Route plan generation device and route plan correction method for route plan generation device
By using a deep learning model in the path planning generation device to identify the white line of the lane and avoid corrections when the white line does not meet the parallel conditions, the problem of path planning accuracy degradation caused by changes in lane relationships is solved, and a more accurate path planning generation is achieved.
Patent Information
- Application Number
- CN202410677721.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-06-01
- Filing Date
- 2024-05-29
- Publication Date
- 2024-12-03
AI Technical Summary
When using the white line identification results to correct the path plan, the path plan may be reduced, especially at branch roads, intersections, etc.
Through the path planning generation device, the white lines of the lane are identified using the first deep learning model and the second deep learning model, and when it is determined that the white lines do not meet the parallel conditions, avoid using the white lines identification results for correction, or use the first deep learning model. The second deep learning model with different learning models is corrected to suppress the reduction in path planning accuracy.
It effectively suppresses the reduction in path planning accuracy caused by correction of white line identification results, ensuring the accuracy of path planning, especially in areas where lane relationship changes are complex.
Smart Images

Figure CN119068431A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a path plan generating device and a path plan correcting method of the path plan generating device. Background Art
[0002] Conventionally, as a technical document related to the path planning of the vehicle, Japanese Patent Publication No. 2018-116370 is known. This publication shows a technology for recognizing the white line of the road of the vehicle based on the image captured by the vehicle-mounted camera.
[0003] Prior art literature
[0004] Patent Literature
[0005] [Patent Document 1] Japanese Patent Application Publication No. 2018-116370 Summary of the invention
[0006] Technical problem to be solved by the invention
[0007] Consider using the results of white line recognition based on the vehicle's camera to generate a vehicle path plan. However, in the area where the relationship between the left and right white lines of the lane changes, the correction using the results of white line recognition may cause the accuracy of the path plan to decrease. The area where the relationship between the left and right white lines of the lane changes refers to branch roads, intersections, areas where the number of lanes increases or decreases, etc.
[0008] Technical solutions to solve problems
[0009] A technical solution of the present invention is a path plan generation device that uses a first deep learning model to generate a path plan for the vehicle based on an image captured by a vehicle-mounted camera. The path plan generation device identifies the left and right white lines that form the driving lane of the vehicle based on the image captured by the vehicle-mounted camera, and uses the recognition result of the white lines to correct the path plan. The path plan generation device determines whether the left and right white lines meet a pre-set parallel condition. If the path plan generation device determines that the left and right white lines do not meet the parallel condition, the path plan generation device does not correct the path plan using the recognition result of the white lines.
[0010] According to a path plan generating device according to one embodiment of the present invention, when the left and right white lines do not satisfy the parallel condition, the path plan using the recognition result of the white lines is not corrected, thereby suppressing the decrease in the accuracy of the path plan caused by the correction.
[0011] In the path plan generation device, the second deep learning model may be used to recognize the left and right white lines based on the image captured by the vehicle-mounted camera. In this case, the second deep learning model may be a deep learning model different from the first deep learning model.
[0012] According to the path plan generation device, the left and right white lines are recognized based on the image captured by the vehicle-mounted camera using a second deep learning model different from the first deep learning model. Thus, even if there is a generation error of the path plan caused by the first deep learning model, the left and right white lines are recognized and corrected using the second deep learning model, thereby suppressing the decrease in the accuracy of the path plan.
[0013] In the path plan generating device, when the driving lane is a straight road, if the angle formed by the left and right white lines is equal to or greater than a permissible angle threshold, it may be determined that the left and right white lines do not satisfy the parallel condition.
[0014] According to the path plan generating device, the angle formed by the left and right white lines becomes larger at the position where the relationship between the left and right white lines changes, and when the angle formed by the left and right white lines is greater than the allowable angle threshold, it is determined that the left and right white lines do not meet the parallel condition. Thus, it is possible to suppress the decrease in the accuracy of the path plan caused by the correction of the recognition result using the white lines.
