Feature point generation method

Abstract

The present invention provides a feature point generation method, a feature point generation device, a feature point generation program, and a recording medium storing the feature point generation program, all of which are capable of generating feature points on a road. [Solution] The server acquires the driving trajectory from the in-vehicle device via the network. The server extracts linear components from the driving trajectory, finds the intersection point between the extension of the extracted linear component and the extension of other linear components adjacent to the linear component in the driving trajectory, and if the driving trajectory passes through a predetermined range from that intersection point, generates that intersection point as a feature point.

Description

[Technical Field]

[0001] The present invention relates to a method for generating feature points. [Background Art]

[0002] When a vehicle travels on a new road that does not exist on the map represented by existing map data, such as a newly opened road, there is a known technique for updating map data by estimating the shape of the new road from the position information and traveling direction of the vehicle (Patent Document 1).

[0003] When applying the above technique to a road with many intersections and turning corners, such as a road in a parking lot, in addition to the shape, the generation of feature points for identifying intersections and turning corners is essential for using map data in navigation and autonomous driving. However, as an example, there is a problem that in the technique of Patent Document 1, only the shape of the new road is estimated and new feature points are not generated. [Prior Art Documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Unexamined Patent Application Publication No. 2014-235510 [Summary of the Invention] [Problems to be Solved by the Invention]

[0005] An example of the problem to be addressed by the present invention is to deal with such issues. That is, an object of the present invention is to provide, for example, a feature point generation method, a feature point generation device, a feature point generation program, and a recording medium recording the feature point generation program that can generate feature points of a road. [Means for Solving the Problems]

[0006] A feature point generation method according to claim 1, made to solve the above-mentioned problems, is a feature point generation method that generates feature points that are candidates for nodes of a travel path based on the travel trajectory of a moving object, and is characterized by comprising: an acquisition step of acquiring the travel trajectory; an extraction step of extracting a linear component from the travel trajectory; and a first generation step of generating the intersection of the extension of the extracted linear component and the extension of another linear component adjacent to the linear component in the travel trajectory as the feature point.

[0007] Furthermore, the feature point generation device according to claim 5 is a feature point generation device that generates feature points that are candidates for nodes on a travel path based on the travel trajectory of a moving body, and is characterized by including an acquisition unit that acquires the travel trajectory, an extraction unit that extracts linear components from the travel trajectory, and a first generation unit that generates the intersection of the extension of the extracted linear component and the extension of another linear component adjacent to the linear component in the travel trajectory as the feature point.

[0008] Furthermore, the feature point generation program described in claim 6 is a feature point generation program that causes a computer to generate feature points that are candidates for nodes on a travel path based on the travel trajectory of a moving object, characterized in that the computer is configured to function as an acquisition unit that acquires the travel trajectory, an extraction unit that extracts linear components from the travel trajectory, and a first generation unit that generates the intersection points of the extension of the extracted linear components and the extension of other linear components adjacent to the linear components in the travel trajectory as feature points.

[0009] Furthermore, the recording medium described in claim 7 is characterized in that it has the feature point generation program described in claim 7 recorded on it. [Brief explanation of the drawing]

[0010] [Figure 1] This is a system configuration utilizing a feature point generation device and a link information generation device according to one embodiment of the present invention. [Figure 2] Figure 1 is a flowchart illustrating the operation of the feature point generation device. [Figure 3]This is an explanatory diagram showing a concrete example of feature point generation. [Figure 4] This is an explanatory diagram showing a concrete example of feature point generation. [Figure 5] This is an explanatory diagram showing a concrete example of feature point generation. [Figure 6] This is an explanatory diagram showing a modified example of feature point generation. [Figure 7] This is an explanatory diagram illustrating a specific example of feature point integration. [Figure 8] Figure 1 is a flowchart illustrating the operation of the feature point generation device. [Figure 9] (A) shows the combined result when there are few travel trajectories, and (B) shows the combined result when there are many travel trajectories. [Figure 10] This is an explanatory diagram illustrating a specific example of link information generation. [Figure 11] This is an explanatory diagram illustrating a specific example of link information generation. [Figure 12] Figure 1 is a flowchart illustrating the operation of the link information generation device. [Figure 13] This is an explanatory diagram illustrating a specific example of link information generation. [Figure 14] This is an explanatory diagram illustrating a specific example of link information generation. [Modes for carrying out the invention]

[0011] A feature point generation method according to one embodiment of the present invention is a feature point generation method that generates feature points that are candidates for nodes of a travel path based on the travel trajectory of a moving object, and is characterized by comprising: an acquisition step of acquiring the travel trajectory; an extraction step of extracting linear components from the travel trajectory; and a first generation step of generating feature points at the intersection of the extension of the extracted linear component and the extension of another linear component adjacent to the linear component in the travel trajectory. This makes it possible to generate feature points of a travel path suitable for use in navigation and autonomous driving.

