Vehicle travel guide line generation method, device, equipment and storage medium

Abstract

This application relates to a method, apparatus, computer device, computer-readable storage medium, and computer program product for generating vehicle driving guide lines. The method includes: obtaining a road complexity evaluation value for a road at a preset distance along the current driving path; the road complexity evaluation value characterizes the complexity of the road; if the road complexity evaluation value is greater than or equal to a preset evaluation value threshold, obtaining lane line points of a road at a preset distance matching the current driving path from a preset lightweight map, and obtaining a first guide line matching the current driving path based on the lane line points; obtaining a second guide line generated by a road perception system to guide the vehicle along the current driving path, and correcting the second guide line using the first guide line to obtain the vehicle driving guide line for the current driving path at the preset distance. This can improve the accuracy of generating vehicle driving guide lines on complex roads, thereby reducing driving risks.

Description

Technical Field

[0001] This application relates to the field of intelligent driving technology, and in particular to a method, apparatus, computer device, computer-readable storage medium, and computer program product for generating vehicle driving guide lines. Background Technology

[0002] Navigate Of Assist (NOA) is a hallmark of advanced intelligent driving, enabling autonomous overtaking, obstacle avoidance, and ramp entry and exit. Early NOA systems primarily relied on the combined use of high-precision maps and high-precision positioning to achieve driver assistance. However, due to the slow update cycle of high-precision maps and the high resource consumption associated with high-precision maps and positioning, the high-precision map solution has gradually been replaced by a lightweight map solution.

[0003] However, lightweight map solutions cannot provide accurate and stable vehicle guidance lines in a timely manner when facing complex road scenarios during assisted driving. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can accurately and stably generate vehicle driving guide lines in response to the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for generating vehicle driving guide lines, including:

[0006] Obtain the road complexity evaluation value of the road at a preset distance on the current driving path; the road complexity evaluation value is used to characterize the complexity of the road;

[0007] If the road complexity evaluation value is greater than or equal to a preset evaluation value threshold, the lane line points of a preset distance road that matches the current driving path are obtained from the preset lightweight map, and a first guide line that matches the current driving path is obtained based on the lane line points.

[0008] A second guideline generated by a road perception system is obtained to guide the vehicle to travel on the current driving path. The second guideline is corrected using the first guideline to obtain the vehicle driving guideline on the road at a preset distance from the current driving path.

[0009] In one embodiment, obtaining the road complexity evaluation value of the vehicle at a preset distance along the current driving path includes:

[0010] The system obtains the current location information of the vehicle, determines the current location in the preset navigation map based on the current location information, queries the road structure information of the road at a preset distance from the current location in the preset navigation map, and obtains the lane line clarity assessment result in real time from the road perception system.

[0011] Based on the road structure information and the lane line clarity assessment results, the road complexity evaluation value of the road at the preset distance is obtained.

[0012] In one embodiment, the road structure information includes average lane radius, road slope, ramp lane spacing, and lane number variation information; the lane line clarity assessment result includes lane line visibility; the lane line visibility is obtained based on pixels identified as lane line areas in the road image acquired by the road perception system and a preset pixel brightness threshold;

[0013] The step of obtaining the road complexity evaluation value of the road at the preset distance based on the road structure information and the lane line clarity evaluation result includes:

[0014] The curve evaluation value is obtained based on the average radius of the lane, the road slope, and the visibility of the lane lines;

[0015] The ramp evaluation value is obtained based on the ramp lane spacing, as well as the preset upper limit threshold and the preset lower limit threshold.

[0016] Based on the lane number change information and the preset lane number change threshold, a lane number change evaluation value is obtained;

[0017] The road complexity evaluation value of the preset distance road is obtained based on the curve evaluation value, the ramp evaluation value, and the lane number change evaluation value.

[0018] In one embodiment, obtaining the curve evaluation value based on the average lane radius, the road slope, and the lane line visibility includes:

[0019] Based on the average lane radius, and the preset upper radius threshold and the preset lower radius threshold, the lane radius evaluation value is obtained;

[0020] Based on the road slope, and the preset upper slope threshold and the preset lower slope threshold, the road slope evaluation value is obtained;

[0021] Based on the lane line visibility, and the preset upper and lower visibility thresholds, a lane line visibility evaluation value is obtained.

[0022] The curve evaluation value is obtained based on the lane radius evaluation value, the road slope evaluation value, and the lane line visibility evaluation value.

[0023] In one embodiment, the current driving route is obtained based on a preset navigation map;

[0024] The step of obtaining lane line points matching the current driving path from a preset lightweight map, and obtaining a first guide line matching the current driving path based on the lane line points, includes:

[0025] Obtain the set of lane line points corresponding to the current driving path in the preset lightweight map, and obtain the lane line points corresponding to each lane from the set of lane line points. Based on the lane line points of each lane, obtain the lane center point of each lane.

[0026] By using the center points of each lane to perform curve fitting, the first guide line of each lane is obtained;

[0027] The first guide line of the lane corresponding to the current driving path is used as the first guide line for matching the current driving path.

[0028] In one embodiment, obtaining the set of lane route points in the preset lightweight map corresponding to the current driving path includes:

[0029] The coordinates of the starting point, ending point, and target point of the current driving route are obtained from the preset navigation map; the coordinates of the target point are any other point on the current driving route other than the starting point and ending point.

[0030] Based on the coordinates of the starting point and the ending point of the path, a candidate waypoint region is generated in the preset lightweight map;

[0031] Based on the coordinates of the target path point, the target path point with the smallest distance to the target path point is obtained from each path point in the candidate waypoint area of ​​the preset lightweight map; each path point is obtained based on the lane line point of each lane.

[0032] Based on the lane line points of each lane corresponding to each target path point, the set of lane line points in the preset lightweight map corresponding to the current driving path is obtained.

[0033] In one embodiment, acquiring the second guide line generated by the road perception system for guiding the vehicle along the current driving path includes:

[0034] The system obtains candidate guide lines corresponding to multiple lanes in the current driving direction generated by the road perception system, obtains the similarity between each candidate guide line and the first guide line, and uses the candidate guide line with the highest similarity as the second guide line to guide the vehicle to drive on the current driving path.

[0035] In one embodiment, the step of correcting the second guide line using the first guide line to obtain the vehicle driving guide line for the current driving path on a road at a preset distance includes:

[0036] Obtain multiple interpolation points of the first guide line and the second guide line, and determine multiple geometric center points based on the multiple interpolation points;

[0037] By fitting multiple geometric center points, a vehicle driving guide line is obtained for the current driving path on a road at a preset distance.

[0038] In one embodiment, acquiring candidate guide lines corresponding to multiple lanes in the current driving direction generated by the road perception system includes:

[0039] The road perception system is used to collect multiple road images in the current driving direction;

[0040] Based on multiple road images, a multi-lane panoramic image from an overhead perspective is obtained;

[0041] The multi-lane panoramic image is input into a pre-trained lane centerline generation model to obtain the lane centerlines of multiple lanes in the current driving direction. The multiple lane centerlines are then used as candidate guide lines for multiple lanes in the current driving direction.

[0042] In one embodiment, obtaining a multi-lane panoramic image from an overhead view based on a plurality of road images includes:

[0043] Feature points are matched on multiple road images, and the multiple road images are stitched together based on the matched feature points to obtain a multi-lane panoramic image from a perspective view.

[0044] The perspective transformation of the multi-lane panoramic image from the perspective view is performed to obtain a multi-lane panoramic image from an overhead perspective.

[0045] Secondly, this application also provides a vehicle driving guide line generating device, comprising:

[0046] The road complexity evaluation value acquisition module is used to acquire the road complexity evaluation value of the road at a preset distance on the current driving path; the road complexity evaluation value is used to characterize the complexity of the road.

[0047] The first guide line acquisition module is used to acquire lane line points of a preset distance road that matches the current driving path from a preset lightweight map when the road complexity evaluation value is greater than or equal to a preset evaluation value threshold, and to obtain a first guide line that matches the current driving path based on the lane line points.

[0048] The vehicle driving guide line generation module is used to acquire a second guide line generated by the road perception system to guide the vehicle to drive on the current driving path, and to correct the second guide line using the first guide line to obtain the vehicle driving guide line of the current driving path on the road at a preset distance.

[0049] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect.