[0015] In the path plan generation device, the lane width of the driving lane may be calculated based on the recognition result of the white line each time the recognition result of the white line is updated. The path plan generation device may also calculate the difference between the lane width before the update and the lane width after the update, that is, the lane width difference. In this case, the path plan generation device may also determine that the left and right white lines do not meet the parallel condition when the lane width difference is greater than the allowable difference threshold.
[0016] According to the path plan generating device, in the portion where the relationship between the left and right white lines changes, the lane width difference between the lane width before the update and the lane width after the update becomes larger, and when the lane width difference is greater than the allowable difference threshold, it is determined that the left and right white lines do not meet the parallel condition. Thus, it is possible to suppress the decrease in the accuracy of the path plan caused by the correction of the recognition result using the white lines.
[0017] Another technical solution of the present invention is a path plan correction method of a path plan generation device that uses a first deep learning model to generate a path plan for the vehicle based on images captured by a vehicle-mounted camera. In the path plan correction method, the left and right white lines that form the driving lane of the vehicle are identified based on the images captured by the vehicle-mounted camera. In the path plan correction method, it is determined whether the left and right white lines meet a pre-set parallel condition. In the path plan correction method, if it is determined that the left and right white lines do not meet the parallel condition, the path plan using the recognition result of the white lines is not corrected.
[0018] According to another technical solution of the present invention, a path plan correction method of a path plan generating device does not correct the path plan using the recognition result of the white lines when the left and right white lines do not meet the parallel condition. This can suppress the decrease in the accuracy of the path plan caused by the correction.
[0019] Effects of the Invention
[0020] According to the aspects of the present invention, it is possible to suppress a decrease in the accuracy of the path plan due to correction of the path plan using the recognition result of the white line. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 It is a block diagram showing the path plan generating device according to the present embodiment.
[0022] Figure 2 This is a plan view for explaining a situation where the left and right white lines do not satisfy the parallel condition.
[0023] Figure 3 This is a flowchart showing an example of a path plan correction method of the path plan generation device.
[0024] Figure 4 A is a flowchart showing an example of parallel condition determination processing. Figure 4 B is a flowchart showing another example of parallel condition determination processing. DETAILED DESCRIPTION
[0025] Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0026] Figure 1 Detailed Description of the Invention It is a block diagram showing the path plan generating device 100 according to the present embodiment. Figure 1The path plan generation device 100 shown is a device that generates a path plan for the vehicle based on the image captured by the vehicle-mounted camera 1. The path plan generation device 100 can be mounted on the vehicle, or a part of the function can be executed by a server that can communicate with the vehicle. The path plan generation device 100 outputs the generated path plan to an automatic driving device, a driving assistance device, etc. The path plan generation device 100 can also be a part of the automatic driving device or the driving assistance device.
[0027] The path plan is a plan related to the path traveled by the vehicle. The path plan is used, for example, for autonomous driving or driving assistance of the vehicle. The path plan may be continuous coordinate data on a map. The path plan may also be data including a target longitudinal position (as the position of the control target in the extension direction of the driving lane) and a target lateral position (as the position of the control target in the width direction of the driving lane) relative to the driving lane. The path plan may also be data including a target steering angle relative to the target longitudinal position of the driving lane. The path plan may also be time series data of a target lateral position relative to the lane and time series data of a target steering angle. The target steering torque may also be used instead of the target steering angle.
[0028] [Configuration of the Path Plan Generating Device]
[0029] The following describes the structure of the path plan generating device 100 according to the present embodiment. Figure 1 As shown, the path plan generating device 100 has an ECU (Electronic Control Unit) 10 that manages the device in an overall manner. The ECU 10 is an electronic control unit having a CPU (Central Processing Unit) and a storage unit such as a ROM (Read Only Memory) or a RAM (Random Access Memory). In the ECU 10, various functions are realized by, for example, the CPU executing a program stored in the storage unit. The ECU 10 may also be composed of a plurality of electronic units.
[0030] The ECU 10 is connected to a vehicle-mounted camera 1 , a radar sensor 2 , a GNSS receiving unit 3 , and a map database 4 .
