Road map generation device, road map generation method, and road map generation program

Abstract

To generate road map data properly corrected in consideration of a relation between adjacent lanes.SOLUTION: A road map generation device comprises: a position information extraction unit 1 that extracts position information measured by a moving body sensor; a lane unit statistic processing unit 3 that performs statistic processing of a first position coordinate point sequence for each lane to generate a second position coordinate point sequence; an approximate line calculation unit 4 that calculates an approximate line of the second position coordinate point sequence; a statistic processing reliability calculation unit 5 that calculates reliability of each second position coordinate point; an approximate line reliability calculation unit 6 that calculates reliability of the approximate line; an adjacent lane information generation unit 7 that identifies an adjacent lane group to which that the approximate line belongs; and an adjacent lane shape correction unit 8 that unifies a shape of each adjacent approximate line on the basis of the reliability of each adjacent approximate line, and moves a position of each adjacent approximate line in a direction vertical to an azimuth of each adjacent approximate line by a movement distance to be determined based on a lane width defined by an interval between two adjacent approximate lines, and the reliability of each adjacent approximate line.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present disclosure relates to the generation of a road map appropriately corrected in consideration of the relationship between adjacent lanes.

Background Art

[0002] In a driving support system such as a navigation system, an automatic driving system, or an ADAS (Advanced Driver Assistance System) that provides route guidance, road map data is used. If the road map data deviates from the actual road conditions or the road environment, restrictions may occur in the services provided or the safety may be impaired. Therefore, a method of always maintaining the latest state has been considered.

[0003] Patent Document 1 describes a technique for generating updated map data using in-vehicle sensor information as an input, calculating a reliability that serves as an accuracy index, and performing an update based on the reliability and the number of times the updated map data is detached from the map database.

[0004] In addition, Patent Document 2 describes a technique for collecting GNSS positioning data, calculating an approximate line using a straight line, an arc, or a clothoid curve, and generating a road map having a shape in which the curvature smoothly changes.

[0005] Furthermore, Patent Document 3 describes a technique for correcting information on a lane having an azimuth angle or a curvature different from that of an adjacent lane by referring to the information on the adjacent lane.

[0006] In addition, Patent Document 4 describes a technique for generating road and lane structure information from sensor information and correcting map data according to the reliability.

Prior Art Documents

Patent Documents

[0007]

Patent Document 1

Patent Document 2

[0008] In the technology described in Patent Document 1, in areas where the sensor's environmental recognition accuracy is high, the assigned data is updated with high reliability. On the other hand, in areas where the sensor's environmental recognition accuracy is consistently low, such as intersections or lanes where blurring occurs, there is no effect in estimating and correcting the map data for that area from the surrounding map shape.

[0009] The technology described in Patent Document 2 can generate a road map with smoothly changing curvature by calculating an approximation line using a straight line, arc, or clothoid curve, even if the GNSS positioning data contains missing values ​​or positioning errors. Furthermore, the technology in Patent Document 2 calculates an approximation line in which the error with the original positioning data is within a certain range, and various items such as road width, curvature, and gradient, which are elements specified in design specifications such as the Road Structure Ordinance, fall within the range. However, cases may occur where the shape or lane width differs between adjacent lanes. In addition, if the original positioning data contains large errors or incorrect information, it may be greatly affected.

[0010] The technology described in Patent Document 3 modifies map data by referring to the azimuth angle or curvature of adjacent lanes, but since the data for adjacent lanes is not always accurate, there is a possibility of making incorrect modifications. Furthermore, as a result of the modifications, a shape in which the curvature changes abruptly may be generated.

[0011] The technology described in Patent Document 4 determines whether to correct road and lane structure information generated based on sensor data according to its reliability. However, correcting only a portion of the road and lane structure information may result in the generation of shapes that are rare in real roads, such as unnatural lane connections or differences in shape or lane width between adjacent lanes.

[0012] This disclosure is made to solve the above-mentioned problems and aims to generate road map data that is appropriately corrected considering the relationship between adjacent lanes. [Means for solving the problem]

[0013] The road map generation device of this disclosure includes: a position information extraction unit that extracts position information representing the position of a mobile body measured by a mobile body sensor mounted on the mobile body as a sequence of first position coordinate points consisting of a plurality of first position coordinate points; a driving lane determination unit that determines the lane the mobile body was traveling in at each first position coordinate point based on images captured by a camera mounted on the mobile body and assigns it to the first position coordinate point; a lane unit statistical processing unit that performs statistical processing on the sequence of first position coordinate points for each lane to generate a sequence of second position coordinate points consisting of a plurality of second position coordinate points; an approximation line calculation unit that calculates an approximation line for the sequence of second position coordinate points for each lane; a statistical processing reliability calculation unit that calculates the reliability of each second position coordinate point based on the number of source first position coordinate points and the accuracy of acquisition of the source first position coordinate points by the mobile body sensor; and the reliability of the approximation line. The system includes: an approximation line reliability calculation unit that calculates the difference of the approximation line to the original second position coordinate point sequence and the reliability of each second position coordinate point constituting the second position coordinate point sequence from which the approximation line was created; an adjacent lane information generation unit that identifies the group of adjacent lanes to which the approximation line belongs based on the distance between the second position coordinate point sequences from which each approximation line was created and the direction of travel of the moving object along each approximation line; and an adjacent lane shape correction unit that unifies the shape of each adjacent approximation line based on the reliability of each adjacent approximation line which constitutes each adjacent lane group, corrects the group of adjacent lanes by moving the position of each adjacent approximation line in a direction perpendicular to the orientation of each adjacent approximation line by a distance determined based on the lane width defined by the distance between two adjacent adjacent approximation lines and the reliability of each adjacent approximation line, and stores the corrected adjacent lane group data in a storage device. [Effects of the Invention]