[0050] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0051] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0052] The aforementioned vehicle driving guide line generation method, apparatus, computer equipment, computer-readable storage medium, and computer program product first obtain a road complexity evaluation value for a road at a preset distance on the current driving path. If the road complexity evaluation value is greater than or equal to a preset evaluation value threshold, lane line points of a road at a preset distance matching the current driving path are obtained from a preset lightweight map. Based on the lane line points, a first guide line matching the current driving path is obtained. A second guide line generated by a road perception system for guiding the vehicle on the current driving path is obtained. The second guide line is corrected using the first guide line to obtain the vehicle driving guide line for the current driving path on the road at a preset distance. In this embodiment, by acquiring a dynamic road complexity evaluation value, when the road is determined to be complex, a first guideline matching the current driving path is obtained through static lightweight map data. A second guideline for guiding the vehicle on the current driving path is obtained through a dynamic road perception system. The first guideline obtained from the static map data is then combined with the second guideline obtained from the dynamic data to correct the vehicle driving guideline, thereby improving the accuracy of generating vehicle driving guidelines on complex roads and reducing driving risks. In addition, by evaluating road complexity to determine whether to generate the first guideline, if the road complexity evaluation value is less than or equal to a preset evaluation value threshold, the intelligent driving system does not need to generate the first guideline, which can save data processing and transmission resources and reduce the operating resource consumption of the intelligent driving system. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart illustrating a method for generating vehicle driving guide lines in one embodiment;

[0055] Figure 2 This is a schematic diagram of the process for obtaining the road complexity evaluation value of a vehicle at a preset distance from the current driving path in one embodiment.

[0056] Figure 3 This is a structural block diagram of the intelligent driving system to which the vehicle driving guide line generation method is applied in one embodiment;

[0057] Figure 4 Here is a flowchart of a method for generating vehicle driving guide lines in another embodiment;

[0058] Figure 5 This is a schematic diagram showing the distribution of vehicle cameras on a vehicle in one embodiment;

[0059] Figure 6 This is a schematic diagram illustrating the positional relationship between the lane line point set and the lane line center point in one embodiment;

[0060] Figure 7 This is a schematic diagram illustrating the effect of lane line fitting in a panoramic image from a perspective view in one embodiment;

[0061] Figure 8 This is a schematic diagram of visual guide lines and map guide lines projected onto a top-down view in one embodiment;

[0062] Figure 9 This is a structural block diagram of a vehicle driving guide line generation device in one embodiment;

[0063] Figure 10 This is an internal structural diagram of the vehicle-side control device in one embodiment. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0065] In existing technologies, intelligent driving systems based on lightweight map architecture have insufficient performance in complex road scenarios. For example, in scenarios such as on-ramp, off-ramp, and continuous curves, there are phenomena such as unsmooth lane changing actions, unreasonable lane changing timing, and non-centering of curves.

[0066] After research and analysis, the applicant found that the reason for the above phenomenon is that the lightweight map has deleted the center line of the lane in the high-precision map. The center line generation method based on vehicle sensors is limited by the perception stability of the sensor and the convergence speed of the perception system when facing high speed, continuous ramps with many road branches, continuous curves or other complex road scenarios, resulting in the inability to provide vehicle driving guidance lines in a timely and accurate manner.

[0067] Therefore, in order to improve the performance of intelligent driving systems under lightweight map architecture in complex road scenarios and enable them to achieve the same performance as those under high-precision maps in complex road scenarios, it is urgent to provide a vehicle driving guide line generation method that can improve the performance of intelligent driving systems under lightweight map architecture.

[0068] In one exemplary embodiment, such as Figure 1As shown, a method for generating vehicle driving guide lines is provided. This method is applied to an intelligent driving system, which is set in a vehicle-side control device. The vehicle-side control device can be an intelligent driving domain controller, which plays a central control and coordination role in autonomous driving and assisted driving scenarios. The method for generating vehicle driving guide lines in an intelligent driving system provided in this embodiment includes the following steps S102 to S106, wherein:

[0069] Step S102: Obtain the road complexity evaluation value of the road at a preset distance on the current driving path; wherein, the road complexity evaluation value is used to characterize the complexity of the road.

[0070] The current driving route can refer to the driving route at the current driving time within the global route generated based on the destination selected by the user on the navigation map.

[0071] The preset distance road can refer to roads within a preset distance, such as 2 kilometers. Therefore, the preset distance road for a vehicle on its current driving path can be the road within 2 kilometers from its current location in the direction of the path. Then, the road complexity evaluation value for this 2-kilometer range can be obtained. It should be noted that the road complexity evaluation value characterizes the complexity of the road and can also be used to assess the difficulty of the driving environment, thereby determining whether additional information is needed to further generate vehicle driving guidance lines. Vehicle driving guidance lines are specific path lines used to guide vehicles on the road. Specifically, the road complexity evaluation value can be obtained based on vehicle perception systems (such as onboard sensors), road environment information in a preset map, or traffic information from a cloud server.

[0072] Step S104: If the road complexity evaluation value is greater than or equal to the preset evaluation value threshold, obtain the lane line points of the preset distance road that matches the current driving path from the preset lightweight map, and obtain the first guide line that matches the current driving path based on the lane line points.

[0073] Among them, the preset evaluation value threshold can be a pre-set threshold used to determine whether the complexity of the road has reached a level that requires additional measures.

[0074] In this context, a preset lightweight map can refer to a lightweight version of a high-precision map. This can be achieved by extensively cropping elements from a high-precision map, thereby reducing the overall data volume while retaining relatively important map elements. For example, a preset lightweight map can contain key information required for autonomous driving, such as lane geometry and traffic sign locations.

[0075] Lane line points can refer to a series of points on a map that represent lane lines, and these points can be used to construct the geometry of the lanes.

[0076] Specifically, when the road complexity evaluation value exceeds or equals a preset evaluation value threshold, the lanes of roads at a preset distance that match the current driving path are first determined from a preset lightweight map. After determining the lanes, the lane line points of the determined lanes are extracted. Then, the corresponding lane center point is determined based on the lane line points. Next, the lane centerline of the lane is obtained based on the lane center point, and the lane centerline is used as the first guide line matching the current driving path. The first guide line reflects the ideal driving path based on static map data and can provide a benchmark reference guide line when the road complexity is high.

[0077] Step S106: Obtain the second guide line generated by the road perception system to guide the vehicle to travel on the current driving path, and use the first guide line to correct the second guide line to obtain the vehicle driving guide line on the road at a preset distance from the current driving path.

[0078] Among them, the road perception system can refer to a system that integrates multiple sensors and algorithms, which can be used to monitor and understand the environment around the vehicle in real time.

[0079] Specifically, in this embodiment, the road perception system may include a camera and a road modeling system. The camera can acquire the current environment image, and then, based on the current environment image, it can identify elements such as road edges and lane lines of each lane in the current driving direction, including the current driving path. Based on the identified road edges and lane lines of each lane, the road modeling system outputs the lane center lines of each lane in the current environment image, and uses these lane center lines as second guide lines to guide the vehicle to drive on the current driving path. Because the road complexity evaluation value is greater than or equal to a preset evaluation value threshold, it is necessary to use the first guide line obtained above to correct the second guide line, thereby obtaining the vehicle driving guide line of the current driving path on the road at a preset distance. Then, the intelligent driving system can perform vehicle driving control according to the vehicle driving guide line.

[0080] In the above-mentioned method for generating vehicle driving guide lines, the road complexity evaluation value of the vehicle at a preset distance on the current driving path is first obtained. If the road complexity evaluation value is greater than or equal to a preset evaluation value threshold, lane line points of the preset distance road that match the current driving path are obtained from a preset lightweight map. Based on the lane line points, a first guide line matching the current driving path is obtained. A second guide line generated by the road perception system to guide the vehicle to drive on the current driving path is obtained. The second guide line is corrected using the first guide line to obtain the vehicle driving guide line of the current driving path on the preset distance road. In this embodiment, by acquiring a dynamic road complexity evaluation value, when the road is determined to be complex, a first guideline matching the current driving path is obtained through static lightweight map data. A second guideline for guiding the vehicle on the current driving path is obtained through a dynamic road perception system. The first guideline obtained from the static map data is then combined with the second guideline obtained from the dynamic data to correct the vehicle driving guideline. This can improve the accuracy of generating vehicle driving guidelines on complex roads, thereby reducing driving risks. In addition, by evaluating road complexity to determine whether to generate the first guideline, if the road complexity evaluation value is less than a preset evaluation value threshold, the intelligent driving system does not need to generate the first guideline, which can save data processing and transmission resources and reduce the operating resource consumption of the intelligent driving system.

