Sparse map matching method and system based on structural combination features

Abstract

The present invention discloses a sparse map matching method and system based on constructing combined features. The method sets different object combinations to construct combined features, selects corresponding combined features from two sparse maps to be aligned, transforms the combined features in one of the two sparse maps to the coordinate system of the other sparse map through Euclidean transformation, distinguishes true and false matches of the corresponding combined features after transformation based on distance and positional relationship, obtains true matching combined features, calculates the optimal Euclidean transformation based on the true matching combined features, and realizes registration of the two sparse maps through the optimal Euclidean transformation. The present invention constructs three kinds of combined features with adjustable parameters and uses these combined features to match map feature points. Compared with general features, the combined features are more unique and can effectively combat the periodicity of road scenes. They also reduce the probability of false matching and increase the robustness of the algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of crowdsourcing map production, and in particular to a sparse map matching method and system based on structural combination features. Background Art

[0002] Crowdsourcing mapping offers the advantages of low cost and high data freshness. However, the integration and updating of crowdsourced data relies on the identification, matching, and alignment of the collected map fragments. Map alignment is a complex pattern recognition problem, and its high computational overhead, low accuracy, and low recall have long hindered the large-scale application of crowdsourcing mapping.

[0003] Feature maps using sparse representations require less data and lower bandwidth for data upload, making them widely used in large-scale crowdsourced mapping. However, sparse representation can lead to loss of detail. Furthermore, the involvement of algorithms such as image segmentation and semantic recognition in sparse representations can introduce errors in shape, existence, and type, further complicating map alignment.

[0004] How to find sufficient number of high-confidence and high-specificity feature matching point pairs is the core problem of using feature matching technology to achieve sparse feature map matching and alignment. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned technical deficiencies and propose a sparse map matching method and system based on constructed combined features to solve the problem that when the shape and type of feature maps have large errors and repeated patterns, conventional feature extraction algorithms are difficult to find feature points with sufficient specificity and high confidence.

[0006] To achieve the above technical objectives, the technical solution of the present invention provides a sparse map matching method based on constructed combined features, which includes the following steps:

[0007] Set different object combinations to construct combined features;

[0008] Filter out corresponding combined features from the two sparse maps to be aligned;

[0009] The combined features of one of the two sparse maps to be aligned are transformed into the coordinate system of the other sparse map through Euclidean transformation;

[0010] Distinguish true and false matches of the corresponding combined features after transformation based on the distance and position relationship, and obtain true matching combined features;

[0011] The optimal Euclidean transformation is calculated based on the true matching combination features, and the two sparse maps are aligned through the optimal Euclidean transformation.

[0012] A second aspect of the present invention provides a sparse map matching system based on constructed combined features, which includes the following functional modules:

[0013] A combination feature setting module is used to set the combination features of different object combinations;

[0014] A combination feature screening module is used to screen out corresponding combination features from the two sparse maps to be aligned;

[0015] The combined feature transformation module is used to transform the combined features in one of the two sparse maps to be aligned into the coordinate system of the other sparse map through Euclidean transformation;

[0016] The true and false feature distinguishing module is used to distinguish the true and false matches of the corresponding combined features after transformation based on the distance and position relationship, and obtain the true matching combined features;

[0017] The optimal transformation registration module is used to calculate the optimal Euclidean transformation based on the true matching combination features, and to align two sparse maps through the optimal Euclidean transformation.

[0018] A third aspect of the present invention provides a server comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned sparse map matching method based on constructing combined features when executing the computer program.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned sparse map matching method based on constructed combined features.

