A method, system, and medium for lane positioning on a public roadway

Abstract

The application relates to a lane positioning method, system and medium on a public road, which comprises the following steps: S1: acquiring current vehicle position information and orientation information based on GNSS, inputting the current vehicle position information into a map component to obtain road information of the current vehicle position, filtering the road information of the current vehicle position according to the current vehicle orientation information, and outputting the filtered road information; S2: extracting middle starting point, middle point and end point information of the road according to the filtered road information, calculating a projection distance and an included angle with a projection line segment, if a preset distance threshold and an angle threshold are met, listing the current checking road as a positioning candidate road and outputting positioning candidate road information, and otherwise discarding. The application uses the confidence of sensor data to weight and calculate various cost factors of lane matching, and can comprehensively and accurately perform lane positioning by using various sensor data for information redundancy.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a lane positioning method, system, and medium on public roads. Background Technology

[0002] This invention relates to lane positioning methods in the field of autonomous driving. Lane-level positioning is a fundamental requirement in autonomous driving. Currently, commonly used vehicle lane-level positioning systems include components such as GPS, LiDAR, smart cameras, and millimeter-wave radar to acquire vehicle location information. Then, multi-sensor fusion algorithms are used to calculate the vehicle's lane-level positioning information. Some autonomous driving solutions use LiDAR to create feature maps of the environment and locate the vehicle within these maps. While LiDAR technology can provide accurate and relatively stable positioning methods, it requires high-end hardware, consumes more computing resources, and has high implementation costs, hindering the large-scale production and promotion of autonomous vehicles.

[0003] When initializing lane positioning for autonomous driving, the following issues arise: GNSS provides the vehicle's absolute position, but its positioning accuracy is low in special road sections (such as tunnels, under viaducts, and densely populated city centers with high-rise building density). Combining GNSS with high-precision maps cannot obtain accurate lane-level positioning information, and direct use of this information may pose risks. The accuracy of camera lane line recognition is affected by ambient lighting conditions, the clarity of the lane lines on the ground, and obstacle occlusion, resulting in unstable output with significant errors. If the matching of camera feature data with high-precision maps to determine the current lane position is incorrect, the lane positioning result may deviate from the correct lane, leading to risks. Summary of the Invention

[0004] In view of the above problems, the present invention provides a lane positioning method, system and medium on public roads. The probability of positioning a lane is calculated by using continuous multi-frame sensor data and motion models. The result is more stable and reliable than the result obtained by a single calculation, and can reduce the occurrence of lane positioning errors during the initialization process.

[0005] To achieve the above and other related objectives, the present invention provides the following technical solution: a lane positioning method on a public road, comprising the following steps:

[0006] S1: Obtain the current vehicle location information and orientation information based on GNSS, input the current vehicle location information into the map component to obtain the road information of the current vehicle location, filter the road information of the current vehicle location according to the current vehicle orientation information, and output the filtered road information;

[0007] S2: Based on the filtered road information, extract the starting point, intermediate point and ending point information of the road, project them onto the three sides of the triangle formed by the GNSS positioning position and the starting point, intermediate point and ending point, calculate the projection distance and the angle with the projection line segment. If the preset distance threshold and angle threshold are met, the currently checked road is listed as a positioning candidate road and the positioning candidate road information is output; otherwise, it is discarded.

[0008] S3: Based on the onboard camera, acquire road image data of the current vehicle position, identify and process it to obtain lane line description equation information, lane line type information and lane line quantity information, and calculate the positioning cost p and corresponding weight w for each lane according to the location candidate road information, lane line type information, lane line description equation information and lane line quantity information.

[0009] S4: Based on the positioning cost p and corresponding weight w of each lane, the function to obtain the lane cost value is: Cost sum =∑(p*w), where Cost sum For lane cost, based on the transformation function F, the lane cost is converted into lane matching confidence C = F(Cost). sum The output lane matching confidence sequence information is C = (C1, C2, C3, ..., Cn).

[0010] S5: Based on the confidence information of lane matching, the lane positioning algorithm is used to output lane positioning information.

[0011] Furthermore, in step S5, the lane positioning algorithm includes:

[0012] S51: Based on the confidence sequence information C = (C1, C2, C3, ..., Cn) of lane matching, the initial sequence matrix B is obtained according to the number of confidence sequences n.