[0031] The vehicle-mounted camera 1 is a photographing device for photographing the external conditions of the vehicle. The vehicle-mounted camera 1 is, for example, arranged on the inner side of the front windshield of the vehicle to photograph the front of the vehicle. The vehicle-mounted camera 1 may also be composed of a plurality of cameras for photographing the rear and side of the vehicle. The vehicle-mounted camera 1 sends the photographed image of the exterior of the vehicle to the ECU 10.
[0032] The radar sensor 2 is a detection device that uses radio waves (such as millimeter waves) or light to detect objects around the vehicle. The radar sensor 2 includes, for example, a millimeter wave radar or a laser detection and ranging system (LiDAR: Light Detection And Ranging). The radar sensor 2 detects objects by sending radio waves or light around the vehicle and receiving radio waves or light reflected by the object. The radar sensor 2 sends object detection information related to object detection to the ECU 10. The object may also include the white line of the road.
[0033] The GNSS receiving unit 3 receives signals from positioning satellites to measure the position of the vehicle (for example, the latitude and longitude of the vehicle). The GNSS receiving unit 3 transmits the measured position information of the vehicle to the ECU 10 .
[0034] The map database 4 is a database storing map information. The map database 4 is formed in a storage device such as a HDD (Hard Disk Drive) mounted on the vehicle. The map information includes road location information, road shape information (such as the type of curved and straight roads, the curvature of curved roads, etc.), lane width information, intersection and branch point location information, and structure location information. In addition, the map database 4 can also be formed in a server that can communicate with the vehicle.
[0035] Next, the functional configuration of the ECU 10 will be described. Figure 1 As shown, the ECU 10 includes a path plan generating unit 11, a white line recognizing unit 12, and a path plan correcting unit 13. In addition, part of the functions of the ECU 10 described below may be executed in a server that can communicate with the host vehicle.
[0036] The path plan generating unit 11 generates a path plan for the vehicle based on the image captured by the vehicle-mounted camera 1. In addition to the image captured by the vehicle-mounted camera 1, the path plan generating unit 11 may also use the object detection information of the radar sensor 2 for generating the path plan. The path plan generating unit 11 may also use the position information of the vehicle measured by the GNSS receiving unit 3 and the map information of the map database 4 for generating the path plan.
[0037] The path plan generation unit 11 uses the first deep learning model 11a to generate the path plan. The first deep learning model 11a is, for example, a machine learning model that performs deep learning-based learning to output a path plan based on the image captured by the vehicle-mounted camera 1. The neural network constituting the first deep learning model 11a is, for example, a convolutional neural network (CNN) including multiple layers, the multiple layers including multiple convolutional layers and pooling layers. The neural network can also be configured as a recursive neural network (RNN).
[0038] In addition, the captured image that becomes the input of the first deep learning model 11a may also be pre-processed. The pre-processing includes, for example, normalization of data, supplementation of data defects, removal of noise, etc. The pre-processing may also include processing to extract feature quantities from the captured image. As feature quantities, for example, at least one of the white line of the lane, the lane width, the curvature of the lane, obstacles contained in the captured image, road signs, road surface markings, and the lighting status of the traffic light may be included. The obstacles include at least one of moving objects such as other vehicles, guardrails, walls, utility poles, and stationary objects such as parked vehicles. The path plan generation unit 11 may also be a method of inputting the above-mentioned feature quantities together with the captured image into the first deep learning model 11a.
[0039] The path plan generation unit 11 inputs the captured images of the vehicle-mounted camera 1 to the first deep learning model 11a. Thus, the path plan generation unit 11 obtains the path plan output from the first deep learning model 11a. The path plan generation unit 11 may also input time series data of captured images in a certain period of time. The path plan generation unit 11 may also additionally input object detection information of the radar sensor 2 and map information around the vehicle to the first deep learning model 11a. The map information around the vehicle may also be map information related to the road on which the vehicle is traveling.
[0040] The white line recognition unit 12 recognizes the left and right white lines forming the driving lane of the host vehicle based on the image captured by the vehicle-mounted camera 1. The white line recognition unit 12 performs white line recognition independently of the path plan generation unit 11. Specifically, the white line recognition unit 12 performs white line recognition using a second deep learning model 12a.