[0020] Compared with the prior art, the sparse map matching method and system based on constructing combined features described in the present invention sets different object combinations to construct combined features, selects corresponding combined features from the two sparse maps to be aligned, and transforms the combined features in one of the two sparse maps to the coordinate system of the other sparse map through Euclidean transformation. The true and false matches of the corresponding combined features after the transformation are distinguished based on the distance and position relationship to obtain true matching combined features. The optimal Euclidean transformation is calculated based on the true matching combined features, and the two sparse maps are aligned through the optimal Euclidean transformation. The present invention constructs three types of combined features with adjustable parameters and uses these combined features to match map feature points. Compared with general features, the combined features are more unique and can effectively combat the periodicity of road scenes. They also reduce the probability of false matches and increase the robustness of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1This is a flowchart of a sparse map matching method based on constructed combined features according to an embodiment of the present invention;

[0022] Figure 2 is a schematic diagram of the combined features of a sign and a lane line according to an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of the combined features of intersecting solid lane lines according to an embodiment of the present invention;

[0024] Figure 4 is a schematic diagram of the combined features of a zebra crossing and a solid lane line according to an embodiment of the present invention;

[0025] Figure 5 yes Figure 1 Flow chart of step S3 in step S3;

[0026] Figure 6 yes Figure 1 4. The step-by-step flow chart of step S4;

[0027] Figure 7 yes Figure 1 Flow chart of step S5 in step S5;

[0028] Figure 8 This is a module block diagram of a sparse map matching system based on constructed combined features according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0030] like Figures 1 to 7 As shown, an embodiment of the present invention provides a sparse map matching method based on constructing combined features, which includes the following steps:

[0031] S1. Set different object combinations to construct combined features.

[0032] The combination features include: combination features of signboards and lane lines, combination features of intersecting solid lane lines, and combination features of zebra crossings and solid lane lines.

[0033] Specific as Figure 2 As shown, the descriptor of the combined feature of the sign and lane line is composed of the associated vector and lane line direction vector Indicates that .

[0034] in: The direction is the vector pointing from the geometric center of the sign to the geometric center of the nearest lane line, and the length is the distance between the two; Indicates that the direction is the long axis direction of the lane line (the azimuth is limited to ), a vector whose length is the length of the major axis of the lane line.

[0035] The position of F is represented by the geometric center point of the sign, which is denoted as .

[0036] like Figure 3 As shown, the descriptor of the intersecting solid lane line combination feature is set as the angle between the adjacent solid lane lines. , recorded as The geometric position of this feature is defined as the intersection of the extension lines of the two lane lines, denoted as .

[0037] like Figure 4 As shown, the descriptor of the combined feature of the zebra crossing and the solid lane line is expressed by the distance d between the zebra crossing and the nearest solid lane line and the long axis direction vector of the adjacent lane line. Indicates that The geometric position of the feature is defined as the foot of the zebra crossing’s geometric center perpendicular to the lane’s long axis, denoted as .

[0038] S2. Select corresponding combination features from the two sparse maps to be aligned.

[0039] Specifically, the corresponding combination features are selected from the two sparse maps to be aligned according to the following priorities:

[0040] Combined features of signboards and lane lines > Combined features of intersecting solid lane lines > Combined features of zebra crossings and solid lane lines.

[0041] That is, first filter out the combined features of signboards and lane lines from the two sparse maps to be aligned:

[0042] Scan all the signs in the entire image, find the dashed lane line closest to them, and calculate the associated vector using the geometric center of the sign and the geometric center and direction of the lane line. and lane line direction vector , and record the geometric center point of the sign as the feature point of the combined feature.

[0043] Then, the intersecting solid lane line combination features are filtered out from the two sparse maps to be aligned:

[0044] Scan all lane lines in the entire image, calculate the closest distance and the angle between the center axis, and if the closest distance is less than the solid line proximity threshold , and the angle is greater than the solid line divergence threshold , then it is determined that an intersection feature is found and the angle is recorded , and calculate the intersection of the extended lines of the solid lines’ medial axes as the feature points of the combined feature.

[0045] Finally, the zebra crossing and solid lane line combination features are selected from the two sparse maps to be aligned:

[0046] Scan all zebra crossings in the entire image, find the closest solid lane line, and calculate the distance from the zebra crossing to the lane line , calculate the long axis direction vector of the lane line (Azimuth is limited to A perpendicular line is drawn from the center point of the zebra crossing to the long axis of the solid lane line, and the foot of the perpendicular line is the feature point of the combined feature.