[0081] In one exemplary embodiment, such as Figure 2 As shown, step S102 includes steps S202 to S204, wherein:

[0082] Step S202: Obtain the vehicle's current location information, determine the current location in the preset navigation map based on the current location information, query the road structure information of the road at a preset distance from the current location in the preset navigation map, and obtain the lane line clarity assessment result in real time from the road perception system.

[0083] The current location information can refer to the exact geographical coordinates obtained from a pre-configured positioning tool on the vehicle. The current location position can refer to the vehicle's current position on a preset navigation map, determined based on the coordinates in the current location information. The preset navigation map can be used to allow users to select a navigation destination and generate a global path based on the map; it can also be used to display the vehicle's position on the navigation map and nearby traffic events. Road structure information can include lane geometry features (such as curve curvature, slope, etc.), number of lanes, road type, traffic signs, etc. Lane line clarity assessment results can refer to the evaluation results obtained by capturing lane images through a road perception system and then using image processing technology to quantitatively assess the visibility and clarity of lane lines; this can be used to reflect the recognizability of lane lines.

[0084] Step S204: Based on the road structure information and lane line clarity assessment results, obtain the road complexity evaluation value of the road at the preset distance.

[0085] Specifically, in this embodiment, the vehicle's current location information is first obtained through a pre-configured positioning tool. Then, based on the location coordinates in the current location information, the current location with the same coordinates is determined in a preset navigation map. Next, road result information for a preset distance road segment starting from the current location is queried from the preset navigation map, and lane line clarity assessment results within the perception range are obtained in real time from the road perception system. Finally, based on the road structure information and lane line clarity assessment results, a preset processing method or algorithm is used to perform comprehensive calculations to obtain the road complexity evaluation value for the road at the preset distance.

[0086] In this embodiment, by comprehensively considering road structure information and lane line clarity, the accuracy of road complexity assessment can be improved, thereby improving the accuracy of subsequent measures taken based on road complexity evaluation values.

[0087] In an exemplary embodiment, the road structure information includes the average lane radius, road slope, ramp lane spacing, and lane number variation information; the lane line clarity assessment result includes lane line visibility; the lane line visibility is obtained based on the pixels identified as lane line areas in the road image acquired by the road perception system and a preset pixel brightness threshold; the above step S204 may specifically include:

[0088] Step S2041: Obtain the curve evaluation value based on the average radius of the lane, the road slope, and the visibility of the lane lines;

[0089] Step S2042: Obtain the ramp evaluation value based on the ramp lane interval, the preset upper limit threshold, and the preset lower limit threshold.

[0090] Step S2043: Based on the lane number change information and the preset lane number change threshold, obtain the lane number change evaluation value;

[0091] Step S2044: Based on the curve evaluation value, ramp evaluation value, and lane number change evaluation value, obtain the road complexity evaluation value of the preset distance road.

[0092] The road structure information may include the average radius of lanes, road slope, ramp lane spacing, and lane number variation information. The lane line clarity assessment result includes lane line visibility. Lane line visibility is obtained based on the pixels identified as lane line areas in the road image acquired by the road perception system and a preset pixel brightness threshold. Specifically, after the road perception system acquires the road image, it identifies the image area corresponding to the lane line in the road image, and then compares the pixels in the image area corresponding to the lane line one by one. Pixels with a brightness greater than the preset pixel brightness threshold in the area are identified as visible pixels. Then, the lane line visibility is obtained based on the ratio of visible pixels to all pixels in the lane line image area.

[0093] In this embodiment, the road complexity described by the road complexity evaluation value is assessed from the perspective of whether the system can quickly predict the guide line based on vision. On the one hand, predicting the future guide line based on the image seen by the camera (i.e., the second guide line generated by the road perception system to guide the vehicle on the current driving path) is limited by the distance and clarity of the visual view. Therefore, when driving on curves or ramps, weather, lighting, traffic flow, road curvature, road undulations, etc., should be considered as factors. On the other hand, when the road topology is complex and the system has a large computational load, and cannot solve the optimal solution in a short time, the visual guide line will be uneven, jumpy, etc. Therefore, whether it merges and exits continuously, intersects, or forks also needs to be considered as factors. Specifically, the above steps S2041-S2044 can be obtained as follows:

[0094] 1) For curves, the main considerations are the reduced visibility of lane lines caused by lighting and weather, and the limited visual visibility and clarity of curves due to small lane radius or excessive slope. Therefore, the factors to be considered in curve scenarios include lane line visibility, average lane radius, and road slope. In this embodiment, the curve evaluation value is obtained based on lane line visibility, average lane radius, and road slope. For example, the curve evaluation value can be obtained from the lane line visibility evaluation value T1, the average lane radius evaluation value T2, and the road slope evaluation value T3.

[0095] 2) For ramps, the main consideration is that the visual guide lines may not have enough time to converge and stabilize due to multiple lane separation scenarios over short distances. Therefore, for ramp scenarios, if lane separation occurs, the interval S (ramp lane separation interval) is determined. In this embodiment, the ramp evaluation value is obtained based on the ramp lane separation interval, as well as the preset upper and lower interval thresholds, which can be obtained by the following formula:

[0096] T

[0097] Wherein, T4 is the ramp evaluation value; S is the interval where lanes diverge. When the lane divergence is less than or equal to 100m, 100 points are awarded; when the lane divergence is greater than 200m, 0 points are awarded; when the interval is between 100 and 200m, the score is -S+200. The preset upper limit threshold for the interval can be 200m, and the preset lower limit threshold for the interval can be 100m. It should be noted that the preset upper limit threshold and the preset lower limit threshold for the interval can also be other values, which can be set or adjusted according to actual needs.

[0098] 3) The increase or decrease of lanes on expressways or some roads will cause the visual guide lines to jump; In this embodiment, the lane number change evaluation value is obtained based on the lane number change information and the preset number change threshold. The specific steps can be: set the lane number change threshold to 1, that is, the occurrence of lane increase or lane decrease will directly accumulate 100 points, and then obtain the lane number change evaluation value T5.

[0099] In summary, based on the curve evaluation value, ramp evaluation value, and lane number change evaluation value, the road complexity evaluation value of the road at the preset distance is obtained, which is the road complexity evaluation value T = T1 + T2 + T3 + T4 + T5.

[0100] In this embodiment, by considering lane visibility, average lane radius, and road gradient, the system can better predict potential risks in curves. The ramp evaluation value corresponding to ramp lane spacing allows for proactive driving responses to complex traffic conditions, and the lane number change evaluation value enables timely adjustments to vehicle guidance lines to prevent driving discontinuities caused by lane changes. By comprehensively considering multiple factors such as average lane radius, road gradient, lane visibility, ramp lane spacing, and lane number change information, the above embodiment provides a comprehensive road complexity scoring system. This multi-dimensional evaluation method can more accurately reflect the impact of road conditions on autonomous driving, thereby improving the adaptability and accuracy of intelligent driving systems on complex road sections.

[0101] In an exemplary embodiment, step S2041 may specifically include:

[0102] Step S21: Obtain the lane radius evaluation value based on the average lane radius, the preset upper radius threshold, and the preset lower radius threshold.

[0103] Step S22: Obtain the road slope evaluation value based on the road slope, as well as the preset upper slope threshold and the preset lower slope threshold.

[0104] Step S23: Based on the lane line visibility, and the preset upper limit threshold and the preset lower limit threshold, obtain the lane line visibility evaluation value;

[0105] Step S24: Obtain the curve evaluation value based on the lane radius evaluation value, road slope evaluation value, and lane line visibility evaluation value.

[0106] In this process, after acquiring a road image, the road perception system identifies the image region corresponding to the lane lines in the road image. Then, it compares the pixels in the image region corresponding to the lane lines one by one, and identifies the pixels in the region whose brightness is greater than a preset pixel brightness threshold as visible pixels. Then, based on the ratio of visible pixels to all pixels in the lane line image region, it uses the visibility of the lane lines as the visibility.