[0041] The second deep learning model 12a is a machine learning model that performs deep learning-based learning to output a recognition result of a white line based on an image captured by the vehicle-mounted camera 1. The neural network constituting the second deep learning model 12a may be a convolutional neural network or a recursive neural network. The constituting neural network may be of the same type as the first deep learning model 11a or a different type of neural network from the first deep learning model 11a.
[0042] The white line recognition unit 12 obtains the recognition result of the white line output from the second deep learning model 12a by inputting the captured image of the vehicle-mounted camera 1 to the second deep learning model 12a. The white line recognition unit 12 may also input time series data of the captured image in a certain period of time. The white line recognition unit 12 may also additionally input the object detection information of the radar sensor 2 to the second deep learning model 12a. The white line recognition unit 12 may also be a method of additionally inputting only the information that has passed the filter for extracting features associated with the white line among the object detection information of the radar sensor 2.
[0043] The white line recognition unit 12 does not necessarily need to use the second deep learning model 12a for white line recognition. The white line recognition unit 12 may also recognize the white line of the driving lane by a white line recognition algorithm using the gradient of brightness in the captured image. The white line recognition unit 12 may also recognize the white line of the driving lane by a known image processing method. In this case, the white line recognition unit 12 may not have the second deep learning model 12a. The known image processing methods include at least one of edge detection, filter smoothing, color space conversion, binarization, Hough transform, and pattern matching.
[0044] The path plan correction unit 13 uses the white line recognition result of the white line recognition unit 12 to correct the path plan generated by the path plan generation unit 11. For example, when the distance between the path of the vehicle in the path plan and the center position between the left and right white lines in the lane width direction is greater than a certain distance, the path plan correction unit 13 corrects the path plan so that the path of the vehicle approaches the center position. The correction method of the path plan is not limited to the above content, and a known correction method can be used.
[0045] The path plan correction unit 13 determines whether the left and right white lines of the driving lane identified by the white line recognition unit 12 satisfy the parallel condition. The parallel condition is a condition preset to determine whether the left and right white lines are in a substantially parallel relationship. When the left and right white lines do not satisfy the parallel condition, the path plan correction unit 13 does not correct the path plan using the white line recognition result of the white line recognition unit 12. When the left and right white lines satisfy the parallel condition, the path plan correction unit 13 corrects the path plan using the white line recognition result. The path plan correction unit 13 determines the parallel condition of the white lines, for example, at a certain interval in the extension direction of the driving lane. The front and rear direction of the vehicle may also be used instead of the extension direction of the driving lane.
[0046] Specifically, when the driving lane of the vehicle is a straight road, the path plan correction unit 13 determines that the left and right white lines do not meet the parallel condition when the angle formed by the left and right white lines identified by the white line recognition unit 12 is greater than the allowable angle threshold. The allowable angle threshold is a threshold set to determine the parallel condition of the white lines. The allowable angle threshold can be 5°, 3°, or 1°. The allowable angle threshold can be set to a value larger than the error amount of the angle calculation, which is caused by the recognition accuracy of the white lines in the white line recognition unit 12.
[0047] The path plan correction unit 13 determines whether the driving lane of the host vehicle is a straight road based on the position information of the host vehicle measured by the GNSS receiving unit 3 and the map information of the map database 4. The path plan correction unit 13 can also determine whether the driving lane of the host vehicle is a straight road based on the curvature of the path in the path plan generated by the path plan generation unit 11. The path plan correction unit 13 can also determine whether the driving lane of the host vehicle is a straight road based on the average value of the curvature of the left and right white lines recognized by the white line recognition unit 12.
[0048] The path plan correction unit 13 calculates the angle between the left and right white lines when it is determined that the driving lane of the vehicle is a straight road. Based on the recognition result of the white line by the white line recognition unit 12, the path plan correction unit 13 recognizes the extension direction of the left and right white lines at a certain interval by a known image processing method. The path plan correction unit 13 extends each white line and calculates the angle when the intersection is used as the angle of the left and right white lines. When the angle between the left and right white lines is greater than the allowable angle threshold, the path plan correction unit 13 determines that the left and right white lines do not meet the parallel condition.