[0014] The road map generation device of this disclosure can generate road map data that has been appropriately corrected, taking into account the relationship between adjacent lanes. [Brief explanation of the drawing]

[0015] [Figure 1] This is a diagram showing the configuration of the road map generation system according to Embodiment 1-5. [Figure 2] This is a diagram showing the server configuration according to Embodiment 1. [Figure 3]It is a diagram showing an operation image example of a lane unit statistical processing unit according to Embodiment 1. [Figure 4] It is a diagram showing an operation image example of an adjacent lane information generation unit according to Embodiment 1. [Figure 5] It is a diagram showing an operation image example of an adjacent lane shape correction unit according to Embodiment 1. [Figure 6] It is a diagram showing an operation image example of an adjacent lane shape correction unit according to Embodiment 1. [Figure 7] It is a diagram showing an operation image example of an adjacent lane shape correction unit according to Embodiment 1. [Figure 8] It is a flowchart showing the processing of a server according to Embodiment 1. [Figure 9] It is a configuration diagram of a server according to Embodiment 2. [Figure 10] It is a configuration diagram of a server according to Embodiment 3. [Figure 11] It is a diagram showing an operation image example of an adjacent lane shape correction unit according to Embodiment 3. [Figure 12] It is a diagram showing an operation image example of an adjacent lane shape correction unit according to Embodiment 3. [Figure 13] It is a configuration diagram of a server according to Embodiment 4. [Figure 14] It is a diagram showing an operation image example of a lane shape estimation unit and a lane shape extraction unit according to Embodiment 4. [Figure 15] It is a diagram showing an operation image example of a lane shape estimation unit and a lane shape extraction unit according to Embodiment 4. [Figure 16] It is a diagram showing an operation image example of a lane shape estimation unit and a lane shape extraction unit according to Embodiment 4. [Figure 17] It is a diagram showing an operation image example of a lane shape estimation unit and a lane shape extraction unit according to Embodiment 4. [Figure 18] It is a flowchart showing the processing of a server according to Embodiment 4. [Figure 19] It is a configuration diagram of a server according to Embodiment 5. [Figure 20]It is a configuration diagram of a moving body according to Embodiment 6.

Embodiments for Carrying Out the Invention

[0016] <A. Embodiment 1> <A-1. Configuration> FIG. 1 is a block diagram showing the system configuration of a road map generation system 1000 according to Embodiments 1-5. The road map generation system 1000 includes a server 101 and a moving body 102. There may be one or more moving bodies 102. In this embodiment, a vehicle is assumed as the moving body 102. In Embodiments 1-5, since the configuration of the server 101 is different in each case, in this embodiment, the server 101 is denoted as server 101A to distinguish it from other embodiments.

[0017] FIG. 2 shows the configuration of the server 101A. As shown in FIG. 2, the server 101A includes a processor 100, a storage device 40, a communication interface 20, and a communicator 30. The server 101A is a computer server and can be an edge server, a cloud server, etc. The server 101A may also be referred to as an edge server or a cloud server.

[0018] The processor 100 is connected to other hardware via signal lines and controls these other hardware. The processor 100 is an IC (Integrated Circuit) for executing programs stored in the storage device 40 to perform processes such as data transfer, calculation, processing, control, and management. The processor 100 has an arithmetic circuit, registers, and cache memories in which instructions and information are stored. Specifically, the processor 100 is a CPU (Central Processing Unit), a DSP (Digital Signal Processor), or a GPU (Graphics Processing Unit).

[0019] In Embodiment 1, the processor 100 constitutes a road map generation device. The road map generation device of Embodiment 1 includes a location information extraction unit 1, a driving lane determination unit 2, a lane unit statistical processing unit 3, an approximation line calculation unit 4, a statistical processing reliability calculation unit 5, an approximation line reliability calculation unit 6, an adjacent lane information generation unit 7, and an adjacent lane shape correction unit 8. That is, the processor 100 executes a program stored in the storage device 40, thereby realizing the functions of the location information extraction unit 1, the driving lane determination unit 2, the lane unit statistical processing unit 3, the approximation line calculation unit 4, the statistical processing reliability calculation unit 5, the approximation line reliability calculation unit 6, the adjacent lane information generation unit 7, and the adjacent lane shape correction unit 8.

[0020] The storage device 40 is RAM (Random Access Memory), HDD (Hard Disk Drive), or SSD (Solid State Drive). Alternatively, the storage device 40 may be a portable storage medium such as an SD (Secure Digital, registered trademark) memory card, CF (CompactFlash, registered trademark), NAND flash, flexible disk, optical disk, compact disk, Blu-ray (registered trademark) disk, or DVD.