[0107] Specifically, for curves, the main considerations are the reduced visibility of lane lines due to lighting and weather conditions, and the limited visual distance and clarity of curves caused by excessively small lane radii or excessively large slopes. Therefore, the above steps S21-S24 can be obtained as follows:

[0108] 1) For lane line visibility in curved scenarios, current light intensity, road snow and water accumulation, and wear and tear on the lane lines themselves can all reduce lane line visibility. Specifically, based on the average lane radius, road slope, and lane line visibility, a curve evaluation value is obtained. The evaluation value corresponding to lane line visibility can be obtained from the following formula:

[0109] T

[0110] Where T1 is the lane line visibility evaluation value, and V represents lane line visibility. A score of 100 is awarded when lane line visibility is less than or equal to 40%; a score of 0 is awarded when lane line visibility is greater than or equal to 90%; and a score of -200V+180 is awarded when lane line visibility is between 40% and 90%. The preset upper visibility threshold can be 90%, and the preset lower visibility threshold can be 40%. It should be noted that the preset upper and lower visibility thresholds can also be other values, which can be set or adjusted according to actual needs.

[0111] 2) The lane radius evaluation value in a curve scenario can be obtained based on the average lane radius, as well as the preset upper and lower radius thresholds, specifically using the following formula:

[0112] T

[0113] Where T2 is the lane radius evaluation value, and R is the average lane radius. A score of 100 is awarded when the average lane radius is less than or equal to 30m, 0 points are awarded when the average lane radius is greater than or equal to 80m, and -2R+160 points are awarded when the average lane radius is between 30 and 80m. The preset upper radius threshold can be 80m, and the preset lower radius threshold can be 30m. It should be noted that the preset upper and lower radius thresholds can also be other values, which can be set or adjusted according to actual needs.

[0114] 3) The road slope evaluation value in the curve scenario can be obtained based on the road slope, as well as the preset upper and lower slope thresholds. Specifically, it can be obtained by the following formula:

[0115] T

[0116] Where T3 is the road slope evaluation value, and G is the road slope; when the road slope is less than or equal to 3°, 0 points are awarded; when the road slope is greater than or equal to 8°, 100 points are awarded; when the road slope is between 3° and 8°, the score is 20G-60. The preset upper slope threshold can be 8°, and the preset lower slope threshold can be 3°. It should be noted that the preset upper and lower slope thresholds can also be other values, which can be set or adjusted according to actual needs.

[0117] In summary, the curve evaluation value is obtained based on the lane radius evaluation value, road slope evaluation value, and lane line visibility evaluation value. Specifically, the curve evaluation value can be obtained by the sum of the lane line visibility evaluation value T1, the lane radius evaluation value T2, and the road slope evaluation value T3, i.e., T1+T2+T3.

[0118] In this embodiment, the curve evaluation value is obtained by using the average lane radius, road slope, and lane line visibility. This fully considers the impact of reduced lane line visibility caused by lighting and weather, as well as the limited visual visibility distance and clarity of the curve due to excessively small lane radius or excessive slope, thereby improving the accuracy and reliability of the curve evaluation value.

[0119] In an exemplary embodiment, the current driving path is obtained based on a preset navigation map. Step S104, which involves obtaining lane line points matching the current driving path from a preset lightweight map and obtaining a first guide line matching the current driving path based on these lane line points, specifically includes: obtaining a set of lane line points corresponding to the current driving path in the preset lightweight map; obtaining lane line points corresponding to each lane from the set of lane line points; obtaining the lane center point of each lane based on the lane line points; performing curve fitting using the lane center points of each lane to obtain the first guide line of each lane; and using the first guide line of the lane corresponding to the current driving path as the first guide line matching the current driving path.

[0120] The current driving route can refer to the future driving route obtained based on the current vehicle location within the global route in the navigation map. The global route in the navigation map can be a global route generated based on the destination selected by the user in the navigation map. Lane line points can refer to a series of points on the map that represent lane lines, and these points can be used to construct the geometry of the lanes.

[0121] Specifically, the current driving route is obtained based on a preset navigation map. Then, based on the current driving route, a set of lane line points corresponding to the current driving route is obtained from the preset lightweight map. These line points can contain lane line points for multiple lanes. Next, the lane line points corresponding to each lane are obtained from the set of lane line points. For example, the set of lane line points contains lane line points for two lanes, namely line points A, B, and C. Line points A and B surround the first lane, and line points B and C surround the second lane. Then, the center point of the first lane can be obtained from line points A and B, and the center point of the second lane can be obtained from line points B and C. Line point A includes the lane line points along the driving route. The path points A1, A2, A3, etc., in the direction of the driving path, and the path points B also include multiple path points B1, B2, B3, etc., along the driving path. Therefore, the lane center point of the first lane can be obtained through the path points A1 and B1 corresponding to the first lane. Based on the other path points A2, B2, A3, B3, etc., multiple lane center points of the first lane can be obtained. By using multiple lane center points to perform curve fitting, the first guide line of the first lane can be obtained. Similarly, the first guide lines of the other lanes can be obtained. Then, based on the current driving path, the specific lanes included in the path are determined, and the first guide line of the lane corresponding to the current driving path is used as the first guide line matching the current driving path.

[0122] In this embodiment, the set of lane line points is obtained from the preset lightweight map through the current driving path. Then, the lane center point is obtained based on the lane line points of each lane. The lane center point is then used for curve fitting to obtain the first guide line of each lane. The accuracy of the first guide line generation can be improved by using the precise lane information in the preset lightweight map.

[0123] In an exemplary embodiment, the above-mentioned acquisition of the set of lane line points corresponding to the current driving path in the preset lightweight map may specifically include: acquiring the coordinates of the starting point, the ending point, and the target path point of the current driving path at a preset distance from the preset navigation map; the target path point coordinates are the coordinates of any other path point in the current driving path other than the starting point and the ending point; generating a candidate road point area in the preset lightweight map based on the starting point and ending point coordinates; acquiring the target path point with the smallest distance from each path point in the candidate road point area of ​​the preset lightweight map based on the target path point coordinates; each path point is obtained based on the lane line points of each lane; and obtaining the set of lane line points corresponding to the current driving path in the preset lightweight map based on the lane line points of each lane corresponding to each target path point.

[0124] The preset navigation map can be used to allow users to select a navigation destination and generate a global route based on the navigation map. It can also be used to display the vehicle's position on the navigation map and show nearby traffic events. The path start coordinates can be the geographic coordinates of the starting point of the current driving path within a preset distance road range, and the path end coordinates can be the geographic coordinates of the ending point of the current driving path within the preset distance road range. The candidate waypoint area can be a region within the lightweight map that includes relevant lane line points based on the path start and end coordinates. The target path point can be the actual path point found within the candidate waypoint area that has the smallest distance to the given target path point coordinates; the target path point can correspond to the lane line points of each lane.

[0125] For example, the coordinates of the starting point (A, B), ending point (C, D), and target waypoint (X, Y) of the current driving route are obtained from a preset navigation map. Then, based on the starting point (A, B) and ending point (C, D) coordinates, a rectangular or polygonal area can be roughly delineated on the preset lightweight map. This area is the candidate waypoint area, which should cover all possible road segments from the starting point to the ending point. Next, within the candidate waypoint area, based on the target waypoint coordinates (X, Y), the closest actual waypoint is found. This can be done by calculating the Euclidean distance or Manhattan distance between each waypoint and the target waypoint. To implement this, assuming a nearest path point (M, N) is found, it is used as the target path point. The lane line points of each lane where the target path point (M, N) is located are obtained. For example, the lane line points corresponding to the target path point (M, N) in the first lane include lane line point a and lane line point b, the lane line points corresponding to the second lane include lane line point b and lane line point c, and so on for the third lane, etc. Then, all target path points in the current driving path are traversed, and the lane line points of each lane corresponding to all target path points corresponding to the current driving path are obtained to form a set of lane line points. This gives the set of lane line points corresponding to the current driving path in the preset lightweight map.

[0126] In this embodiment, the preset navigation map and the preset lightweight map are mapped by coordinates to obtain a set of lane line points. This can accurately determine the lane line points related to the current driving path, improve the path accuracy of matching paths in different maps, and ensure the accuracy of the guide lines generated based on the set of lane line points.

[0127] In an exemplary embodiment, the step 106 above, which involves obtaining a second guide line generated by the road perception system to guide the vehicle on the current driving path, may specifically include: obtaining candidate guide lines generated by the road perception system corresponding to multiple lanes in the current driving direction; obtaining the similarity between each candidate guide line and the first guide line; and using the candidate guide line with the highest similarity as the second guide line to guide the vehicle on the current driving path.

[0128] Here, candidate guide lines can refer to possible driving paths generated by the road perception system based on multiple lanes in the current driving direction. Similarity can be used to measure the degree of matching between two guide lines, and can be determined by geometric distance, curve fitting error, or other distance correlation analysis algorithms.