[0047] After selecting the corresponding combination features from the two sparse maps to be aligned, the combination features of the two sparse maps are freely combined and paired, and the similarity between each pair of combination features is calculated. The combination feature pairs with similarity less than the similarity threshold are eliminated to obtain a matching feature queue.

[0048] The similarity function between a pair of signboard and lane line combination features is defined as:

[0049]

[0050] In the above formula, These are all adjustable input parameters pre-configured by the algorithm and need to be adjusted before the algorithm runs based on the quality of data collection; Related to the positioning error of the signboard collection, Related to the azimuth error of lane line acquisition; Used to adjust the contribution ratio of the association vector and lane direction vector in the similarity function; They are the feature descriptors of the combination of two signboards and lane lines; They are The correlation vector of They are The lane line direction vector.

[0051] The similarity function between the combined features of a pair of intersecting solid lane lines is defined as:

[0052]

[0053] In the above formula, It is an algorithm-adjustable input parameter related to the angle measurement error of lane line data acquisition; They are the feature descriptors of the combined features of two intersecting solid lane lines; They are of horn.

[0054] The similarity function between a pair of zebra crossing and solid lane line combination features is defined as:

[0055]

[0056] The above formula, These are all adjustable input parameters pre-configured by the algorithm and need to be adjusted before the algorithm runs based on the quality of data collection; Related to the azimuth error of lane line acquisition; Related to the positioning error of zebra crossings; Used to adjust the contribution ratio of the two terms in the similarity function; They are the feature descriptors of the combination of two zebra crossings and solid lane lines; They are The lane line direction vector; They are The zebra crossing lane line association distance.

[0057] Set appropriate parameters according to the sensor error characteristics The recommended parameters are as follows:

[0058]

[0059] Assume that P features are extracted from the first sparse map and Q features are extracted from the second sparse map.

[0060]

[0061] Need to traverse this Each time, a feature feature_a is taken from the first sparse map, and a feature feature_b is taken from the second sparse map. The similarity between feature feature_a and feature feature_b is calculated. If the similarity is greater than the feature similarity threshold , then the two-tuple (denoted as ) is added to the matching feature queue (recorded as matched_features).

[0062] S3. Transform the combined features of one of the two sparse maps to be aligned into the coordinate system of the other sparse map through Euclidean transformation.

[0063] Among them, such as Figure 5 As shown, step S3 includes the following sub-steps:

[0064] S31, randomly extracting two sets of combined feature pairs from the matching feature queue;

[0065] S32, establishing a linear equation system based on the coordinates of the feature points of the two sets of combined feature pairs, and solving the linear equation system to obtain a Euclidean transformation matrix;

[0066] S33. Transform the combined features in one of the two sparse maps to be aligned into the coordinate system of the other sparse map through a Euclidean transformation matrix.

[0067] Although the feature similarity threshold At this time, there may still be a large number of false matches in the matching pairs, because the road scene has a very obvious periodic pattern.

[0068] In order to remove false matches, it is necessary to utilize the relative position relationship between feature points.

[0069] Use the loop traversal method (if there are too many matched feature point pairs, change to random M times, the size of M depends on the computing power of the execution environment and the probability of false matching, which needs to be determined through experiments), each time extract two sets of combined feature pairs from the matching feature queue matched_feature, assuming that a certain time is and .

[0070] According to their anchor points, the corresponding relationship between the coordinate points of the two sets of combined feature pairs is obtained:

[0071]

[0072] Set up a system of linear equations:

[0073]

[0074] Solving the above system of equations gives the Euclidean transformation matrix:

[0075] .

[0076] Use the Euclidean transformation matrix R to transform the anchor points of all feature points in the first sparse map to the coordinate system of the second sparse map to obtain a new feature set .

[0077] S4. Distinguish true and false matches of the corresponding combined features after transformation based on the distance and position relationship, and obtain true matching combined features.