[0012] Also, when the travel locus does not exist within a predetermined range centered on the intersection point, or when the intersection point does not exist, an intersection point between an extension line of the linear component and a tangent line on the locus from a point deviating from the linear component to the other linear component among the travel locus, and an intersection point between the tangent line and an extension line of the other linear component are each generated as the feature point in a second generation step. Thereby, it is possible to generate feature points more suitable for use in navigation and autonomous driving.

[0013] Also, the tangent line may be a tangent line at the center on the locus.

[0014] Also, it may include a third generation step of generating, as the feature point, a point where the change in the inclination angle of the travel locus is a predetermined angle or more. Thereby, it is possible to generate feature points more suitable for use in navigation and autonomous driving.

[0015] Also, a feature point generation device according to an embodiment of the present invention is a feature point generation device that generates a feature point that is a candidate for a node of a travel route based on a travel locus of a moving body, and includes an acquisition unit that acquires the travel locus, an extraction unit that extracts a linear component from the travel locus, and a first generation unit that generates, as the feature point, an intersection point between an extension line of the extracted linear component and an extension line of another linear component adjacent to the linear component in the travel locus. Thereby, it is possible to generate feature points of a new road.

[0016] Also, it may be a feature point generation program for causing a computer to execute the above-described feature point generation method. Since it is a program executed by a computer in this way, dedicated hardware or the like is not required, and it can be installed and made to function in a general-purpose information processing device.

[0017] Also, the above-described feature point generation program may be stored in a computer-readable recording medium. By doing so, the program can be distributed not only when incorporated into a device but also as a single entity, and version updates and the like can be easily performed. [Examples]

[0018] The feature point generation device and link information generation device of the present invention will be described with reference to Figures 1 to 14. A server 10, which serves as a feature point generation device and link information generation device according to one embodiment of the present invention, comprises a communication unit 11, a feature point generation unit 12, a link information generation unit 13, and a storage unit 14.

[0019] The communication unit 11 communicates with the in-vehicle unit 20, which will be described later, via a network N such as the Internet. The communication unit 11 receives, for example, the vehicle's (mobile object's) location information (latitude, longitude), direction of travel, and inclination angle, as well as other driving trajectories, acquired by the in-vehicle unit 20. In addition, if the in-vehicle unit 20 is a device that requires map data, such as a navigation system, the communication unit 11 transmits the map data stored in the storage unit 14, which will be described later, in response to a request from the in-vehicle unit 20.

[0020] The feature point generation unit 12 generates feature points that are candidates for nodes on the road based on the road trajectory received from the communication unit 11.

[0021] The link information generation unit 13 generates links that connect the feature points generated by the feature point generation unit 12.

[0022] The memory unit 14 stores feature points generated by the feature point generation unit 12 and links generated by the link information generation unit 13. The memory unit 14 also stores and accumulates driving trajectories received from multiple in-vehicle devices 20. In addition, the memory unit 14 stores existing map data in which the vehicle's driving route is indicated by nodes, links, etc.

[0023] In the above description, the feature point generation unit 12 and the link information generation unit 13 function as the CPU (Central Processing Unit) or other computing devices of the server 10. The communication unit 11 functions as the network interface or other functions of the server 10. The storage unit 14 functions as the storage medium such as the hard disk of the server 10. Therefore, the flowcharts shown in Figures 2, 8, and 12, which are executed by the feature point generation unit 12 and the link information generation unit 13, can be configured as computer programs executed by the CPU to form a feature point generation program and a link information generation program. Furthermore, by having the feature point generation unit 12 and the link information generation unit 13 execute the flowcharts shown in Figures 2, 8, and 12, the server 10 implements the feature point generation method and the link information generation method of the present invention.