[0129] Specifically, the road perception system obtains possible driving paths generated by multiple lanes in the current driving direction, i.e. candidate guide lines for multiple lanes. Then, the similarity between each candidate guide line and the first guide line is obtained through a preset algorithm. The candidate guide line with the highest similarity is used as the second guide line to guide the vehicle to drive on the current driving path.

[0130] In this embodiment, by obtaining the similarity between the candidate guide line and the first guide line, the candidate guide line that is closest to the first guide line can be selected, thereby improving the reliability of using the first guide line to correct the second guide line in subsequent steps.

[0131] In an exemplary embodiment, the step 106 above, which involves correcting the second guide line using the first guide line to obtain the vehicle driving guide line for the current driving path on the road at a preset distance, may specifically include: obtaining multiple interpolation points of the first guide line and the second guide line; determining multiple geometric center points based on the multiple interpolation points; and fitting the multiple geometric center points to obtain the vehicle driving guide line for the current driving path on the road at a preset distance.

[0132] Interpolation points can refer to a series of coordinate points selected from the first and second guide lines, which are used for interpolation calculations between the two guide lines. The geometric center point can be the point obtained by geometrically averaging the interpolation points at the same location on the first and second guide lines.

[0133] Specifically, multiple interpolation points of the first guide line and the second guide line are obtained. Then, the geometric center point of the corresponding interpolation point is determined based on the multiple interpolation points. Finally, all geometric center points are fitted to obtain the vehicle driving guide line of the current driving path on the road at a preset distance.

[0134] In this embodiment, by combining the information from the first guide line and the second guide line, the generated vehicle driving guide line takes into account both static map data and real-time perception data (including data reflecting real road conditions and traffic conditions), thus improving the accuracy of the vehicle driving guide line.

[0135] In an exemplary embodiment, the step of obtaining candidate guide lines corresponding to multiple lanes in the current driving direction generated by the road perception system may include: acquiring multiple road images in the current driving direction using the road perception system; obtaining a multi-lane panoramic image from an overhead view based on the multiple road images; inputting the multi-lane panoramic image into a pre-trained lane centerline generation model to obtain lane centerlines of multiple lanes in the current driving direction; and using the multiple lane centerlines as candidate guide lines for multiple lanes in the current driving direction.

[0136] The road perception system can acquire multiple road images along the current driving direction using multiple onboard cameras; the multi-lane panoramic image refers to converting multiple road images into a single panoramic image viewed from above using image stitching and transformation techniques. The pre-trained lane centerline generation model can be a pre-trained machine learning or deep learning model used to extract lane centerlines from the input multi-lane panoramic image.

[0137] Specifically, a road perception system is used to collect multiple road images in the current driving direction. Then, image stitching and image transformation techniques are used to convert the multiple road images into a multi-lane panoramic image from an overhead view. The multi-lane panoramic image is then input into a pre-trained lane centerline generation model to obtain the lane centerlines of multiple lanes in the multi-lane panoramic image, which are the lane centerlines of multiple lanes in the current driving direction. These multiple lane centerlines are then used as candidate guide lines for multiple lanes in the current driving direction.

[0138] In this embodiment, by acquiring a multi-lane panoramic image from an overhead view, more comprehensive road information can be obtained, thereby improving the accuracy of extracting lane center lines, which in turn improves the accuracy of candidate guide lines.

[0139] In an exemplary embodiment, the process of obtaining a multi-lane panoramic image from an overhead view based on multiple road images in the above embodiment may specifically include: matching feature points of multiple road images; stitching the multiple road images together based on the matched feature points to obtain a multi-lane panoramic image from a perspective view; and performing perspective transformation on the multi-lane panoramic image from the perspective view to obtain a multi-lane panoramic image from an overhead view.

[0140] Feature point matching can be a computer vision technique used to find corresponding points between different images. These points are typically regions in an image with unique properties, such as corners or edges. Image stitching refers to the process of merging two or more partially overlapping images into a wide-view or panoramic image. Perspective transformation is a two-dimensional mapping that simulates how objects in three-dimensional space are projected onto a two-dimensional plane. Perspective transformation can change the viewpoint of an image, allowing it to be viewed from different angles.

[0141] Specifically, feature point matching is performed on multiple road images. Matching feature points are found between adjacent road images. A suitable geometric transformation matrix is ​​calculated using the found matching feature points. The geometric transformation matrix is ​​then applied to stitch the multiple road images together in the correct order and position to form a multi-lane panoramic image from a perspective view. Finally, the perspective transformation matrix is ​​used to convert the multi-lane panoramic image from a perspective view into an overhead view view to obtain a multi-lane panoramic image from an overhead view view.

[0142] In this embodiment, by performing feature point matching, image stitching, and perspective transformation on road images, a series of local road images can be effectively combined into a global panoramic image from an overhead view, thereby improving the comprehensiveness of road information.

[0143] Figure 3 This is a structural block diagram of an intelligent driving system used in the vehicle driving guide line generation method in this application embodiment. The intelligent driving system may include a camera (vehicle-mounted camera), a navigation map SD map (preset navigation map), a lightweight high-precision map HD lite (lightweight map), an intelligent cockpit domain controller CDC (cockpit CDC), an intelligent driving domain controller ADC (intelligent driving ADC), and a PBOX (Positioning Box, high-precision integrated positioning system). Please refer to... Figure 3 The system comprises several components: the vehicle camera, which primarily handles driving perception, identifying road edges, lane lines, and other elements for subsequent perception mapping; the navigation map (SD map), running in the cockpit domain controller (CDC), which is used to select the destination for the Highway Navigation Assist (NOA) and generate a global path based on the navigation map, displaying the vehicle's position on the navigation map and nearby traffic events; the PBOX, which is used to pinpoint the vehicle's precise location and feeds back the location information to the map module to update the vehicle's position; and the lightweight high-precision map (HD lite), running in the intelligent driving domain controller (ADC), which primarily stores the road network of a fixed area and information on other road elements attached to it. The navigation map (SD map) sends the navigation destination and the global path based on the navigation map to the lightweight high-precision map (HD lite). After receiving the data, the lightweight high-precision map (HD lite) performs path matching based on its own data to obtain a global path based on the lightweight high-precision map (HD lite). As the vehicle's position is updated, the navigation map and the lightweight high-precision map output the path and other traffic element information for the next 2 kilometers (preset distance) of the vehicle's vicinity. The intelligent driving ADC's road modeling system, based on the perception results, outputs perception guidance lines (second guidance lines) based on HD... The map guide lines (first guide lines) output by the lite data processing are merged to output the final guide lines (vehicle driving guide lines), which will be used by the subsequent path tracking module for vehicle control.

[0144] Figure 4 This is a flowchart of a method for generating vehicle driving guide lines in an embodiment of this application. Please refer to it. Figures 1-3 Reference Figure 4 This application also provides an optional implementation including steps S1-S9, specifically:

[0145] Step S1, Initialization: After the user gets in the vehicle, if a destination is set based on the navigation map, the system will perform global route planning and provide multiple optional routes for the user to choose from. When the user selects a route and the NOA function is enabled in an area where the conditions for highway NOA are met, the system will generate visual guidance lines and simultaneously score the road complexity.

[0146] Step S2, Road Complexity Scoring: When the system enters the NOA (Noise of Assessment) function, the road complexity scoring module receives information from the navigation map and high-precision positioning. Based on the positioning, it queries the navigation map for road information up to 2 kilometers ahead, obtaining information such as the average curvature, slope, lane merging and diverging of the future road. It also receives lane line clarity assessment results from the perception module for complexity scoring. The "complex road" described in this embodiment is evaluated from the perspective of whether the system can quickly predict guide lines visually. On one hand, predicting future guide lines based on images seen by the camera is limited by the distance and clarity of the visual view. Therefore, when driving on curves or ramps, weather, lighting, traffic flow, road curvature, and road undulations should be considered. On the other hand, when the road topology is complex and the system's computational load is large, making it unable to find the optimal solution quickly, visual guide lines may appear uneven or jerky. Therefore, whether there are continuous merging and merging, intersections, or bifurcations also needs to be considered. To make the complex road judgment programmatic, an integral strategy is designed to integrate multiple factors for judgment, using integration to reflect the complexity of the current road for quickly generating guide lines visually. The specific strategies are shown in the table below:

[0147]

[0148] As described in the table above, this embodiment categorizes complex road types into three types: curves, ramps, and others. Curves are primarily considered due to reduced lane visibility caused by lighting and weather conditions, as well as limitations in visual visibility and clarity caused by excessively small lane radii or steep gradients. Ramps are primarily considered because visual guide lines may not have enough time to converge and stabilize in short-distance multi-lane scenarios. Additionally, the variation in visual guide lines caused by lane increases or decreases on expressways or certain sections of the road is also taken into account.