[0078] Among them, such as Figure 6 As shown, step S4 includes the following sub-steps:

[0079] S41, calculating the Euclidean distance between the feature points of each set of combined feature pairs in the matching feature queue after Euclidean transformation;

[0080] S42, comparing the Euclidean distance between the feature points with the feature position offset threshold, and determining that the combined feature pair whose Euclidean distance is less than the feature position offset threshold is a valid match;

[0081] S43. Perform traversal extraction tests on the matching feature queue, check the test with the most valid matches among these extractions, and use the valid matches obtained in this test as the true match set.

[0082] Specifically, traverse the matched feature queue matched_features and check the Euclidean distance d between the geometric position anchors of each set of combined feature pairs in the queue after transformation:

[0083]

[0084] If d is greater than the feature position offset threshold D, the set of combined feature pairs is considered to be an invalid match (called an outlier). Otherwise, it is a valid match, called an inlier.

[0085] After completing the traversal extraction of matched_feature or M random extraction tests, check the test with the largest number of inliers among these extractions and take the inliers obtained in this test as the true matching set true_matched_feature= .

[0086] S5. Calculate the optimal Euclidean transformation based on the true matching combination features, and align the two sparse maps through the optimal Euclidean transformation.

[0087] Among them, such as Figure 7 As shown, step S5 includes the following sub-steps:

[0088] S51, obtaining a set of feature point correspondences between the two sparse maps according to the true matching set;

[0089] S52, according to the corresponding relationship of the feature points, the optimal Euclidean transformation is calculated based on the Umeyama algorithm;

[0090] S53. Use the optimal Euclidean transformation to transform all coordinate points in one of the two sparse maps into the coordinate system of the other sparse map to achieve registration of the two maps.

[0091] The present invention constructs combined features by setting different object combinations, selects corresponding combined features from the two sparse maps to be aligned, transforms the combined features in one of the two sparse maps to the coordinate system of the other sparse map through Euclidean transformation, distinguishes true and false matches of the corresponding combined features after transformation based on distance and positional relationship, obtains true matching combined features, calculates the optimal Euclidean transformation based on the true matching combined features, and aligns the two sparse maps through the optimal Euclidean transformation. The present invention constructs three types of combined features with adjustable parameters and uses these combined features to match map feature points; compared to general features, the combined features are more unique and can effectively combat the periodicity of road scenes; and reduce the probability of false matches, thereby increasing the robustness of the algorithm.

[0092] Furthermore, the present invention can complete the registration of two frames of 1500-meter-long road sparse feature maps in less than 100 milliseconds, and can also work normally when the location and type of POIs are not accurate enough and their existence is uncertain.

[0093] like Figure 8 As shown, the embodiment of the present invention further discloses a sparse map matching system based on constructed combined features, which includes the following functional modules:

[0094] A combination feature setting module 10 is used to set combination features of different object combinations;

[0095] A combination feature screening module 20 is used to screen corresponding combination features from the two sparse maps to be aligned;

[0096] The combined feature transformation module 30 is used to transform the combined features in one of the two sparse maps to be aligned into the coordinate system of the other sparse map through Euclidean transformation;

[0097] A true and false feature distinguishing module 40 is used to distinguish true and false matches of the corresponding combined features after transformation based on the distance and position relationship, and obtain true matching combined features;

[0098] The optimal transformation registration module 50 is used to calculate the optimal Euclidean transformation based on the true matching combination features, and to register the two sparse maps through the optimal Euclidean transformation.

[0099] The execution method of the sparse map matching system based on constructed combined features in this embodiment is basically the same as the sparse map matching method based on constructed combined features described above, so it will not be described in detail.

[0100] The server in this embodiment is a device that provides computing services, typically a computer with high computing power that is provided to multiple consumers via a network. The server in this embodiment includes memory, a processor, and a system bus. The memory includes executable programs stored thereon. Those skilled in the art will appreciate that the terminal device structure in this embodiment does not limit the terminal device and may include more or fewer components than shown, or combinations of certain components, or different component arrangements.

[0101] The memory can be used to store software programs and modules. The processor executes the various functional applications and data processing functions of the terminal by running the software programs and modules stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store the operating system and at least one application required for a function (such as sound playback and image playback). The data storage area can store data generated based on the use of the terminal (such as audio data and phone book). In addition, the memory can include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state memory device.