[0024] The in-vehicle unit 20 is installed in the vehicle and collects and transmits the driving trajectory to the server 10. Furthermore, if the in-vehicle unit 20 is configured as a navigation system, it requests generated map data from the server 10 as needed.

[0025] The in-vehicle unit 20 includes a communication unit 21, a control unit 22, and a vehicle information acquisition unit 23.

[0026] The communication unit 21 communicates with the server 10 via a network N such as the Internet. The communication unit 21 transmits the vehicle's trajectory, including its location information (latitude, longitude), direction of travel, and inclination angle, which has been acquired by the vehicle information acquisition unit 23. In this embodiment, the trajectory is composed of point cloud data representing the positions (latitude, longitude) on the trajectory traveled by the vehicle, as shown in Figure 3, for example, with the direction of travel and inclination angle at each point representing a position added to it.

[0027] Furthermore, if the in-vehicle device 20 is a device that requires map data, such as a navigation system, the communication unit 21 requests map data from the server 10 under control from the control unit 22, etc., and receives the map data transmitted from the server 10.

[0028] The control unit 22 is responsible for the overall control of the in-vehicle unit 20. The control unit 22 causes the communication unit 21 to transmit the vehicle's location information (latitude, longitude), direction of travel, inclination angle, etc., collected by the vehicle information acquisition unit 23 to the server 10 as a driving trajectory.

[0029] The vehicle information acquisition unit 23 is connected to various sensors and receivers such as a GPS (Global Positioning System) receiver, a gyro sensor, or a vehicle speed pulse sensor. The vehicle information acquisition unit 23 collects information such as vehicle position information (latitude, longitude), direction of travel, and tilt angle from the aforementioned sensors.

[0030] Furthermore, the in-vehicle unit 20 is not limited to a car navigation system installed on the instrument panel of a vehicle, but may also be a portable device such as a smartphone. In other words, the in-vehicle unit 20 may be detachably installed inside the vehicle.

[0031] In the above description, the control unit 22 functions as the CPU and other computing devices of the in-vehicle unit 20. The communication unit 21 functions as the communication circuits and antennas of the in-vehicle unit 20. The vehicle information acquisition unit 23 functions as the interface circuits for various sensors.

[0032] Next, the operation of the server 10 and the in-vehicle unit 20 with the above configuration will be explained. The in-vehicle unit 20 transmits the driving trajectory to the server 10 at predetermined intervals, such as every certain amount of time or every time it has traveled a certain distance. The server 10 functions as an acquisition unit and receives (acquires) the driving trajectory from the in-vehicle unit 20 and stores it sequentially in the storage unit 14.

[0033] The server 10 (feature point generation unit 12) described above performs the feature point generation process shown in Figure 2 at predetermined timings, such as each time a driving trajectory is received or at regular intervals. The server 10 reads one of the multiple driving trajectories stored in the storage unit 14 for which the feature point generation process has not yet been performed (step S1). Subsequently, the server 10 uses a well-known map matching technique to determine whether the read driving trajectory deviates from the driving route included in the map data (step S2).

[0034] When server 10 determines that the driving trajectory is a road included in the map data (N in step S2), it terminates the process.

[0035] On the other hand, if server 10 determines that the driving trajectory deviates from the road included in the map data (Y in step S2), it proceeds to steps S3 to S13 to generate feature points that are candidates for road nodes from that driving trajectory.

[0036] First, the server 10 acts as an extraction unit and extracts linear components from the travel trajectory using a well-known linear extraction technique (step S3). Next, as shown in Figure 3(A), the server 10 finds the intersection point between the extension LPn of the extracted linear component LSn and the extension LPn+1 of another linear component LSn+1 adjacent to the linear component LSn in the travel trajectory (step S4). If the server 10 can find the intersection point in step S4 (Y in step S5), it sets the intersection point found in step S4 as a feature point candidate Pcn, as shown in Figure 3(B) (step S6).

[0037] Next, the server 10 determines whether or not a travel trajectory exists within a predetermined range A1 centered on the feature point candidate Pcn (step S7). In this embodiment, the predetermined range A1 is the area within a circle centered on the feature point candidate Pcn, but it is not limited to this. The predetermined range A1 can have any shape as long as it is centered on the feature point candidate Pcn.