[0021] The communication interface 20 is a device that includes a receiver for receiving data from the mobile device 102 and a transmitter for transmitting data. Specifically, the communication interface 20 is a communication chip or a NIC (Network Interface Card). The communication interface 20 uses communication protocols such as DSRC (Dedicated Short Range Communication) and IEEE 802.11p, which are dedicated to vehicle communication. The communication interface 20 may also use cellular networks such as LTE (Long Term Evolution, registered trademark) or 5G. The communication interface 20 may also use Bluetooth (registered trademark) or wireless LAN such as IEEE 802.11a / b / g / n / ac.

[0022] The server 101 and the mobile unit 102 may not be connected by communication. In that case, the sensor information stored in the storage device of the mobile unit 102 shall be manually stored in the storage device 40 of the server 101.

[0023] The storage device 40 stores the program to be executed by the processor, as well as the mobile sensor information management table 9 and the road map data 10. The road map data 10 is the output result of the adjacent lane shape correction unit 8. The mobile sensor information management table 9 stores the GNSS sensor information and camera sensor information received by the server 101 from the mobile body 102 via the communication interface 20.

[0024] The GNSS sensor information stored in the mobile sensor information management table 9 includes the identification number, position, speed, and direction of the mobile object 102, the date and time of information acquisition, and accuracy information at the time of information acquisition. The position information of the mobile object 102 included in the GNSS sensor information may also include altitude information of the mobile object 102.

[0025] The camera sensor information stored in the mobile sensor information management table 9 includes the identification number of the mobile body 102 and the camera image. The camera sensor information may also include accuracy information at the time of information acquisition, driving lane number, driving lane position information, driving lane type information, driving lane color information, driving lane width information, driving lane recognition accuracy information, lane position information of adjacent lanes, lane type information of adjacent lanes, lane color information of adjacent lanes, lane width information of adjacent lanes, recognition accuracy information of adjacent lanes, shoulder boundary position information, and shoulder width information.

[0026] Road map data 10 is map information relating to road maps, i.e., road map information. Road map information includes the shape of lane centerlines, latitude, longitude, curvature, lane centerline connection information, adjacent lane information, and reliability information. Road map information may also include gradient.

[0027] The location information extraction unit 1 acquires location information that is obtained at a date and time specified by the user and that falls within a location range specified by the user. Among the acquired location information, the location information extraction unit 1 assigns sequential numbers to location information belonging to the same journey of the same mobile object in order of acquisition date and time. The same journey refers to, for example, driving data from engine start to engine stop. For example, two pieces of location information whose acquisition dates and times are separated by a certain amount of time are determined to belong to different journeys. The location information extraction unit 1 also deletes location information while the mobile object 102 is stopped and location information that exists at distance intervals shorter than the assumed accuracy of the generated map.

[0028] The lane determination unit 2 acquires the camera image closest to the location information extracted by the location information extraction unit 1, determines the lane the moving object 102 is traveling in using image recognition technology such as machine learning, and adds information indicating the determined lane (hereinafter referred to as lane information) to the location information. The lane information indicates, for example, which lane the moving object 102 is traveling in from the left. If the camera sensor information includes information other than the camera image that can be used to determine the lane the moving object 102 is traveling in, the lane determination unit 2 may use that information to determine the lane the moving object 102 is traveling in.

[0029] The lane-unit statistical processing unit 3 performs statistical processing on the location information extracted by the location information extraction unit 1 for each driving lane determined by the driving lane determination unit 2. One example of statistical processing is weighted averaging. For example, if the location information extracted by the location information extraction unit 1 includes two types of GNSS sensor information: GNSS sensor information with an accuracy of approximately 10m and GNSS sensor information with an accuracy of approximately 1m, the lane-unit statistical processing unit 3 sets the weight of the GNSS sensor information with an accuracy of approximately 10m to 0.1 and the weight of the GNSS sensor information with an accuracy of 1m to 1. In addition, the lane-unit statistical processing unit 3 increases or decreases the weight of the GNSS sensor information according to the accuracy information at the time of information acquisition included in the GNSS sensor information.

[0030] Figure 3 shows an example of the operation of the lane-unit statistical processing unit 3. In Figure 3, the position information extracted by the position information extraction unit 1 is represented as position coordinate points p1 and p2. The position coordinate points p1 and p2 measured by the mobile sensor and extracted by the position information extraction unit 1 are also referred to as first position coordinate points, and the sequence of first position coordinate points is also referred to as the first position coordinate point sequence. If position coordinate points p1 and p2 are included in the same journey d1 of the same mobile body 102, the lane-unit statistical processing unit 3 generates a link l1 connecting position coordinate points p1 and p2 in order of acquisition date and time. The lane-unit statistical processing unit 3 also generates a perpendicular line pl1 passing through position coordinate point p1, which is the starting point of l1, with a predetermined length. The lane-unit statistical processing unit 3 then acquires a link l2 that has an intersection with perpendicular line pl1, where the error with the azimuth angle of link l1 is within a predetermined range, the driving lane determined by the driving lane determination unit 2 matches, and the link l2 has an intersection with perpendicular line pl1. The lane-unit statistical processing unit 3 calculates the intersection point pi1 between link l2 and perpendicular line pl1. The lane-unit statistical processing unit 3 sets the weight of intersection point pi1 using the position coordinate points p3 and p4 at both ends of link l2 that are closer to intersection point pi1. The lane-unit statistical processing unit 3 calculates point pj1 by weighted average of position coordinate point p1 and intersection point pi1.