[0049] In addition, the path plan correction unit 13 may also use the lane width to determine the parallel condition. Whenever the white line recognition unit 12 updates the recognition result of the white line, the path plan correction unit 13 calculates the lane width of the driving lane based on the recognition result of the white line. The lane width is equivalent to the interval between the left and right white lines in the width direction of the driving lane. In addition, the width direction of the vehicle may be used instead of the width direction of the driving lane.
[0050] The path plan correction unit 13 calculates the lane width difference between the lane width before the update and the lane width after the update. The lane width difference is an absolute value. The lane width difference can be an indicator that increases as the lane width after the update increases relative to the lane width before the update. When the lane width difference is greater than the allowable difference threshold, the path plan correction unit 13 determines that the left and right white lines of the driving lane do not meet the parallel condition. The allowable difference threshold is a threshold set to determine the parallel condition of the white lines. The allowable difference threshold can be set to a value larger than the error amount of the lane width difference, which is caused by the recognition accuracy of the white lines in the white line recognition unit 12. When the left and right white lines do not meet the parallel condition, the path plan correction unit 13 does not correct the path plan using the recognition result of the white lines.
[0051] Here, Figure 2 This is a plan view for explaining a situation where the left and right white lines do not satisfy the parallel condition. Figure 2 The vehicle enters the branch point of the road. Figure 2 The host vehicle M, the path C in the path plan of the host vehicle M, the driving lane L1, the branch lane L2, and the white lines W1 to W3 are shown. The path C in the path plan is a path corresponding to the path plan generated by the path plan generation unit 11. The driving lane L1 is the lane in which the host vehicle M is traveling. The branch lane L2 is a lane that branches from the driving lane L1 and faces a direction different from the driving lane L1.
[0052] Figure 2 The white line W1 shown forms the white line on the left side of the driving lane L1 until the branching. The white line W1 becomes the white line on the left side of the branch lane L2 after the branching. The range where the white line W1 is inclined relative to the extending direction of the driving lane L1 to form the branch lane L2 is represented as the branch range Es. The white line W2 is the white line on the right side of the driving lane L1. The white line W3 is a dotted line that divides the driving lane L1 and the branch lane L2.
[0053] exist Figure 2 In the situation shown, the parallel condition of the white line W1 and the white line W2 is not satisfied within the branch range Es. Therefore, the path plan correction unit 13 does not correct the path plan using the recognition results of the white line W1 and the white line W2 by the white line recognition unit 12. In the branch range Es, the white line W1 extends obliquely relative to the driving lane L1. As a result, the angle formed by the white line W1 and the white line W2 becomes greater than the allowable angle threshold, and therefore the parallel condition is not satisfied. In addition, in the branch range Es, the white line W1 is inclined in the direction away from the white line W2. The lane width difference between the white line W1 and the white line W2 also becomes greater than the allowable difference threshold. Therefore, if the path plan is corrected using the white line recognition result of the branch range Es, it may become an inappropriate correction. An inappropriate correction is represented by an arrow Ca.
[0054] When the parallel condition of the white line W1 and the white line W2 is not satisfied, the path plan correction unit 13 does not correct the path plan using the recognition result of the white line W1 and the white line W2 by the white line recognition unit 12. Thus, it is possible to avoid inappropriate correction of the path plan. When the path plan is corrected or not corrected, the path plan correction unit 13 outputs the path plan to the automatic driving device, the driving assistance device, etc.
[0055] [Path plan correction method of path plan generation device]
[0056] Next, a path plan correction method of the path plan generating device 100 according to the present embodiment will be described with reference to the drawings. Figure 3 2 is a flowchart showing an example of a path plan correction method of the path plan generating device 100. The path plan correction method is executed, for example, when driving assistance control or automatic driving control is performed.
[0057] like Figure 3 As shown, as S1, the ECU 10 of the route plan generating device 100 acquires an image captured by the vehicle-mounted camera 1. The ECU 10 acquires an image of the exterior of the vehicle including the front of the host vehicle M captured by the vehicle-mounted camera 1. Thereafter, the ECU 10 moves to S2.