[0038] As shown in Figure 3(B), if a travel trajectory exists within a predetermined range A1 centered on the feature point candidate Pcn (Y in step S7), the server 10 functions as a first generation unit, as shown in Figure 3(C), confirms the feature point candidate Pcn as the feature point Pfn (step S8), and then proceeds to step S13.

[0039] On the other hand, if the server 10 does not have a travel trajectory within a predetermined range A1 centered on the feature point candidate Pcn, for example as shown in Figure 4(A), the server 10 proceeds to step S9. Also, if the server 10 does not have an intersection point for the extension lines LPn and LPn+1, for example as in the case of a trajectory that moves in a crank shape as shown in Figure 5, the server 10 proceeds to step S9 (N in step S5).

[0040] In step S9, as shown in Figures 4(B) and 5, the server 10 finds the tangent line Ltn at the center of the trajectory from point Pn where the trajectory deviates from the linear component LSn to point Pn+1 where it lies on another linear component LSn+1. The server 10 then designates the intersection point of the found tangent line Ltn and the extension of the linear component LSn, LPn, as feature point candidate Pcn1, and the intersection point of the found tangent line Ltn and the extension of the linear component LSn+1, LPn+1, as feature point candidate Pcn2 (step S10).

[0041] Subsequently, the server 10 determines, similar to step S7, whether or not a travel trajectory exists within a predetermined range A1 centered on feature point candidates Pcn1 and Pcn2 (step S11). If the server 10 determines that a travel trajectory exists within the predetermined range A1 centered on feature point candidates Pcn1 and Pcn2 (Y in step S11), it functions as a second generation unit, confirms feature point candidates Pcn1 and Pcn2 as feature points Pfn, and then proceeds to step S13. On the other hand, if the server 10 determines that no travel trajectory exists within the predetermined range A1 centered on feature point candidates Pcn1 and Pcn2 (N in step S11), it proceeds to step S13 without confirming them as feature points Pfn.

[0042] Subsequently, server 10 determines whether it has executed the process of determining feature points for all adjacent linear components in the travel trajectory from steps S4 to S12 (step S13). If feature points have not been determined for all adjacent linear components (N in step S13), it returns to step S4 and executes the process of determining feature points for the next adjacent linear components LSn+1 and LSn+2 on the travel trajectory.

[0043] According to the feature point generation process described above, server 10 extracts linear components from the driving trajectory and generates a feature point candidate Pcn as a feature point Pfn, which is the intersection point of the extension LPn of the extracted linear component LSn and the extension LPn+1 of another linear component LSn+1 adjacent to the linear component LSn in the driving trajectory. This makes it possible to generate feature points at locations that are candidates for nodes, such as corners and intersections in the driving path, and to generate feature points for driving paths that are suitable for use in navigation and autonomous driving.

[0044] Furthermore, according to the feature point generation process described above, if there is no travel trajectory within a predetermined range A1 centered on feature point candidate Pcn, which is the intersection of extension lines LPn and LPn+1, or if there is no intersection of extension lines LPn and LPn+1, feature point candidates Pcn1 and Pcn2, which are the intersections of tangent line Ltn and extension lines LPn and LPn+1 respectively, are generated as feature point Pfn.

[0045] As a result, as shown in Figure 4, when the radius of curvature of the driving trajectory is large, feature points Pfn can be generated near the corners of the road and close to the driving trajectory. This allows for more accurate map matching and enables the generation of feature points for roads that are even more suitable for navigation and autonomous driving.

[0046] In the above embodiment, if no travel trajectory exists within a predetermined range A1 centered on the intersection of extension lines LPn and LPn+1, a feature point Pfn is not generated at that intersection. However, this is not the only example. Without determining whether or not a travel trajectory exists within the predetermined range A1, if there is an intersection of extension lines LPn and LPn+1, that intersection may be used as a feature point Pfn.

[0047] Furthermore, if no travel trajectory exists within a predetermined range A1 centered on the intersection of extension lines LPn and LPn+1, it is not necessary to generate feature point Pfn based on extension lines LPn and LPn+1 without newly determining feature point candidates Pcn1 and Pcn2.

[0048] Furthermore, in the above embodiment, if there is no travel trajectory within a predetermined range A1 centered on the intersection of the tangent line Ltn and the extension lines LPn and LPn+1 (feature point candidates Pcn1 and Pcn2), feature point Pfn is not generated, but this is not the only example. Without determining whether or not there is a travel trajectory within a predetermined range A1 centered on feature point candidates Pcn1 and Pcn2, the intersection of the tangent line Ltn and the extension lines LPn and LPn+1 may be generated as feature point Pfn.