[0013] Right now

[0014] S52: Based on the initialized sequence matrix B, obtain the lane matching prediction probability B0. t ,

[0015] B t =M*B T Where M is an n*n order motion matrix, B T To initialize the transpose of sequence matrix B;

[0016] S53: Calculate the observation update probability G, and compare each item in C with B. p Multiplying the corresponding terms by their subscripts, we get the probability G = (C1 * B) p1 C2*B p2 ,…,Cn*B pn );

[0017] S54: Normalize the data in G to obtain the updated positioning lane probability G = ηG, where η is the update factor;

[0018] S55: Check if there is a value in G with a probability greater than the set threshold ξ. If so, the corresponding lane is the positioning lane; otherwise, return to step S3 to continue execution.

[0019] Furthermore, in step S1, the road information of the current vehicle location is filtered to remove oncoming roads in the vehicle's direction of travel.

[0020] Furthermore, in step S2, the projection distance is the distance from the starting point of the road midpoint to the plane where the triangle is located, and the angle with the projection line segment is the angle between the two line segments.

[0021] Furthermore, in step S3, calculating the positioning cost p and corresponding weight w for each lane includes the following steps:

[0022] S31: Calculate the difference between the camera recognition width in the current lane, the left lane, and the right lane and the width of the corresponding lane in the map. The ratio of the difference to the maximum lane width among the candidate lanes is p1, and the weight is w1.

[0023] S32: Based on the matching difference between the types of lane lines on the left and right sides of the current lane and the types of lane lines in the map, if they are the same, it is recorded as 0, and if they are different, it is recorded as p2, with a weight of w2.

[0024] S33: Based on the left lane line type of the left lane and the right lane line type of the right lane, the difference between the matching difference and the lane line type at the corresponding location in the map is recorded as 0 if they are the same and p3 if they are different, with a weight of w4.

[0025] S34: Based on the difference between the number of lane lines identified by the camera and the number of lane lines contained in the current lane and its left and right lanes, if they are the same, it is recorded as 0; if they are different, it is recorded as p4, and the weight is w4.

[0026] S35: The ratio of the distance from the GNSS positioning signal to the center line of the currently calculated lane to the lane with the largest width among the candidate lanes is p5, and the weight is w5;

[0027] S36: p obtained from steps S31-S35 i and w i The value of obtaining lane location n=5 and the corresponding weights n = 5.

[0028] Furthermore, the transformation function F = β T Xβ, where β is an n-order power matrix, β TLet X be the transpose of β, and let X be the matrix of variables to be transformed.

[0029] To achieve the above and other related objectives, a lane positioning system for public roads includes a computer device, characterized in that the computer device is programmed or configured to perform the steps of any of the lane positioning methods for public roads described above.

[0030] To achieve the above and other related objectives, a computer-readable storage medium is provided, characterized in that the computer-readable storage medium stores a computer program programmed or configured to perform any of the lane positioning methods on public roads described above.

[0031] The present invention has the following positive effects:

[0032] 1. This invention uses GNSS to obtain the current vehicle position and uses map output lane information to filter the oncoming lanes in the vehicle's direction of travel, thereby narrowing the lane positioning range and reducing the consumption of computing resources.

[0033] 2. This invention uses the confidence level of sensor data to perform weighted calculations on various cost factors of lane matching. By using multiple sensor data for information redundancy, lane positioning can be performed more comprehensively and accurately.

[0034] 3. This invention calculates the probability of lane positioning using continuous multi-frame sensor data and motion models, which is more stable and reliable than the result obtained from a single calculation, and can reduce the occurrence of lane positioning errors during the initialization process. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0036] Figure 2 This is a flowchart of the candidate lane extraction process of the present invention;

[0037] Figure 3 This is a schematic diagram of the lane positioning algorithm of the present invention. Detailed Implementation

[0038] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0039] Example 1: As Figure 1 As shown, a lane positioning method on a public road includes the following steps:

[0040] S1: Obtain the current vehicle location information and orientation information based on GNSS, input the current vehicle location information into the map component to obtain the road information of the current vehicle location, filter the road information of the current vehicle location according to the current vehicle orientation information, and output the filtered road information;

[0041] S2: Based on the filtered road information, extract the starting point, intermediate point and ending point information of the road, project them onto the three sides of the triangle formed by the GNSS positioning position and the starting point, intermediate point and ending point, calculate the projection distance and the angle with the projection line segment. If the preset distance threshold and angle threshold are met, the currently checked road is listed as a positioning candidate road and the positioning candidate road information is output; otherwise, it is discarded.