[0031] The weighted average can be expressed, for example, by the following equation (1). In equation (1), ω i The position coordinate point I i (x i ,y i This represents the weight of ). Note that i is a natural number from 1 to n.

[0032]

number

[0033] The lane-unit statistical processing unit 3 calculates the average bearing of the position coordinate point p1 used to calculate point pj1 and links l1 and l2 to which the intersection point pi1 belongs, and calculates position coordinate point pj2 which is advanced a distance shorter than the assumed accuracy of the generated map in the direction of the average bearing. Then, the lane-unit statistical processing unit 3 generates a perpendicular line pl2 of a predetermined length passing through position coordinate point pj2. Then, the lane-unit statistical processing unit 3 obtains a link where the error with the azimuth angle of link l2 is within a predetermined range, the driving lane determined by the driving lane determination unit 2 matches, and the link has an intersection with perpendicular line pl2. The lane-unit statistical processing unit 3 calculates the next point by taking the weighted average of the intersection point and position coordinate point pj2. The lane-unit statistical processing unit 3 generates a sequence of position coordinate points calculated in this way. The position coordinate points calculated by the lane unit statistical processing unit 3 through statistical processing of the first position coordinate points are also referred to as second position coordinate points, and the sequence of second position coordinate points is also referred to as the second position coordinate point sequence.

[0034] Furthermore, when the lane-unit statistical processing unit 3 calculates a weighted average, if there are more data points than a predetermined threshold, it may remove points with weights smaller than the predetermined threshold to improve speed.

[0035] The approximation line calculation unit 4 generates a set of approximation lines, such as straight lines, circular arcs, or clothoid curves, for the second position coordinate point sequence for each lane generated by the lane unit statistical processing unit 3.

[0036] The statistical processing reliability calculation unit 5 calculates the reliability of the second position coordinate point generated by the lane unit statistical processing unit 3 and assigns it to the second position coordinate point. For example, the reliability of the second position coordinate point is calculated based on the number of first position coordinate points used in the statistical processing that generates the second position coordinate point and the accuracy of the measurement of the first position coordinate point by the moving sensor. Alternatively, the statistical processing reliability calculation unit 5 may calculate the reliability using the following equation (2), which calculates the sum of the weights of each first position coordinate point when taking a weighted average. In equation (2), ω i Similar to equation (1), the first position coordinate point I i (x i ,y i This represents the weight of ).

[0037]

number

[0038] The approximation line reliability calculation unit 6 calculates the reliability of the approximation line generated by the approximation line calculation unit 4. For example, the reliability of the approximation line is calculated based on the difference of the approximation line to the second position coordinate point sequence from which the approximation line was created, and the reliability of each second position coordinate point in the second position coordinate point sequence from which the approximation line was created. Alternatively, the approximation line reliability calculation unit 6 may calculate the reliability of the approximation line by summing the reliability of second position coordinate points whose difference from the approximation line is within a predetermined threshold.

[0039] The adjacent lane information generation unit 7 extracts the sequence of second position coordinate points that are the source of each approximate line from the set of approximate lines generated by the approximate line calculation unit 4, and generates adjacent lane information from the distance between the second position coordinate points and the direction of travel of the moving body at the second position coordinate points. If the altitude information of the moving body 102 is stored in the moving body sensor information management table 9, adjacent lane information may be generated based on the altitude at the first position coordinate point that is the source of the second position coordinate points.

[0040] For example, in Figure 4, approximation line A10 is created based on position coordinate points p10-p15, approximation line A11 is created based on position coordinate points p16-p22, and approximation line A12 is created based on position coordinate points p23-p28. Note that approximation lines A10-A12 are omitted from the illustration in Figure 4. The adjacent lane information generation unit 7 generates a perpendicular line pl10 of link l10 connecting position coordinate points p10 and p11 with a predetermined length, and obtains link l11 which intersects with perpendicular line pl10. If the difference in azimuth angles between links l10 and l11 is less than a predetermined threshold, the adjacent lane information generation unit 7 considers link l11 to be the adjacent lane to link l10. Furthermore, if the GNSS sensor information stored in the mobile sensor information management table 9 includes altitude information, in addition to the azimuth angle condition described above, link l11 may be considered an adjacent lane to link l10 if the altitude difference between link l11 and link l10 is within a predetermined range.

[0041] Similarly, the adjacent lane information generation unit 7 generates a perpendicular line for link l11 passing through position coordinate point p16, and obtains link l12 which intersects with this perpendicular line. In this way, it is determined that links l11 and l12 are adjacent lanes to link l10, that is, the approximate line A10-A12 is a group of adjacent lanes.