[0149] For lane line visibility in curved scenes, current light intensity, snow and water accumulation on the road, and wear and tear on the lane lines themselves can all reduce lane line visibility. Lane line visibility is characterized by the ratio of visible lane line pixels to total lane line pixels in the image, and the road complexity integral corresponding to lane line visibility is calculated using the following formula:

[0150] T

[0151] Where T1 is the integral corresponding to lane line visibility, and V represents lane line visibility. When lane line visibility is less than or equal to 40%, 100 points are awarded; when lane line visibility is greater than or equal to 90%, 0 points are awarded; and when lane line visibility is between 40% and 90%, the integral is -200V + 180. The preset upper visibility threshold can be 90%, and the preset lower visibility threshold can be 40%. It should be noted that the preset upper and lower visibility thresholds can also be other values, which can be set or adjusted according to actual needs.

[0152] The average radius integral evaluation strategy for curve scenarios is shown in the following formula:

[0153] T

[0154] Where T2 is the integral corresponding to the average lane radius, and R is the average lane radius. The integral is 100 points when the average lane radius is less than or equal to 30m, 0 points when the average lane radius is greater than or equal to 80m, and -2R+160 when the average lane radius is between 30 and 80m. The preset upper radius threshold can be 80m, and the preset lower radius threshold can be 30m. It should be noted that the preset upper and lower radius thresholds can also be other values, which can be set or adjusted according to actual needs.

[0155] The road gradient integral evaluation strategy for curve scenarios is shown in the following formula:

[0156] T

[0157] Where T3 is the integral corresponding to the road slope, and G is the road slope. When the slope is less than or equal to 3°, the integral is 0 points; when the slope is greater than or equal to 8°, the integral is 100 points; and when the radius is between 3° and 8°, the integral is 20G-60.

[0158] For ramp scenarios, if a road split (lane separation scenario) occurs, the interval S at which the lanes separate is determined, using the following formula:

[0159] T

[0160] Where T4 is the integral corresponding to the lane separation interval of the ramp, and S is the interval at which lane separation occurs. When the lane separation interval is less than or equal to 100m, the integral is 100 points; when the lane separation interval is greater than 200m, the integral is 0 points; when the interval is between 100 and 200m, the integral is -S+200. The preset upper limit threshold can be 200m, and the preset lower limit threshold can be 100m. It should be noted that the preset upper limit threshold and the preset lower limit threshold can also be other values, which can be set or adjusted according to actual needs.

[0161] 100 points will be awarded directly for any increase or decrease in the number of lanes on expressways or certain roads.

[0162] In summary, the integral T = T1 + T2 + T3 + T4 + T5 + T6.

[0163] Step S3: Start map guide line determination; Step S2 above performed a multi-dimensional complexity determination on the road and finally obtained the integral T. If T>=100, map guide line generation is started; if T<100, only visual guide lines are used.

[0164] Step S4: Image stitching; When the system enables NOA, it will use visual guide lines for vehicle control by default, and the generation of visual guide lines first requires image acquisition. The specific camera arrangement is as follows... Figure 5 As shown, the intelligent driving system uses a total of 7 cameras, 4 of which are used for subsequent BEV-view image stitching and road modeling, such as... Figure 5 Cameras numbered 2, 4, 5, and 6.

[0165] The main steps of the image acquisition process are: the optical signal is converted into an electrical signal by a CMOS sensor; the electrical signal is digitized after being processed by an ADC; the digital signal is converted into a high-speed serial data stream by a serializer; it is transmitted over long distances to the domain controller using technologies such as GMSL; and in the domain controller, the deserializer restores the serial data stream into parallel data.

[0166] The main steps of the image stitching process are: denoising, white balancing, and color correction of the original images captured by the camera; distortion correction of images from different cameras based on camera intrinsic parameters; identification and extraction of key feature points such as corners or edges in the images using algorithms; finding matching feature point pairs between adjacent camera images; calculating the geometric transformation model between images based on the feature point pairs to achieve precise image alignment; stitching together images from multiple cameras after alignment to form a continuous panoramic image; and converting the panoramic image from a perspective view to a BEV (overhead view) view by calculating the perspective transformation matrix.

[0167] Step S5: Visual guide lines (second guide lines) are generated. Based on the stitched BEV view, lane lines are output using an end-to-end centerline generation model. The input to this model is the stitched BEV image, and the output is a 5th-order polynomial lane centerline guide line based on the vehicle coordinate system. The general steps of using the end-to-end visual guide line generation model are: collecting driving data based on human drivers; manually or automatically annotating lane centerlines based on the stitched BEV view; training the end-to-end model based on the annotated image; and generating visual guide lines from the stitched image collected during driving using the end-to-end visual guide line generation model.

[0168] Step S6, map matching and waypoint extraction; to achieve low computational cost, the road complexity assessment is based on the SD map. When the path complexity determination result in step three indicates that the map guidance line needs to be activated, the system will perform path matching between the SD map and the HD lite map, and extract the waypoints corresponding to the path in the next 2 kilometers on the HD lite map. The map matching and waypoint extraction steps are as follows: Extract the latitude and longitude coordinates of the starting point SS (latSS, LonSS) and ending point SE (LatSE, LonSE) of the path points within the next 2 kilometers on the SD map; draw a rectangle on the HD lite map using the latitude and longitude of SS and SE as the range of selectable points (candidate waypoint area) for HD lite map matching; extract all road waypoints Si (LatSi, LonSi) (target path points) within the next 2 kilometers on the SD map, where i is the waypoint number; calculate the distance between Si and all road waypoints Hj (LatHj, LonHj) within the rectangle on the HD lite map, expressed as: Di = (LatSi - LatHj) * (LatSi - LatHj) + (LonSi - LonHj) * (LonSi - LonHj); define the Hj with the smallest distance Di from Si as the corresponding point of Si on the HD lite map on the HD lite map; perform the above operation on all Si sequentially to obtain the set H (lane path point set) of all waypoints on the HD lite map corresponding to the path on the SD map.

[0169] Step S7: Map guide lines (first guide lines) are generated. In step S6, the set of waypoints H for roads deemed complex in the HD lite map has been obtained. Based on these waypoints, HD lite element processing is performed to generate map-based guide lines. The specific steps are as follows:

[0170] Lane line point set extraction: Lane lines are described using points in the lightweight map. The point attributes of the lane lines contain predecessor and successor information. Straight lines can be used to connect the predecessor and successor points of each point to obtain the basic road lane envelope. Select the elements Hk of the waypoint set H corresponding to the path in the HD lite map in sequence, {Hk|Hk∈H (k is the subscript)}, and then extract the lane line point set Lk (k is the subscript) belonging to the waypoints Hk.

[0171] Lane centerline point set calculation: For lane line points Lkj with the same index belonging to the same road point Hk, {Lkj|Lkj∈Lk (j is the subscript)}, the corresponding lane center point can be obtained by averaging the x and y coordinates of two lane line points for a certain lane. Figure 6 As shown, Figure 6The edge lane line point set LK on both sides is the lane line point set. The lane center points LC0, LC1, etc. are the calculated lane center line point sets. There are no explicit lane lines, so the dashed lines in the figure represent the lane lines based on point fitting. The related point with index 1 is used as an example to describe the relationship. S1 is a road point on the SD map, H1 is the corresponding road point mapped to the HD lite map, L11 and L12 are two lane line points of a certain lane corresponding to the H1 road point (there are other Lkj for multiple lanes). With L11 coordinates as (L11x, L11y) and L12 coordinates as (L12x, L12y), the corresponding lane center point LC1(x,y) has the following relationship: the horizontal coordinate of the center line point x = (L11x + L12x) / 2; the vertical coordinate of the center line point y = (L11y + L12y) / 2.

[0172] Lane line fitting: Based on the calculated set of centerline points and lane line sets, a fifth-order polynomial is used to fit the lane line points. Grouping the points into sets of six, each set is substituted into the formula y = ax^5 + bx^4 + cx^3 + dx^2 + e^x + f, where a to f are undetermined coefficients. This yields the lane line and centerline equations for the current lateral coordinate x range. The final effect of lane line fitting in a panoramic image from a perspective view is shown below. Figure 7 As shown in the diagram, 701 and 702 are the boundary lane lines, 703 is the fitted non-boundary lane line, 704 is the fitted lane center line, and 705 is the lane center line where the current vehicle is traveling. The fitted lane center line is used as the map guide line (first guide line) and fed into the subsequent guide line fusion module.