[0042] S3: Based on the onboard camera, acquire road image data of the current vehicle position, identify and process it to obtain lane line description equation information, lane line type information and lane line quantity information, and calculate the positioning cost p and corresponding weight w for each lane according to the location candidate road information, lane line type information, lane line description equation information and lane line quantity information.

[0043] S4: Based on the positioning cost p and corresponding weight w of each lane, the function to obtain the lane cost value is: Cost sum =∑(p*w), where Cost sum For lane cost, based on the transformation function F, the lane cost is converted into lane matching confidence C = F(Cost). sum The output lane matching confidence sequence information is C = (C1, C2, C3, ..., Cn).

[0044] S5: Based on the confidence information of lane matching, the lane positioning algorithm is used to output lane positioning information.

[0045] Among them, such as Figure 3 As shown, in step S5, the lane positioning algorithm includes:

[0046] S51: Based on the confidence sequence information C = (C1, C2, C3, ..., Cn) of lane matching, the initial sequence matrix B is obtained according to the number of confidence sequences n.

[0047] Right now

[0048] S52: Based on the initialized sequence matrix B, obtain the lane matching prediction probability B0. t ,

[0049] B t =M*B T Where M is an n*n order motion matrix, BT To initialize the transpose of sequence matrix B;

[0050] S53: Calculate the observation update probability G, and compare each item in C with B. p Multiplying the corresponding terms by their subscripts, we get the probability G = (C1 * B) p1 C2*B p2 ,…,Cn*B pn );

[0051] S54: Normalize the data in G to obtain the updated positioning lane probability G = ηG, where η is the update factor;

[0052] S55: Check if there is a value in G with a probability greater than the set threshold ξ. If so, the corresponding lane is the positioning lane; otherwise, return to step S3 to continue execution.

[0053] In step S1, the road information of the current vehicle position is filtered to remove oncoming roads in the vehicle's direction of travel. In step S2, the projection distance is the distance from the starting point of the road midpoint to the plane containing the triangle, and the angle with the projection line segment is the angle between the two lines containing the projection line segment.

[0054] like Figure 2 As shown, in step S3, calculating the positioning cost p and corresponding weight w for each lane includes the following steps:

[0055] S31: Calculate the difference between the camera recognition width in the current lane, the left lane, and the right lane and the width of the corresponding lane in the map. The ratio of the difference to the maximum lane width among the candidate lanes is p1, and the weight is w1.

[0056] S32: Based on the matching difference between the types of lane lines on the left and right sides of the current lane and the types of lane lines in the map, if they are the same, it is recorded as 0, and if they are different, it is recorded as p2, with a weight of w2.

[0057] S33: Based on the left lane line type of the left lane and the right lane line type of the right lane, the difference between the matching difference and the lane line type at the corresponding location in the map is recorded as 0 if they are the same and p3 if they are different, with a weight of w4.

[0058] S34: Based on the difference between the number of lane lines identified by the camera and the number of lane lines contained in the current lane and its left and right lanes, if they are the same, it is recorded as 0; if they are different, it is recorded as p4, and the weight is w4.

[0059] S35: The ratio of the distance from the GNSS positioning signal to the center line of the currently calculated lane to the lane with the largest width among the candidate lanes is p5, and the weight is w5;

[0060] S36: p obtained from steps S31-S35 i and wi The value of obtaining lane location n=5 and the corresponding weights n = 5.