[0042] Similarly, the perpendicular line pl11 of the link passing through position coordinate point p25 identifies approximation line A11 as an adjacent lane to approximation line A12. However, depending on the perpendicular line passing through position coordinate point p26, the difference in azimuth angles of links l13 and l14 exceeds a predetermined threshold, so approximation line A11 is not extracted as an adjacent lane to approximation line A12. In this case, the adjacent lane information generation unit 7 uses position coordinate point p25 as the boundary of the adjacent lane area where approximation line A12 is treated as an adjacent lane to approximation lines A10 and A11, extends the perpendicular line pl11 of position coordinate point p25, and divides approximation lines A10, A11, and A12 at the point where they intersect with the perpendicular line pl11. In this way, the adjacent lane information generation unit 7 identifies the adjacent lane group and adjacent lane area.

[0043] If there are multiple sequences of position coordinates for the same lane within the group of adjacent lanes, the adjacent lane information generation unit 7 may compare the length of the sequence with the sum of the confidence values ​​of the sequences and adopt the one that is longer and has a larger sum of confidence values.

[0044] The adjacent lane shape correction unit 8 extracts the adjacent lane group identified by the adjacent lane information generation unit 7 and corrects each approximation line constituting the adjacent lane group in a direction that unifies the lane shape and lane width within the adjacent lane group. Hereinafter, the approximation lines constituting the adjacent lane group will also be referred to as adjacent approximation lines. The adjacent lane shape correction unit 8 unifies the shape of each adjacent approximation line according to the reliability of each adjacent approximation line. For example, the adjacent lane shape correction unit 8 adopts the shape of the adjacent approximation line with the highest reliability as the shape of the other adjacent approximation lines. Subsequently, the adjacent lane shape correction unit 8 arranges each adjacent approximation line so that the difference with the second position coordinate point sequence from which each adjacent approximation line was created becomes small.

[0045] Furthermore, the adjacent lane shape correction unit 8 determines the movement distance of each adjacent approximation line according to the reliability of each adjacent approximation line, and performs position correction in a direction in which the lane width defined by the interval between adjacent approximation lines becomes equal in the group of adjacent lanes.

[0046] For example, let's consider the approximation lines le101, le102, and le103 shown in Figure 5 as a group of adjacent lanes. Let the confidence levels of the approximation lines le101, le102, and le103 be w101, w102, and w103, respectively. If w101 > w103 > w102, the adjacent lane shape correction unit 8 adopts the shape of the approximation line le101, which has the highest confidence level, as the shape of the other approximation lines le102 and le103. The adjacent lane shape correction unit 8 may also adopt the shape of an approximation line that does not have the highest confidence level among the group of adjacent lanes but is greater than a predetermined threshold.

[0047] As shown in Figure 6, the adjacent lane shape correction unit 8 adopts the shape of approximation line le101 for approximation line le102 and replaces it with approximation line le202, which is positioned to minimize the error with the position coordinate point on which approximation line le102 was based. Similarly, approximation line le103 is replaced with approximation line le203. Note that in this example, there is no change to approximation line le101 in Figure 5, but in Figure 6, the same approximation line as approximation line le101 is denoted as approximation line le201.

[0048] Next, the adjacent lane shape correction unit 8 corrects at least one of the approximation lines le201, le202, and le203 by moving it parallel perpendicular to its orientation so that the lane widths formed by the approximation lines le201 and le202 are equal to the lane widths formed by the approximation lines le202 and le203. The corrected approximation lines le201, le202, and le203 are designated as approximation lines le301, le302, and le303, respectively. Figure 7 shows the result of moving approximation line le202 toward approximation line le301 by a distance of m302. When approximation line le201 is moved parallel, the distance of movement is m301, and when approximation line le203 is moved parallel, the distance of movement is m303. The lane width formed by approximation lines le301 and le302 is set to dw300, and the lane width formed by approximation lines le302 and le303 is set to dw301.

[0049] The adjacent lane shape correction unit 8 may create an evaluation function F1 as shown in Equation (3) using the lane width error, the movement distance, and the reliability, and determine the movement distance of each approximate line so that the evaluation function F1 takes the minimum value.

[0050]

Equation

[0051] <A-2. Operation> FIG. 8 is a flowchart showing the operation of the server 101A according to the first embodiment. Hereinafter, the operation of the server 101A will be described with reference to FIG. 8. The operation of the server 101A according to the first embodiment corresponds to the processing of the road map generation method according to the first embodiment or the road map generation program according to the first embodiment.

[0052] First, in step S101, the position information extraction unit 1 acquires the position information of the moving body 102 from the moving body sensor information management table 9 in the storage device 40. This position information includes a first position coordinate point sequence representing the position of the moving body 102.

[0053] Next, in step S102, the traveling lane determination unit 2 determines the traveling lane of the moving body 102 corresponding to the position information acquired in step S101, and assigns the information on the determined traveling lane to the position information of the moving body 102.

[0054] Next, in step S103, the lane unit statistical processing unit 3 performs statistical processing of the first position coordinate point sequence for each lane, and generates a second position coordinate point sequence for each lane.

[0055] Next, in step S104, the approximate line calculation unit 4 generates an approximate line approximated by a straight line, an arc, a clothoid curve, or the like for the second position coordinate point sequence for each traveling lane.