[0049] Furthermore, if no travel trajectory exists within a predetermined range A1 centered on feature point candidates Pcn1 and Pcn2, tangent lines may be drawn slightly shifted from the center toward points Pn and Pn+1, respectively, and the intersection points of these tangent lines with their extensions LPn and LPn+1 may be designated as new feature point candidates. Then, the tangent lines may be shifted toward points Pn and Pn+1 until a travel trajectory exists within the predetermined range A1 centered on the new feature point candidates.

[0050] Furthermore, in the above-described embodiment, feature points Pfn were generated from the position (latitude, longitude) of the driving trajectory, but this is not the only method. For example, as shown in Figure 6, the server 10 may define feature points Pfn as positions on the driving trajectory where the change in inclination angle is greater than or equal to a predetermined angle. In this embodiment, feature points Pfn and Pfn+1 are defined as positions on the driving trajectory where the change in inclination angle is 1 degree or more. This makes it possible to generate feature points Pfn and Pfn+1 at the boundaries between each floor of a self-propelled multi-story parking garage and the ramps connecting each floor. Feature points could also be generated using altitude data from GPS, but since altitude data from GPS is likely to have a large margin of error, more accurate feature points can be obtained by determining feature points based on the inclination angle as described above. Therefore, it is possible to generate feature points Pfn and Pfn+1 that are even more suitable for use in navigation and autonomous driving.

[0051] Additionally, a camera is installed in the vehicle, and the vehicle transmits images from the camera to the server 10. The server 10 analyzes the camera images, and if landmarks such as toll booths can be detected, it may generate those locations as feature points. Furthermore, the vehicle may be equipped with a narrowband communication device (e.g., DSRC) for communicating with toll booths and other landmarks, and the location information transmitted to the server 10 may be modified to indicate that communication has taken place. The server 10 generates the locations where communication occurred as feature points.

[0052] By performing the feature point generation process described above, feature points can be generated for each of multiple travel trajectories. As a result, as shown in Figure 7, multiple feature points Pf are generated for, for example, intersections on the same travel path.

[0053] Therefore, each time a feature point Pfn is determined by the feature point generation process described above (i.e., each time steps S9 and S12 are executed), the server 10 functions as an integration unit and executes the feature point integration process shown in Figure 8, which integrates multiple feature points Pfn.

[0054] In the feature point integration process, the server 10 determines whether there are any existing feature points Pfn that were determined before the determined feature point Pfn within a predetermined range A2 centered on the determined feature point Pfn (step S21). The predetermined range A2 may be the same as the predetermined range A1 used in the feature point generation process, or it may be different. If there are existing feature points Pfn within the predetermined range A2 (Y in step S21), the server 10 calculates the average position of the two feature points Pfn that exist within the predetermined range A2 (step S22). Then, the server 10 integrates them into a feature point at the average position (step S23).

[0055] Subsequently, the server 10 deletes the two feature points Pfn before integration, saves the integrated feature point Pfn as an existing feature point Pfn in the storage unit 14 (step S24), and terminates the process. On the other hand, if no existing feature point Pfn exists within the predetermined range A2 (N in step S21), the server 10 saves the newly determined feature point Pfn as an existing feature point in the storage unit 14 (step S24), and terminates the process.

[0056] According to the feature point merging process described above, when multiple generated feature points Pfn exist within a predetermined range A2, these multiple feature points Pfn can be merged. This prevents multiple feature points Pfn from being attached to a single corner or intersection on the road, making it possible to generate feature points suitable for use in navigation and autonomous driving.

[0057] Furthermore, the feature point integration process described above integrates the locations (latitude and longitude) of multiple feature points Pfn into a single feature point that is the average of these locations. This makes it possible to generate feature points Pfn that are even more suitable for use in navigation and autonomous driving.

[0058] The feature point integration process described above yields different results depending on the number of travel trajectories. Figure 9(A) shows the feature point integration result when there are few travel trajectories. Figure 9(B) shows the feature point integration result when there are many travel trajectories. When there are many travel trajectories, the position of feature point 1 shifts slightly lower than when there are few. Generally, it is thought that the more travel trajectories there are, the closer the feature points will be to the average of the actual routes traveled by vehicles. By generating feature points that correspond to actual routes in this way and using that map data, the accuracy of map matching can be improved.