[0061] Specifically, the single-lane matching cost function includes: calculating the difference between the camera-recognized width and the width of the corresponding lane in the map for the current lane, left lane, and right lane, and the ratio of this difference to the maximum lane width among the candidate lanes, with a weight of 1; the difference between the type of the left and right lane lines of the current lane and the type of lane lines in the map, denoted as 0 if they are the same, and 0.25 if they are different, with a weight of 4, decreasing to 2 when the camera data confidence is low; the difference between the type of the left lane line of the left lane, the type of the right lane line of the right lane and the type of the corresponding lane line in the map, denoted as 0 if they are the same, and 0.125 if they are different, with a weight of 1, decreasing to 0 when the camera data confidence is low; the difference between the number of lane lines recognized by the camera and the number of lane lines contained in the current lane and its left and right lanes, denoted as 0 if they are the same, and 0.25 if they are different, with a weight of 1; and the ratio of the distance of the GNSS positioning signal from the center line of the currently calculated lane to the maximum width of the candidate lanes, with a weight of 4, decreasing to 2 when the GNSS signal is poor.

[0062] Wherein, the transformation function F = β T Xβ, where β is an n-order power matrix, β T Let X be the transpose of β, and let X be the matrix of variables to be transformed.

[0063] The motion matrix is ​​set as an n*n sparse matrix M, where the diagonal is assumed to be 0.9, indicating the vehicle is moving in the lane; the left and right sides are 0.05, indicating the vehicle is changing lanes; and the remaining positions are all 0. Specifically, the first row is {0.9, 0.1, ..., 0}, and the last row is {0, ..., 0.1, 0.9}. For example, when n = 3, the motion matrix is:

[0064]

[0065] To achieve the above and other related objectives, a lane positioning system for public roads includes a computer device, characterized in that the computer device is programmed or configured to perform the steps of any of the lane positioning methods for public roads described above.

[0066] To achieve the above and other related objectives, a computer-readable storage medium is provided, characterized in that the computer-readable storage medium stores a computer program programmed or configured to perform any of the lane positioning methods on public roads described above.

[0067] Example 2: A lane positioning method, system and medium on a public road in Example 1. The present invention will be further described below.

[0068] like Figure 2 As shown, the process of obtaining candidate road information using the public lane positioning method includes the following steps:

[0069] Step 1. Obtain the current location information from GNSS, input it into the map component, and obtain the surrounding road information from the map component.

[0070] Step 2. Obtain the current orientation information from GNSS and filter out oncoming roads in the vehicle's direction of travel.

[0071] Step 3. Based on the filtered surrounding road information, extract the starting point, intermediate point, and ending point information of the road. Project the GNSS positioning position onto the three sides of the triangle formed by the starting point, intermediate point, and ending point, and calculate the projection distance d and the angle θ with the projection line segment. If certain distance and angle thresholds are met, the currently checked road is listed as a candidate road for positioning; otherwise, it is discarded.

[0072] Step 4. Acquire image information from the camera component, process it to obtain the lane line description equation and lane line type information. The number of lane lines identified is 2-4. Extract lane width information, lane line type information, and lane line quantity information from the candidate roads obtained in Step 3.

[0073] The lane positioning algorithm includes:

[0074] S51: Based on the confidence sequence information C = (C1, C2, C3, ..., Cn) of lane matching, the initial sequence matrix B is obtained according to the number of confidence sequences n.

[0075] Right now

[0076] S52: Based on the initialized sequence matrix B, obtain the lane matching prediction probability B0. t ,

[0077] B t =M*B T Where M is an n*n order motion matrix, B T To initialize the transpose of sequence matrix B;

[0078] S53: Calculate the observation update probability G, and compare each item in C with B. p Multiplying the corresponding terms by their subscripts, we get the probability G = (C1 * B) p1 C2*B p2 ,…,Cn*B pn );

[0079] S54: Normalize the data in G to obtain the updated positioning lane probability G = ηG, where η is the update factor;

[0080] S55: Check if there is a value in G with a probability greater than the set threshold ξ. If so, the corresponding lane is the positioning lane; otherwise, return to step S3 to continue execution.

[0081] The computer-readable storage medium of this application stores a computer program that, when executed, implements a public road positioning method. The computer-readable storage medium may include, but is not limited to, floppy disks, optical disks, CD-ROMs (Read-Only Optical Disk Memory), magneto-optical disks, ROMs (Read-Only Memory), RAMs (Random Access Memory), EPROMs (Erasable Programmable Read-Only Memory), EEPROMs (Electrically Erasable Programmable Read-Only Memory), magnetic cards or optical cards, flash memory, or other types of media / machine-readable media suitable for storing machine-executable instructions. The computer-readable storage medium may be a product not connected to a computer device or a component used with a computer device.