[0056] Next, in step S105, the statistical processing reliability calculation unit 5 calculates the reliability of the second position coordinate points for each lane generated in step S103 and assigns it to the second position coordinate points. The reliability of the second position coordinate points is calculated based on the number of first position coordinate points used in the statistical processing by the lane unit statistical processing unit 3 when creating the second position coordinate points, and the accuracy information at the time of acquiring the information of the first position coordinate points.

[0057] Next, in step S106, the approximation line reliability calculation unit 6 calculates the reliability of the approximation line generated in step S104. The reliability of the approximation line is calculated based on the difference of the approximation line to the second position coordinate point sequence from which the approximation line was generated, and the reliability of the second position coordinate point from which the approximation line was generated.

[0058] Next, in step S107, the adjacent lane information generation unit 7 extracts the second position coordinate point of the source of the approximation line, and identifies the adjacent lane and adjacent lane area based on the direction of travel of the moving body 102 on the approximation line and the distance between the second position coordinate point of the source of the approximation line. If altitude information is stored as the position information of the moving body 102 in the moving body sensor information management table 9, the adjacent lane information generation unit 7 may also use the altitude information to identify the adjacent lane and adjacent lane area.

[0059] Next, in step S108, the adjacent lane shape correction unit 8 extracts an adjacent lane group which is a plurality of approximate lines that are adjacent lanes to each other, and generates the road map data 10 by correcting the shape and position of the adjacent lane group. Specifically, the adjacent lane shape correction unit 8 compares the reliability of the adjacent lane group, and sets the shape of one approximate line selected according to the reliability as the shape of all the approximate lines in the adjacent lane group. Further, the adjacent lane shape correction unit 8 arranges each approximate line in the adjacent lane group so that the error from the position coordinate point of the generation source of each approximate line becomes small. Furthermore, the adjacent lane shape correction unit 8 moves at least one approximate line in the adjacent lane group in a direction perpendicular to its orientation so that the lane widths defined by two adjacent approximate lines in the adjacent lane group are equal for each lane, according to the reliability of each approximate line in the adjacent lane group. The road map data 10 generated by the adjacent lane shape correction unit 8 in this way is stored in the storage device 40.

[0060] The road map generation system 1000 may repeatedly perform the processes of steps S101 to S108 at regular time intervals.

[0061] <A-3. Effect> As described above, the road map generation device according to the present embodiment can select the adjacent lane shape according to the reliability and correct it so that the widths of the adjacent lanes are equal. Therefore, even if the sensor information contains inaccurate information or an error occurs during the determination process of the driving lane, it is possible to suppress the occurrence of cases where the shapes of the adjacent lanes are different or the widths of the adjacent lanes are different, and accurate correction can be performed.

[0062] By determining the moving distance according to the reliability calculated from the data amount and the accuracy of the sensor information, it becomes possible to generate accurate road map data.

[0063] In addition, for the generated road map data, by applying approximation processing using arcs, straight lines, and clothoid curves used in road design, a shape with a smoothly changing curvature is adopted, and by assigning a reliability calculated from the data amount and the accuracy of the sensor information to the approximate lines, it becomes possible to make a selection according to the intended use.

[0064] <B. Embodiment 2> <B-1. Configuration> FIG. 9 shows the configuration of the server 101B in the road map generation system 1000 according to Embodiment 2. The server 101B is a road map generation device according to Embodiment 2. In this embodiment, the same components as those in Embodiment 1 are denoted by the same reference numerals, and the descriptions overlapping with Embodiment 1 are omitted as appropriate.

[0065] In addition to the configuration of Embodiment 1, the processor 100 of the server 101B includes a road structure information setting unit 11. The road structure information setting unit 11 sets the lane width conditions for each road type according to the design specifications defined by the road structure order. The lane width conditions include, for example, the conditions of the maximum value and the minimum value of the lane width.

[0066] Also, when correcting the positions of each approximate adjacent line constituting the adjacent lane group, the adjacent lane shape correction unit 8 in this embodiment makes the lane width satisfy the lane width conditions set by the road structure information setting unit 11.

[0067] <B-2. Operation> The flow of the server 101A in Embodiment 1 shown in FIG. 6 is also applicable to the server 101B in Embodiment 2. However, when determining the moving distances of each approximate adjacent line constituting the adjacent lane group in step S108 of FIG. 6, the adjacent lane shape correction unit 8 makes the lane width satisfy the conditions of the maximum value and the minimum value set by the road structure information setting unit 11.

[0068] For example, the adjacent lane shape correction unit 8 creates an evaluation function F2 represented by the following formula (4), and calculates the moving distances m301, m302, m303 of each approximate line so that the evaluation function F2 takes the minimum value. In formula (4), the evaluation function F2 is represented using the difference in lane width |dw300 - dw301| between adjacent lanes, the moving distances m of approximate lines 301, m302, m303, the reliability w101, w102, w103 of the approximate lines, the maximum value dwmax of the lane width, and the minimum value dwmin of the lane width.