[0173] Step S8, Guide Line Fusion: The visual guide line module does not know the vehicle's future global path and planning; it recommends guide lines for various possible lanes based on the vehicle's current lane. Limited by camera recognition distance and FOV (Field of View), visual guide lines are generally shorter than map guide lines, but they are generated based on real-time road conditions, reflecting the current road and traffic situation more realistically. Map guide lines are fitted based on known global path planning solutions, with clear road selections and a longer consideration distance, resulting in more reasonable choices. However, they use non-real-time data from the cloud, which may lead to unreasonable planning (such as temporary road damage or obstacles). To fully utilize the advantages of both guide lines, in complex road sections where map guide lines are enabled, the visual and map guide lines need to be fused and output to the subsequent vehicle planning and control module. A schematic diagram of projecting visual and map guide lines onto the BEV view is shown below. Figure 8 As shown, Figure 8Guide lines 801, 802, and 803 are visual guide lines, while guide line 804 is a map guide line. The arrows in the lane markings only indicate the direction of travel. The following steps will be taken to merge the guide lines:

[0174] Visual guide line selection: First, based on the map guide lines, suitable visual guide lines are selected as future road guide lines. The specific method can be distance correlation analysis, that is, interpolation points are taken for each visual and map guide line at the same horizontal x interval. Then, the distance between the corresponding points of different visual guide lines and map guide lines is calculated. Finally, the distance between each visual guide line and map guide line is summed, and the visual guide line with the smallest total distance to the map guide line is selected as the representative to participate in the subsequent guide line fusion.

[0175] Guide Line Fusion: After selecting suitable visual guide lines for fusion, based on the interpolation points from the previous step, the geometric center points of both the map guide lines and the visual guide lines at the same index are calculated sequentially. The specific method is the same as that used in step seven for solving the lane center points. Finally, based on the calculated sets of visual and map center points, a fifth-order polynomial fitting is performed to obtain the final guide lines. The specific fitting method is the same as that used in step seven for lane line fitting.

[0176] Step S9, End; After the vehicle passes through a complex route section, that is, after the vehicle passes through the current map guidance line range, the entire map guidance line related program will exit, and the system will revert to the default visual guidance line, waiting for the next wake-up.

[0177] In this embodiment, the above technical solution can achieve the following beneficial effects:

[0178] Improving the accuracy of road complexity assessment: This embodiment provides a comprehensive road complexity scoring system by comprehensively considering multiple factors such as lane visibility, road curvature, gradient, and lane spacing. This multi-dimensional assessment method can more accurately reflect the impact of road conditions on autonomous driving systems, thereby improving the adaptability and accuracy of autonomous driving systems on complex road sections.

[0179] Enhancing the safety of autonomous driving systems: The system can activate map guidance lines when it identifies highly complex roads, which helps provide additional safety on complex road sections where visual guidance lines may be inaccurate or unreliable. By combining visual guidance lines and map guidance lines, the system can provide drivers with more reliable navigation information, reducing the risk of accidents.

[0180] Optimized resource utilization and cost-effectiveness: By accurately assessing road complexity and intelligently selecting between visual guide lines and fused guide lines, the system can utilize computing resources more effectively. On road sections where high-precision map guide lines are not required, the system relies solely on visual guide lines, thereby saving resources on data processing and transmission and reducing system operating costs.

[0181] Enhancing system compatibility and scalability: The technical solution in this embodiment can be seamlessly integrated with existing autonomous driving systems without requiring large-scale modifications to existing hardware. Furthermore, the system's scoring and guideline generation strategies can be adjusted according to different autonomous driving levels and vehicle models, exhibiting good compatibility and scalability, which facilitates widespread application.

[0182] Real-time response: The system can quickly assess road complexity during real-time driving and dynamically adjust the guide line generation strategy based on the assessment results to ensure that autonomous vehicles can generate the best guide lines under various road conditions.

[0183] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0184] Based on the same inventive concept, this application also provides a vehicle driving guide line generation device for implementing the vehicle driving guide line generation method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more vehicle driving guide line generation device embodiments provided below can be found in the limitations of the vehicle driving guide line generation method described above, and will not be repeated here.

[0185] In one exemplary embodiment, such as Figure 9 As shown, a vehicle driving guide line generation device 900 is provided, including: a road complexity evaluation value acquisition module 901, a first guide line acquisition module 902, and a vehicle driving guide line generation module 903, wherein:

[0186] The road complexity evaluation value acquisition module 901 is used to acquire the road complexity evaluation value of the vehicle at a preset distance on the current driving path; the road complexity evaluation value is used to characterize the complexity of the road.

[0187] The first guide line acquisition module 902 is used to obtain lane line points of a preset distance road that matches the current driving path from a preset lightweight map when the road complexity evaluation value is greater than or equal to a preset evaluation value threshold, and obtain the first guide line that matches the current driving path based on the lane line points.

[0188] The vehicle driving guide line generation module 903 is used to acquire the second guide line generated by the road perception system to guide the vehicle to drive on the current driving path, and to correct the second guide line using the first guide line to obtain the vehicle driving guide line of the current driving path on the road at a preset distance.

[0189] In an exemplary embodiment, the road complexity evaluation value acquisition module 901 is further configured to acquire the vehicle's current positioning information, determine the current positioning location in a preset navigation map based on the current positioning information, query the road structure information of a road at a preset distance from the current positioning location from the preset navigation map, and acquire the lane line clarity assessment result from the road perception system in real time; and obtain the road complexity evaluation value of the road at the preset distance based on the road structure information and the lane line clarity assessment result.

[0190] In an exemplary embodiment, the road structure information includes the average lane radius, road slope, ramp lane spacing, and lane number change information; the lane line clarity assessment result includes lane line visibility; the lane line visibility is obtained based on the pixels identified as lane line areas in the road image acquired by the road perception system and a preset pixel brightness threshold; the road complexity evaluation value acquisition module 901 is further used to obtain a curve evaluation value based on the average lane radius, road slope, and lane line visibility; obtain a ramp evaluation value based on the ramp lane spacing, and a preset upper and lower interval thresholds; obtain a lane number change evaluation value based on lane number change information and a preset number change threshold; and obtain a road complexity evaluation value for a preset distance road based on the curve evaluation value, ramp evaluation value, and lane number change evaluation value.

[0191] In an exemplary embodiment, the road complexity evaluation value acquisition module 901 is further configured to obtain a lane radius evaluation value based on the average lane radius and a preset upper radius threshold and a preset lower radius threshold; obtain a road slope evaluation value based on the road slope and a preset upper slope threshold and a preset lower slope threshold; obtain a lane line visibility evaluation value based on the lane line visibility and a preset upper visibility threshold and a preset lower visibility threshold; and obtain a curve evaluation value based on the lane radius evaluation value, the road slope evaluation value, and the lane line visibility evaluation value.

[0192] In an exemplary embodiment, the current driving path is obtained based on a preset navigation map; the first guide line acquisition module 902 is further configured to acquire a set of lane line points corresponding to the current driving path in the preset lightweight map, and acquire lane line points corresponding to each lane from the set of lane line points, and obtain the lane center point of each lane based on the lane line points of each lane; perform curve fitting using the lane center points of each lane to obtain the first guide line of each lane; and use the first guide line of the lane corresponding to the current driving path as the first guide line matching the current driving path.

[0193] In an exemplary embodiment, the first guide line acquisition module 902 is further configured to acquire from a preset navigation map the coordinates of the starting point, ending point, and target path point of the current driving path at a preset distance from the road; the target path point coordinates are the coordinates of any other path point in the current driving path other than the starting point and ending point; generate a candidate waypoint region in a preset lightweight map based on the starting point and ending point coordinates; acquire the target path point with the smallest distance from each path point in the candidate waypoint region of the preset lightweight map based on the target path point coordinates; each path point is obtained based on the lane line points of each lane; and obtain the set of lane line points corresponding to the current driving path in the preset lightweight map based on the lane line points of each lane corresponding to each target path point.