[0102] An executable program for a sparse map matching method based on constructing combined features is included in a memory. The executable program can be divided into one or more modules / units, which are stored in the memory and executed by a processor to complete the information acquisition and implementation process. The one or more modules / units can be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the server. For example, the computer program can be divided into a combined feature setting module 10, a combined feature screening module 20, a combined feature transformation module 30, a true and false feature distinction module 40, and an optimal transformation registration module 50.

[0103] The processor is the control center of the server, connecting the various components of the entire terminal device using various interfaces and lines. By running or executing software programs and / or modules stored in memory and accessing data stored in memory, it performs various terminal functions and processes data, thereby providing overall terminal monitoring. Optionally, the processor may include one or more processing units; preferably, the processor may integrate an application processor and a modem processor, wherein the application processor primarily processes the operating system, application programs, etc., while the modem processor primarily handles wireless communications. It is understood that the modem processor may not be integrated into the processor.

[0104] The system bus connects the various functional components within a computer and transmits data, address information, and control information. It can be a PCI bus, ISA bus, or VESA bus. Processor instructions are transmitted to memory via the bus, and memory feeds data back to the processor. The system bus is responsible for the exchange of data and instructions between the processor and memory. Of course, the system bus can also connect to other devices, such as network interfaces and display devices.

[0105] The server should at least include a CPU, a chipset, a memory, a disk system, etc. Other components will not be described in detail here.

[0106] In an embodiment of the present invention, the executable program executed by the processor included in the terminal is specifically: a sparse map matching method based on constructing combined features, which includes the following steps:

[0107] Set different object combinations to construct combined features;

[0108] Filter out corresponding combined features from the two sparse maps to be aligned;

[0109] The combined features of one of the two sparse maps to be aligned are transformed into the coordinate system of the other sparse map through Euclidean transformation;

[0110] Distinguish true and false matches of the corresponding combined features after transformation based on the distance and position relationship, and obtain true matching combined features;

[0111] The optimal Euclidean transformation is calculated based on the true matching combination features, and the two sparse maps are aligned through the optimal Euclidean transformation.

[0112] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0113] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0114] Those skilled in the art will appreciate that the modules, units, and / or method steps of the various embodiments described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0115] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sparse map matching method based on constructed combined features, characterized in that: The steps include: Set different object combinations to construct combined features; Filter out corresponding combined features from the two sparse maps to be aligned; The combined features of one of the two sparse maps to be aligned are transformed into the coordinate system of the other sparse map through Euclidean transformation; Distinguish true and false matches of the corresponding combined features after transformation based on the distance and position relationship, and obtain true matching combined features; The optimal Euclidean transformation is calculated based on the true matching combination features, and the two sparse maps are aligned through the optimal Euclidean transformation; The step of selecting corresponding combination features from the two sparse maps to be aligned includes selecting combination features of signs and lane lines, combination features of intersecting solid lane lines, and combination features of zebra crossings and solid lane lines from the two sparse maps to be aligned, which specifically includes: Scan all the signs in the sparse map and find the dashed lane line closest to them. Calculate the association vector and lane line direction vector using the geometric center of the sign and the geometric center and direction of the lane line. Record the geometric center of the sign as the feature point of the combined feature. Scan all lane lines in the sparse map and calculate the closest distance and the angle between the two lane lines. If the closest distance is less than the solid line proximity threshold and the angle is greater than the solid line divergence threshold, a combined feature of intersecting solid lane lines is detected. The angle between the two lane lines is recorded, and the intersection of the extended center axes of the two lane lines is calculated as the feature point of the combined feature. Scan all zebra crossings in the sparse map, find the closest solid lane line, calculate the distance from the zebra crossing to this lane line, calculate the long axis direction vector of this lane line, draw a perpendicular line from the center point of the zebra crossing to the long axis of the solid lane line, and record the foot of the perpendicular as the feature point of the combined feature.