[0059] The feature point integration process described above was performed each time a feature point was generated by the feature point generation process, but it is not limited to this. The feature point integration process may be performed after the feature point generation process has been executed for multiple travel trajectories (i.e., after the feature point generation process has been executed multiple times). In this case as well, the latitude and longitude of multiple feature points within a predetermined range A2 are averaged and integrated into that averaged point. Furthermore, if there are multiple feature points within the predetermined range A2 that have the same latitude and longitude, they may be integrated into the latitude and longitude with the largest number of such points. In addition, the feature points may be integrated into the center position of the figure formed by connecting multiple feature points within the predetermined range A2.

[0060] Furthermore, while the feature point integration process described above integrated feature points generated from the extensions of the driving trajectory LPn and LPn+1, it is conceivable that feature points generated from the slope angle could also be integrated in a similar manner. For feature points generated from the slope angle, the slope angle is added to the feature point. If there are multiple feature points with the same slope angle within a predetermined range A2, the feature points may be integrated by averaging their latitude and longitude.

[0061] Next, the server 10 generates links connecting the feature points that have been integrated by the feature point integration process described above. The server 10 (specifically its link information generation unit 13) basically connects the feature points Pf7 and Pf18 to each other in the order in which they pass through a single travel trajectory R1 that passes through a predetermined range A3, as shown in Figure 10. The predetermined range A3 may be the same as or different from the predetermined ranges A1 and A2 used in the feature point generation process and the feature point integration process. In addition, the connection direction of the feature points (the direction in which the vehicle passes) may be added to the link data at this time.

[0062] However, in this link generation method, when a link connecting feature point Pf3 and feature point Pf2 is generated along the travel trajectory R2, as shown in Figure 11(B), the travel trajectory R2 passes within a predetermined range A3 centered on feature point Pf8, which was generated based on the travel trajectory R1. As a result, a link connecting feature point Pf2 → feature point Pf8 → feature point Pf3 is generated, as shown in Figure 11(A).

[0063] Therefore, in this embodiment, the link generation process shown in Figure 12 is executed. First, the server 10 extracts three feature points from among the multiple feature points generated in the feature point generation process and integrated in the feature point integration process, where the same travel trajectory passes through a predetermined range A3 and the order in which the trajectory passes is consecutive (step S30). For example, in the example shown in Figure 11, feature points Pf2, Pf8, and Pf3 are extracted as three feature points where the travel trajectory R2 (one of the multiple travel trajectories) passes through a predetermined range A3 and the order in which the trajectory passes is consecutive.

[0064] Next, the server 10 calculates a first evaluation result (step S31) by evaluating the distance between the line L1 connecting the three extracted consecutive feature points Pf2, Pf8, and Pf3, and the travel trajectory R2 that passes through a predetermined range A3 from each of the three feature points Pf2, Pf8, and Pf3. In this embodiment, as shown in Figure 11(A), the server 10 takes the maximum value of the distance until the normal vector on the travel trajectory R2 (indicated by the arrow in the figure) intersects with line L1 as the first evaluation result. In the example shown in Figure 11(A), the maximum value is the distance until the normal vector intersects with feature point Pf8.

[0065] Next, the server 10 calculates a second evaluation result by evaluating the distance between the line L2, which connects the first and last feature points Pf2 and Pf3 among the three feature points Pf2, Pf8, and Pf3, and the travel trajectory R2 (step S32). In this embodiment, as shown in Figure 11(B), the server 10 takes the maximum value among the distances until the normal on the travel trajectory R2 intersects with line L2 as the second evaluation result. In the example shown in Figure 11(B), the maximum value is the distance until the normal intersects with feature point Pf2.

[0066] Next, the server 10 compares the first evaluation result with the second evaluation result. If the first evaluation result is greater than the second evaluation result by a certain value or more (Y in step S33), it does not connect the middle feature point Pf8, but instead generates a link connecting the preceding and succeeding feature points Pf2 and Pf3, and terminates the process (step S34). On the other hand, if the first evaluation result is not greater than the second evaluation result by a certain value or more (N in step S33), the server 10 connects the three consecutive feature points Pf2, Pf8, and Pf3 and terminates the process (step S35).