[0194] In an exemplary embodiment, the vehicle driving guide line generation module 903 is further configured to obtain candidate guide lines corresponding to multiple lanes in the current driving direction generated by the road perception system, obtain the similarity between each candidate guide line and the first guide line, and use the candidate guide line with the highest similarity as the second guide line for guiding the vehicle to drive on the current driving path.

[0195] In an exemplary embodiment, the vehicle driving guide line generation module 903 is further configured to acquire multiple interpolation points of the first guide line and the second guide line, determine multiple geometric center points based on the multiple interpolation points, and fit the multiple geometric center points to obtain the vehicle driving guide line of the current driving path on the road at a preset distance.

[0196] In an exemplary embodiment, the vehicle driving guide line generation module 903 is further configured to acquire multiple road images in the current driving direction using a road perception system; obtain a multi-lane panoramic image from an overhead view based on the multiple road images; input the multi-lane panoramic image into a pre-trained lane centerline generation model to obtain the lane centerlines of multiple lanes in the current driving direction, and use the multiple lane centerlines as candidate guide lines for multiple lanes in the current driving direction.

[0197] In an exemplary embodiment, the vehicle driving guide line generation module 903 is further configured to perform feature point matching on multiple road images, stitch the multiple road images together based on the matched feature points to obtain a multi-lane panoramic image from a perspective view; and perform perspective transformation on the multi-lane panoramic image from a perspective view to obtain a multi-lane panoramic image from an overhead view.

[0198] Each module in the aforementioned vehicle driving guide line generation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0199] In one exemplary embodiment, a vehicle-side control device is also provided, the internal structure of which can be shown in the following diagram. Figure 10 As shown, it includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0200] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the vehicle-side control device to which the present application is applied. The specific vehicle-side control device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0201] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0202] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0203] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0204] It should be noted that the user information (including but not limited to user device information, user personal information, user navigation information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0205] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0206] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0207] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for generating vehicle driving guide lines, characterized in that, The method includes: Obtain the road complexity evaluation value of the road at a preset distance on the current driving path; the road complexity evaluation value is used to characterize the complexity of the road; If the road complexity evaluation value is greater than or equal to the preset evaluation value threshold, the lane of the road at a preset distance that matches the current driving path is determined from the preset lightweight map, and the lane center point corresponding to the lane line point of the lane is determined. The lane center line obtained based on the lane center point is used as the first guide line that matches the current driving path. The second guide line generated by the road perception system is obtained to guide the vehicle to travel on the current driving path. The second guide line is corrected using the first guide line to obtain the vehicle driving guide line of the current driving path on the road at a preset distance. The second guide line is the lane center line of the lane identified by the road perception system based on the current environmental image.

2. The method according to claim 1, characterized in that, The process of obtaining the road complexity evaluation value of the vehicle at a preset distance along the current driving path includes: The system obtains the current location information of the vehicle, determines the current location in a preset navigation map based on the current location information, queries the road structure information of a road at a preset distance from the current location in the preset navigation map, and obtains lane line clarity assessment results in real time from the road perception system. Based on the road structure information and the lane line clarity assessment results, the road complexity evaluation value of the road at the preset distance is obtained.

3. The method according to claim 2, characterized in that, The road structure information includes average lane radius, road slope, ramp lane spacing, and lane number variation information; the lane line clarity assessment result includes lane line visibility; the lane line visibility is obtained based on the pixels identified as lane line areas in the road image acquired by the road perception system and a preset pixel brightness threshold. The step of obtaining the road complexity evaluation value of the road at the preset distance based on the road structure information and the lane line clarity evaluation result includes: The curve evaluation value is obtained based on the average radius of the lane, the road slope, and the visibility of the lane lines; The ramp evaluation value is obtained based on the ramp lane spacing, as well as the preset upper limit threshold and the preset lower limit threshold. Based on the lane number change information and the preset lane number change threshold, a lane number change evaluation value is obtained; The road complexity evaluation value of the preset distance road is obtained based on the curve evaluation value, the ramp evaluation value, and the lane number change evaluation value.

4. The method according to claim 3, characterized in that, The process of obtaining a curve evaluation value based on the average lane radius, the road slope, and the lane line visibility includes: Based on the average lane radius, and the preset upper radius threshold and the preset lower radius threshold, the lane radius evaluation value is obtained; Based on the road slope, and the preset upper slope threshold and the preset lower slope threshold, the road slope evaluation value is obtained; Based on the lane line visibility, and the preset upper and lower visibility thresholds, a lane line visibility evaluation value is obtained. The curve evaluation value is obtained based on the lane radius evaluation value, the road slope evaluation value, and the lane line visibility evaluation value.

5. The method according to claim 1, characterized in that, The current driving route is obtained based on a preset navigation map; The step of obtaining lane line points matching the current driving path from a preset lightweight map, and obtaining a first guide line matching the current driving path based on the lane line points, includes: Obtain the set of lane line points corresponding to the current driving path in the preset lightweight map, and obtain the lane line points corresponding to each lane from the set of lane line points. Based on the lane line points of each lane, obtain the lane center point of each lane. By using the center points of each lane to perform curve fitting, the first guide line of each lane is obtained; The first guide line of the lane corresponding to the current driving path is used as the first guide line for matching the current driving path.

6. The method according to claim 5, characterized in that, The step of obtaining the set of lane line points in the preset lightweight map corresponding to the current driving path includes: The coordinates of the starting point, ending point, and target point of the current driving route are obtained from the preset navigation map; the coordinates of the target point are any other point on the current driving route other than the starting point and ending point. Based on the coordinates of the starting point and the ending point of the path, a candidate waypoint region is generated in the preset lightweight map; Based on the coordinates of the target path point, the target path point with the smallest distance to the target path point is obtained from each path point in the candidate waypoint area of ​​the preset lightweight map; each path point is obtained based on the lane line point of each lane. Based on the lane line points of each lane corresponding to each target path point, the set of lane line points in the preset lightweight map corresponding to the current driving path is obtained.

7. The method according to claim 1, characterized in that, The acquisition of the second guide line generated by the road perception system for guiding the vehicle to travel on the current driving path includes: The system obtains candidate guide lines corresponding to multiple lanes in the current driving direction generated by the road perception system, obtains the similarity between each candidate guide line and the first guide line, and uses the candidate guide line with the highest similarity as the second guide line to guide the vehicle to drive on the current driving path.

8. The method according to claim 7, characterized in that, The step of correcting the second guide line using the first guide line to obtain the vehicle driving guide line for the current driving path on the road at a preset distance includes: Obtain multiple interpolation points of the first guide line and the second guide line, and determine multiple geometric center points based on the multiple interpolation points; By fitting multiple geometric center points, a vehicle driving guide line is obtained for the current driving path on a road at a preset distance.

9. The method according to claim 7, characterized in that, The acquisition of candidate guide lines corresponding to multiple lanes in the current driving direction generated by the road perception system includes: The road perception system is used to collect multiple road images in the current driving direction; Based on multiple road images, a multi-lane panoramic image from an overhead perspective is obtained; The multi-lane panoramic image is input into a pre-trained lane centerline generation model to obtain the lane centerlines of multiple lanes in the current driving direction. The multiple lane centerlines are then used as candidate guide lines for multiple lanes in the current driving direction.

10. The method according to claim 9, characterized in that, The process of obtaining a multi-lane panoramic image from an overhead view based on multiple road images includes: Feature points are matched on multiple road images, and the multiple road images are stitched together based on the matched feature points to obtain a multi-lane panoramic image from a perspective view. The perspective transformation of the multi-lane panoramic image from the perspective view is performed to obtain a multi-lane panoramic image from an overhead perspective.

11. A vehicle driving guide line generating device, characterized in that, The device includes: The road complexity evaluation value acquisition module is used to acquire the road complexity evaluation value of the road at a preset distance on the current driving path; the road complexity evaluation value is used to characterize the complexity of the road. The first guide line acquisition module is used to determine the lane of the road at a preset distance that matches the current driving path from the preset lightweight map when the road complexity evaluation value is greater than or equal to the preset evaluation value threshold, and to determine the lane center point corresponding to the lane line point of the lane, and to use the lane center line obtained based on the lane center point as the first guide line that matches the current driving path. The vehicle driving guide line generation module is used to acquire a second guide line generated by the road perception system to guide the vehicle to drive on the current driving path, and to correct the second guide line using the first guide line to obtain the vehicle driving guide line of the current driving path on the road at a preset distance; the second guide line is the lane center line of the lane identified by the road perception system based on the current environmental image.

12. A vehicle-end control device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 10.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 10.

Images