-many translation cost function; Based on the one-to-one translation cost function, the Hungarian algorithm is used to obtain the many-to-many association with the lowest cost, resulting in the many-to-many translation cost function as follows: 。 5. The alignment method according to claim 4, characterized in that, The translation cost function is either a one-to-one translation cost function of coordinate decomposition or a many-to-many translation cost function of coordinate decomposition. The one-to-one translation cost function for x-axis coordinate decomposition is: ； The many-to-many translation cost function for x-axis coordinate decomposition is: ； The one-to-one translation cost function for y-axis coordinate decomposition is: ； The many-to-many translation cost function for y-axis coordinate decomposition is: 。 6. The alignment method according to claim 5, characterized in that, The translation cost function is a translation cost function; the translation cost function includes: the x-axis translation cost function and the y-axis translation cost function; The x-axis translation cost function: ； The y-axis translation cost function: in, Indicates a translation operation The change in coordinate values ​​of all features in Map 2 after being applied to Map 2.

7. The alignment method according to claim 1, characterized in that, The translation cost function is a weighted translation cost function; the calculation process of the weighted translation cost function includes: Map 1 and Map 2 are divided into four layers based on four features, and the x-axis offset cost is calculated for each of the four layers. , , , and y-axis offset cost , , , ; The weighted translation cost function is obtained as follows: ； ； in, , , , These are the translation cost weights of the four features along the x-axis; , , , These are the translation cost weights of the four features along the y-axis.

8. The alignment method according to claim 1, characterized in that, Step 3 includes: Step 301, based on The optimal x-axis translation dx1 between [-40, 40] is searched with a step size of 0.1 meters; the optimal x-axis translation dx2 between [-4, 4] near dx1 is searched with a step size of 0.01 meters. Step 302, based on The optimal y-axis translation dy1 between [-40, 40] is searched with a step size of 0.1 meters; the optimal y-axis translation dy2 between [-4, 4] near dy1 is searched with a step size of 0.01 meters. Step 303: Repeat steps 301-302 a set number of times to obtain the translation transformation. ; Step 304, perform the translation transformation Applying this to Map 2 yields the best match between Map 1 and Map 2.

Images