[0067] In particular, in parking lot pathways, the direction of travel is often precisely defined. In the example shown in Figure 11, a pole cone is present between Pf2 and Pf8, and the link between Pf2 and Pf8 was inappropriate. According to the embodiment described above, in the example shown in Figure 11, a link connecting feature points Pf2 and Pf3 is generated. As a result, the accuracy of the link is improved, and link information suitable for use in navigation and autonomous driving can be generated.

[0068] Note that the first and second evaluation results are not limited to the embodiments described above. The first evaluation result may be, for example, the area enclosed by the travel trajectory R2, line L1, the normal vector L3 on the travel trajectory R2 passing through the first feature point Pf3, and the normal vector L4 on the travel trajectory R2 passing through the last feature point Pf2 (shown by the shaded area in the figure), as shown in Figure 13(A). The second evaluation result may be, for example, the area enclosed by the travel trajectory R2, line L2, and normal vectors L3 and L4 (shown by the shaded area in the figure), as shown in Figure 13(B).

[0069] Furthermore, as shown in Figure 14(A), the first evaluation result may be the maximum distance among the distances from the normal on the travel trajectory R2 to the point where it intersects with each of the three feature points Pf2, Pf8, and Pf3. Also, as shown in Figure 14(B), the second evaluation result may be the maximum distance among the distances from the normal on the travel trajectory R2 to the point where it intersects with the preceding and succeeding feature points Pf2 and Pf3. In this case, similar to the embodiment shown in Figure 11, the distance from the normal on the travel trajectory R2 to the point where it intersects with feature point Pf8 becomes the first evaluation result, and the distance from the normal on the travel trajectory R2 to the point where it intersects with feature point Pf2 becomes the second evaluation result.

[0070] Similarly, by defining first and second evaluation results in this way, the accuracy of the links improves, and link information suitable for use in navigation and autonomous driving can be generated.

[0071] In the above-described embodiment, links were generated based on whether the first evaluation result was greater than the second evaluation result by a certain value or more, that is, whether the difference between the first and second evaluation results was greater than a certain value or more. However, this is not the only way. Links may also be generated based on whether the first evaluation result is greater than the second evaluation result. To explain in more detail, if the first evaluation result > the second evaluation result, a link is generated connecting the feature points Pf2 and Pf3 before and after the middle feature point Pf8, without connecting the middle feature point Pf8. On the other hand, if the first evaluation result ≤ the second evaluation result, a link is generated by connecting the three feature points Pf2, Pf8, and Pf3.

[0072] Furthermore, in the embodiment described above, the server 10 extracted three consecutive feature points in the link information generation process, but this is not limited to this. It may also extract three or more consecutive feature points.

[0073] Furthermore, in the above-described embodiment, the server 10, which can communicate with the in-vehicle device 20 via the network N, generated the feature points and link information, but this is not the only way. The driving trajectory may be stored in a memory that can be attached to and removed from the in-vehicle device 20, and the feature points and link information may be generated using a device that reads the memory after it has been removed from the in-vehicle device 20. Alternatively, the in-vehicle device 20 may generate feature points from its own driving trajectory and transmit them to the server 10.

[0074] Furthermore, while the above-described embodiment involved the generation, integration, and generation of link information of feature points on a single server 10, this is not the only option. The generation, integration, and generation of link information of feature points may be performed on separate servers.

[0075] Furthermore, while the above-described embodiment primarily assumed a driving path within a parking lot, it is not limited to this. The driving path could also be a public road or other public road.

[0076] It should be noted that the present invention is not limited to the embodiments described above. That is, it can be implemented with various modifications without departing from the core principles of the present invention. [Explanation of symbols]

[0077] 10. Server (feature point generation device, link information generation device, acquisition unit) 12. Feature point generation unit (extraction unit, first generation unit, second generation unit, generation unit, integration unit) 13 Link Information Generation Unit

Claims

[Claim 1] A feature point generation device that generates feature points that are candidates for nodes on a travel path based on the travel trajectory of a moving object, The acquisition unit acquires the aforementioned driving trajectory, An extraction unit that extracts a linear component from the aforementioned travel trajectory, A feature point generation device comprising a first generation unit that generates feature points at the intersection of the extension of the extracted linear component and the extension of another linear component adjacent to the linear component in the travel trajectory.

Images