[0069]

Number

[0070] <B-3. Effect> As described above, in the road map generation device according to the second embodiment, the adjacent lane shape correction unit 8 corrects the positions of the respective adjacent approximate lines constituting the adjacent lane group so as to satisfy the lane width conditions defined in the road structure ordinance. Therefore, according to the road map generation device described in the second embodiment, in addition to the effects of the first embodiment, the following effects can be obtained. That is, even if the sensor information includes inaccurate information or an error occurs during the determination process of the traveling lane, it is possible to suppress the lane width in the road map data from deviating from the road structure ordinance.

[0071] <C. Third Embodiment> <C-1. Configuration> FIG. 10 shows the configuration of the server 101C in the road map generation system 1000 according to the third embodiment. The server 101C is a road map generation device according to the third embodiment. In the present embodiment, the same components as those in the first embodiment or the second embodiment are denoted by the same reference numerals, and the descriptions overlapping with the first embodiment or the second embodiment are appropriately omitted.

[0072] In addition to the configuration of the first embodiment, the processor 100 of the server 101C includes a connection information setting unit 12. The connection information setting unit 12 sets the connection information of the approximate line based on the connection information of the second position coordinate point sequence that is the source of creation of the approximate line.

[0073] In addition to the operation of the first embodiment, the adjacent lane shape correction unit 8 of the present embodiment acquires the connection information of the approximate line from the connection information setting unit 12, and corrects each approximate line so that the distance between the end points becomes smaller between the approximate lines to be connected.

[0074] <C-2. Operation> The flow of server 101A in Embodiment 1 shown in Figure 6 also applies to server 101C in Embodiment 3. However, in step S108 of Figure 6, when the adjacent lane shape correction unit 8 determines the movement distance of each adjacent approximation line constituting the adjacent lane group, it sets an evaluation function using the difference in lane width in the adjacent lane group, the movement distance of each adjacent approximation line, the reliability of each adjacent approximation line, the distance between the endpoints of the connecting approximation line, the difference in lane width in the adjacent lane group to which the connecting approximation line belongs, the movement distance of the connecting approximation line, and the reliability of the connecting approximation line.

[0075] Figure 11 shows the approximate lines le401-le406 after shape correction by the adjacent lane shape correction unit 8. Approximate lines le404-le406 are connected to approximate lines le401-le403, respectively. The reliability levels of approximate lines le401-le406 are denoted as w401-w406, respectively.

[0076] As shown in Figure 12, the adjacent lane shape correction unit 8 corrects the lane widths so that they are equal by moving the approximation lines le401-le406 parallel to the other approximation lines perpendicular to the inclination of the connection endpoints. The distances moved by the approximation lines le401-le406 at this time are m501-m506, and the approximation lines after the movement are designated as approximation lines le501-le506. The lane widths based on approximation lines le501 and le502 are designated as dw500, the lane widths based on approximation lines le502 and le503 are designated as dw501, the lane widths based on approximation lines le504 and le505 are designated as dw502, and the lane widths based on approximation lines le505 and le506 are designated as dw503. Furthermore, the shortest distance between the endpoints of approximation lines le501 and le504 is defined as dw504, the shortest distance between the endpoints of approximation lines le502 and le505 is defined as dw505, and the shortest distance between the endpoints of approximation lines le503 and le506 is defined as dw506. The adjacent lane shape correction unit 8 creates an evaluation function F3 represented by the following equation (5), and calculates the movement distances m301, m302, and m303 of each approximation line so that the evaluation function F3 takes its minimum value.

[0077]

number

[0078] In the above description, the connection of two approximate lines was considered. When three or more approximate lines are connected, the adjacent lane shape correction unit 8 recursively extracts the approximate lines to be connected, and sets an evaluation function targeting the distances between the endpoints of the approximate lines to be connected and the approximate lines to which all the extracted approximate lines are connected, as well as the errors, moving distances, and reliabilities of the adjacent lane widths of the approximate lines to be connected.

[0079] <C - 3. Effect> As described above, the road map generation device according to the present embodiment includes a connection information setting unit 12 that sets connection information of approximate lines based on the connection information of the second position coordinate point sequence. Then, the adjacent lane shape correction unit 8 corrects the positions of the adjacent lane group based on the connection information of the approximate lines. Therefore, the adjacent lane shape correction unit 8 can smoothly connect the approximate lines. According to the road map generation device of the present embodiment, even if the sensor information includes inaccurate information or an error occurs during the determination process of the driving lane, it is possible to create road map data 10 in which the curvature of the center line of the lane changes smoothly.

[0080] <D. Embodiment 4> <D - 1. Configuration> FIG. 13 shows the configuration of the server 101D in the road map generation system 1000 according to Embodiment 4. The server 101D is a road map generation device according to Embodiment 4. In the present embodiment, the same components as those in Embodiments 1 to 3 are denoted by the same reference numerals, and the descriptions overlapping with those in Embodiments 1 to 3 are omitted as appropriate.

[0081] In addition to the configuration of Embodiment 1, the processor 100 of the server 101D includes a lane shape estimation unit 13 and a lane shape extraction unit 14.