[0082] In summary, this invention uses the confidence level of sensor data to weight and calculate various cost factors for lane matching. By using multiple sensor data for information redundancy, lane positioning can be performed more comprehensively and accurately.

[0083] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method of lane positioning on a public road, characterized by, The method comprises the following steps: S1. Obtain current vehicle position information and orientation information based on GNSS, input the current vehicle position information into a map component to obtain road information of the current vehicle position, filter the road information of the current vehicle position according to the current vehicle orientation information, and output filtered road information; S2. Extract the information of the middle start point, middle point and end point of the road according to the filtered road information, project onto the three edges of the triangle formed by the start point, middle point and end point of the GNSS positioning position, calculate the projection distance and the angle with the projection line segment, if the distance threshold and the angle threshold are satisfied, list the current check road as a positioning candidate road and output the positioning candidate road information, otherwise discard; S3. Obtain road image data of the current vehicle position based on the vehicle-mounted camera, and obtain lane line description equation information, lane line type information and lane line number information after identification and processing; S4. Calculate the positioning cost value p and the corresponding weight w of each lane according to the positioning candidate road information, the lane line type information, the lane line description equation information and the lane line number information; S5. Obtain the function of the lane cost value according to the positioning cost value p and the corresponding weight w of each lane: Costsum = ∑(p*w), wherein Costsum is the lane cost value, the lane cost value is converted into the confidence of lane matching based on the conversion function F: C = F(Costsum), and the confidence sequence information C = (C1, C2, C3,..., Cn) of lane matching is output; S6. Output lane positioning information according to the confidence sequence information of lane matching; the lane positioning algorithm comprises: S51. Based on the confidence sequence information C = (C1, C2, C3,..., Cn) of lane matching, obtain an initialization sequence matrix B according to the number n of the confidence sequence; ； S52. Obtain the lane matching prediction probability B according to the initialization sequence matrix B t , B t =M*B T , where M is an n*n order motion matrix, B T is the transpose matrix of the initialization sequence matrix B; S53. Compute observation update probability G, multiply each term in C by the corresponding term in B p with subscripts, to get probability G = (C1*B p1 ,C2*B p2 ,…,Cn*B pn ); S54. Normalize the data in G to obtain the updated positioning lane probability G = ηG, wherein η is an update factor; S55. Check whether there is a value with a probability greater than a set threshold ξ in G, if there is, the corresponding lane is a positioning lane; otherwise, return to step S3 for continuous execution.

2. The lane positioning method on an open road according to claim 1, wherein: In step S1, the road information of the current vehicle position is filtered to filter out the opposite road in the vehicle driving direction.

3. The lane positioning method on the disclosed road according to claim 1, wherein: In step S4, the calculation of the positioning cost value p and the corresponding weight w of each lane comprises the following steps: S41. Calculate the difference between the camera recognition width on the current lane, the left lane and the right lane and the width of the corresponding lane in the map, then the ratio of the difference value to the maximum lane width in the candidate lane is p1, and the weight is w1; S42. According to the type matching difference of the left lane and the right lane in the current lane and the type of the lane in the map, the same is recorded as 0, and the different is recorded as p2, and the weight is w2; S43. According to the type matching difference of the left lane and the right lane in the current lane and the type of the lane in the map, the same is recorded as 0, and the different is recorded as p3, and the weight is w3; S44. According to the number of lane lines identified by the camera and the number of lane lines contained in the current lane and its left and right lanes, if they are the same, record 0, if they are different, record p4, and the weight is w4; S45. Based on the distance between the GNSS positioning signal and the current calculated lane center line, the ratio of the maximum width of the lane in the candidate lane is p5, and the weight is w5; S46. The p i and w i , to obtain the positioning cost value of the lane , n = 5 and the weight value corresponding thereto , n = 5.

4. The method of lane positioning on a public roadway of claim 1, wherein: The conversion function F = β T Xβ, wherein β is an n-order power matrix, β T is the transpose matrix of β, and X is a variable matrix that needs to be converted.

5. A lane positioning system on a public road, comprising a computer device, characterized in that, The computer device is programmed or configured to perform the steps of the lane positioning method on a disclosed road according to any one of claims 1-4.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program programmed or configured to perform the lane positioning method on a disclosed road according to any one of claims 1-4.

Images