[0058] As S2, the ECU 10 generates a path plan using the first deep learning model 11a through the path plan generation unit 11. The path plan generation unit 11 generates a path plan by inputting the image captured by the vehicle-mounted camera 1 to the first deep learning model 11a. Then, the ECU 10 moves to S3.
[0059] In S3, the ECU 10 performs white line recognition using the second deep learning model 12a through the white line recognition unit 12. The white line recognition unit 12 recognizes the left and right white lines of the driving lane by inputting the captured image of the vehicle-mounted camera 1 to the second deep learning model 12a. In addition, the white line recognition unit 12 does not necessarily need to use the second deep learning model 12a for white line recognition. After that, the ECU 10 moves to S4.
[0060] In S4, the ECU 10 determines whether the left and right white lines of the driving lane meet the parallel condition through the path plan correction unit 13. The process of determining the parallel condition will be described in detail later. When the ECU 10 determines that the left and right white lines of the driving lane meet the parallel condition (S4: YES), the ECU 10 moves to S5. When the ECU 10 does not determine that the left and right white lines of the driving lane meet the parallel condition (S4: NO), the ECU 10 moves to S6.
[0061] In S5, the ECU 10 uses the recognition result of the white line by the white line recognition unit 12 to correct the path plan generated by the path plan generation unit 11. The correction method is not particularly limited. Thereafter, the ECU 10 proceeds to S6.
[0062] In S6, the ECU 10 outputs the path plan to the automatic driving device, the driving assistance device, etc. Then, the ECU 10 completes the path plan correction method. Figure 3 S2 and S3 can also be executed in the reverse order or simultaneously.
[0063] Figure 4 A is a flowchart showing an example of parallel condition determination processing. Figure 3 This corresponds to S4 of the flowchart.
[0064] like Figure 4 As shown in A of FIG. 1 , as S10, the ECU 10 determines whether the driving lane of the vehicle is a straight road through the path plan correction unit 13. The path plan correction unit 13 determines whether the driving lane of the vehicle is a straight road based on the position information of the vehicle measured by the GNSS receiving unit 3 and the map information of the map database 4. When the ECU 10 determines that the driving lane is a straight road (S10: Yes), it moves to S11. When the ECU 10 does not determine that the driving lane is a straight road (S10: No), it ends the parallel condition determination process.
[0065] In S11, the ECU 10 determines whether the angle formed by the left and right white lines is greater than the permissible angle threshold value through the path plan correction unit 13. When the ECU 10 determines that the angle formed by the left and right white lines is greater than the permissible angle threshold value (S11: Yes), the ECU 10 moves to S12. When the ECU 10 does not determine that the angle formed by the left and right white lines is greater than the permissible angle threshold value (S11: No), the parallel condition determination process ends.
[0066] In S12, the ECU 10 determines that the left and right white lines of the driving lane do not satisfy the parallel condition through the path plan correction unit 13. Thereafter, the ECU 10 ends the parallel condition determination process.
[0067] Figure 4 B is a flowchart showing another example of parallel condition determination processing. Figure 4 As shown in FIG. 1B , as S20 , the ECU 10 calculates the lane width of the driving lane through the path plan correction unit 13 . The lane width of the driving lane may be calculated each time the white line recognition unit 12 updates the recognition result of the white line through the parallel condition determination process. Thereafter, the ECU 10 moves to S21 .
[0068] In S21, the path plan correction unit 13 determines whether the lane width difference between the lane width before the update and the lane width after the update is greater than the allowable difference threshold. When the ECU 10 determines that the lane width difference is greater than the allowable difference threshold (S21: Yes), the ECU 10 moves to S22. When the ECU 10 does not determine that the lane width difference is greater than the allowable difference threshold (S21: No), the parallel condition determination process ends.
[0069] In S22, the ECU 10 determines that the left and right white lines of the driving lane do not satisfy the parallel condition through the path plan correction unit 13. Thereafter, the ECU 10 ends the parallel condition determination process.