2. The sparse map matching method based on constructed combined features according to claim 1, characterized in that: After selecting the corresponding combination features from the two sparse maps to be aligned, the combination features of the two sparse maps are freely combined and paired, and the similarity between each pair of combination features is calculated. The combination feature pairs with similarity less than the similarity threshold are eliminated to obtain a matching feature queue.

3. The sparse map matching method based on structural combination features according to claim 2, characterized in that: The step of transforming the combined features of one of the two sparse maps to be aligned into the coordinate system of the other sparse map through Euclidean transformation specifically includes: Randomly extract two sets of combined feature pairs from the matching feature queue; A linear equation system is established based on the coordinates of the feature points of the two sets of combined feature pairs, and the Euclidean transformation matrix is ​​obtained by solving the linear equation system; The combined features of one of the two sparse maps to be aligned are transformed into the coordinate system of the other sparse map through the Euclidean transformation matrix.

4. The sparse map matching method based on constructed combined features according to claim 3 is characterized in that: The step of distinguishing true and false matches of the corresponding combined features after transformation based on the distance and position relationship to obtain true matching combined features specifically includes: Calculate the Euclidean distance between the feature points of each set of combined feature pairs in the matching feature queue after Euclidean transformation; Compare the Euclidean distance between feature points with the feature position offset threshold, and determine that the combined feature pair whose Euclidean distance is less than the feature position offset threshold is a valid match; Traverse the matching feature queue and extract the test, check the test with the most valid matches among these extractions, and take the valid matches obtained in this test as the true match set.

5. The sparse map matching method based on constructed combined features according to claim 4 is characterized in that: The optimal Euclidean transformation is calculated based on the true matching combination features, and the registration of the two sparse maps is achieved through the optimal Euclidean transformation, specifically including: Obtain a set of feature point correspondences between two sparse maps based on the true matching set; According to the corresponding relationship of feature points, the optimal Euclidean transformation is calculated based on the Umeyama algorithm; The optimal Euclidean transformation is used to transform all coordinate points in one of the two sparse maps to the coordinate system of the other sparse map to achieve the registration of the two maps.

6. A sparse map matching system based on constructed combined features, characterized in that: Includes the following functional modules: A combination feature setting module is used to set the combination features of different object combinations; A combination feature screening module is used to screen out corresponding combination features from the two sparse maps to be aligned; The combined feature transformation module is used to transform the combined features in one of the two sparse maps to be aligned into the coordinate system of the other sparse map through Euclidean transformation; The true and false feature distinguishing module is used to distinguish the true and false matches of the corresponding combined features after transformation based on the distance and position relationship, and obtain the true matching combined features; The optimal transformation registration module is used to calculate the optimal Euclidean transformation based on the true matching combination features and to register two sparse maps through the optimal Euclidean transformation; The step of selecting corresponding combination features from the two sparse maps to be aligned includes selecting combination features of signs and lane lines, combination features of intersecting solid lane lines, and combination features of zebra crossings and solid lane lines from the two sparse maps to be aligned, which specifically includes: Scan all the signs in the sparse map and find the dashed lane line closest to them. Calculate the association vector and lane line direction vector using the geometric center of the sign and the geometric center and direction of the lane line. Record the geometric center of the sign as the feature point of the combined feature. Scan all lane lines in the sparse map and calculate the closest distance and the angle between the two lane lines. If the closest distance is less than the solid line proximity threshold and the angle is greater than the solid line divergence threshold, a combined feature of intersecting solid lane lines is detected. The angle between the two lane lines is recorded, and the intersection of the extended center axes of the two lane lines is calculated as the feature point of the combined feature. Scan all zebra crossings in the sparse map, find the closest solid lane line, calculate the distance from the zebra crossing to this lane line, calculate the long axis direction vector of this lane line, draw a perpendicular line from the center point of the zebra crossing to the long axis of the solid lane line, and record the foot of the perpendicular as the feature point of the combined feature.

7. A server comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the sparse map matching method based on constructed combined features according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the sparse map matching method based on constructed combined features according to any one of claims 1 to 5 is implemented.

Images