[0070] If the ECU 10 does not determine that the parallel condition is not satisfied in any parallel condition determination process, it determines that the left and right white lines of the driving lane satisfy the parallel condition. Figure 4 A and Figure 4 The parallel condition determination process shown in B may be executed by both parties or by only one party. The parallel condition determination process may also be executed by other methods.
[0071] According to the path plan generating device 100 involved in the present embodiment described above, when the left and right white lines do not meet the parallel condition, the path plan using the recognition result of the white lines is not corrected. Thus, the accuracy of the path plan caused by the correction can be suppressed. This is because the accuracy of the path plan may be reduced by correcting the path plan using the recognition result of the white lines at locations where the relationship between the left and right white lines changes, such as branch roads, intersections, and locations where the number of lanes increases or decreases.
[0072] In addition, in the path plan generation device 100, the left and right white lines are recognized based on the image captured by the vehicle-mounted camera using a second deep learning model different from the first deep learning model. Thus, even if there is a generation error of the path plan due to the first deep learning model, the accuracy of the path plan can be suppressed by correcting it using the left and right white lines recognized by the second deep learning model.
[0073] In addition, in the path plan generating device 100, the angle formed by the left and right white lines becomes larger at the position where the relationship between the left and right white lines changes, and when the angle formed by the left and right white lines is greater than the allowable angle threshold, it is determined that the left and right white lines do not meet the parallel condition. Thus, it is possible to suppress the decrease in the accuracy of the path plan caused by the correction of the recognition result using the white lines.
[0074] Alternatively, in the path plan generating device 100, since the lane width difference between the lane width before the update and the lane width after the update becomes larger at the position where the relationship between the left and right white lines changes, when the lane width difference is greater than the allowable difference threshold, it is determined that the left and right white lines do not satisfy the parallel condition. Thus, it is possible to suppress the decrease in the accuracy of the path plan caused by the correction using the recognition result of the white lines.
[0075] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments. The present invention can be implemented in various forms by variously changing and improving the above-described embodiments as representatives based on the knowledge of those skilled in the art.
[0076] Furthermore, the path plan generating device 100 does not necessarily need to use the radar sensor 2 , the GNSS receiving unit 3 , and the map database 4 . The path plan generating device 100 may be a system that generates and corrects a path using only images captured by the vehicle-mounted camera 1 .
Claims
1. A path plan generating device, which generates a path plan of a vehicle using a first deep learning model based on an image captured by a vehicle-mounted camera, identifies left and right white lines forming a driving lane of the vehicle based on the image captured by the vehicle-mounted camera, and uses the recognition result of the white lines to correct the path plan, wherein the path plan generating device, Determine whether the left and right white lines meet a preset parallel condition, When it is determined that the left and right white lines do not satisfy the parallel condition, the path plan is not corrected using the recognition result of the white lines.
2. The path plan generating device according to claim 1, A second deep learning model is used to identify the left and right white lines based on the image captured by the vehicle-mounted camera, and the second deep learning model is a deep learning model different from the first deep learning model.
3. The path plan generating device according to claim 1 or 2, When the driving lane is a straight road, if the angle formed by the left and right white lines is equal to or larger than the permissible angle threshold, it is determined that the left and right white lines do not satisfy the parallel condition.
4. The path plan generating device according to any one of claims 1 to 3, Whenever the recognition result of the white line is updated, the lane width of the driving lane is calculated based on the recognition result of the white line, and the difference between the lane width before the update and the lane width after the update, i.e., the lane width difference, is calculated. When the lane width difference is greater than an allowable difference threshold, it is determined that the left and right white lines do not satisfy the parallel condition.
5. A path plan correction method for a path plan generating device, comprising: generating a path plan for a vehicle using a first deep learning model based on an image captured by a vehicle-mounted camera, identifying left and right white lines forming a driving lane of the vehicle based on the image captured by the vehicle-mounted camera, and correcting the path plan using the recognition result of the white lines, wherein: Determine whether the left and right white lines meet a preset parallel condition, When it is determined that the left and right white lines do not satisfy the parallel condition, the path plan is not corrected using the recognition result of the white lines.