[0082] The lane shape estimation unit 13 generates an image plotting the second position coordinate point sequence for each lane, which is generated by the lane unit statistical processing unit 3. The lane shape estimation unit 13 then estimates the lane shape from the plotted image of the second position coordinate point sequence and creates an estimated lane shape image representing the estimated lane shape. For example, the lane shape estimation unit 13 prepares a lane centerline map created based on the position and shape detected by the camera or sensors of the MMS (Mobile Mapping System) measurement vehicle, and generates an image of the lane centerline map covering the same range as the plotted image of the second position coordinate point sequence. The lane shape estimation unit 13 then uses the plotted image of the second position coordinate point sequence and the image of the lane centerline map covering the same range as training data to generate an estimated lane shape image using machine learning. Here, the lane shape estimation unit 13 may also use an image generation algorithm that utilizes a Generative Adversarial Network (GAN) such as pix2pix.

[0083] Figure 14 shows the sequence of second position coordinate points for each lane generated by the lane unit statistical processing unit 3. Figure 15 shows a plotted image of the sequence of second position coordinate points from Figure 14, generated by the lane shape estimation unit 13. Figure 16 shows the estimated lane shape image generated by the lane shape estimation unit 13 based on the image from Figure 15.

[0084] The lane shape extraction unit 14 generates position coordinate point sequence information for each lane from the estimated lane shape image generated by the lane shape estimation unit 13. For example, the lane shape extraction unit 14 extracts position information on the image of contour points that constitute the outline of the lane or road from the estimated lane shape image, and converts this position information into position coordinate point information consisting of latitude and longitude to generate position coordinate point sequence information for each lane. Here, the generated position coordinate point sequence is referred to as the third position coordinate point sequence to distinguish it from the second position coordinate point sequence generated by the lane unit statistical processing unit 3.

[0085] Figure 17 shows the position coordinate point information generated from the estimated lane shape image in Figure 16. The position coordinate point sequence information in Figure 17 includes multiple third position coordinate points and connection information for these third position coordinate points.

[0086] The approximate line calculation unit 4 generates an approximate line for the third position coordinate point sequence generated by the lane shape extraction unit 14, rather than the second position coordinate point sequence generated by the lane unit statistical processing unit 3.

[0087] <D-2. Operation> FIG. 18 is a flowchart showing the operation of the server 101D according to the fourth embodiment. The flowchart of FIG. 18 is obtained by adding steps S103A and S103B between steps S103 and S104 in the flowchart of FIG. 8 showing the operation of the server 101A according to the first embodiment.

[0088] In step S103, the lane unit statistical processing unit 3 performs statistical processing of the first position coordinate point sequence for each lane to generate a second position coordinate point sequence. After step S103, in step S103A, the lane shape estimation unit 13 generates an estimated lane shape image corresponding to the second position coordinate point sequence. The lane shape estimation unit 13 generates an estimated lane shape image by machine learning with the drawn image of the second position coordinate point sequence and the image of the high-precision map of the same geographical range as inputs.

[0089] After step S103A, in step S103B, the lane shape extraction unit 14 converts the estimated lane shape image into position coordinate point sequence information for each lane.

[0090] After step S103B, in step S104, the approximate line calculation unit 4 generates an approximate line that approximates the third position coordinate point sequence included in the position coordinate point sequence information created in step S103B with a straight line, an arc, a clothoid curve, or the like. <00oo382> <D-3. Effect> The road map generation device according to Embodiment 4 generates a drawn image of a second position coordinate point sequence, and creates an estimated lane shape image representing a lane shape corresponding to the second position coordinate point sequence by machine learning using the drawn image and an image of a high-precision map within the same geographical range as the drawn image as teacher data. The device includes a lane shape extraction unit 14 that generates a third position coordinate point sequence including position coordinates and connection information of contour points that form the contour of the lane in the estimated lane shape image. The approximate line calculation unit 4 calculates an approximate line of the third position coordinate point sequence instead of the second position coordinate point sequence. Therefore, according to the present embodiment, the occurrence of cases where the shape or lane width differs between adjacent lanes is suppressed while incorporating the correction result by machine learning.

[0092] <E. Embodiment 5> <E-1. Configuration> FIG. 19 shows the configuration of the server 101E in the road map generation system 1000 according to Embodiment 5. The server 101E is a road map generation device according to Embodiment 5. In the present embodiment, the same components as those in Embodiments 1 to 4 are denoted by the same reference numerals, and the descriptions overlapping with Embodiments 1 to 4 are appropriately omitted.

[0093] In addition to the configuration of Embodiment 1, the processor 100 of the server 101E includes an image shape comparison unit 15.

[0094] The image shape comparison unit 15 acquires a satellite image or an aerial image that includes a measurement date and time range of the position information acquired by the position information extraction unit 1 from the mobile sensor information management table 9 and has the same or approximately the same geographical range photographed at the same time or around the same time. The satellite image or aerial image acquired here is referred to as a reference image.

[0095] Then, the image shape comparison unit 15 extracts position information of contour points that form the contour of the road or lane on the reference image, and converts this position information into a position coordinate point sequence composed of latitude and longitude and connection information of each position coordinate point. The position coordinate point sequence created by the image shape comparison unit 15 here is referred to as a fourth position coordinate point sequence, and the position coordinate points that make up the fourth position coordinate point sequence are referred to as fourth position coordinates.

[0096] In addition, the image shape comparison unit 15 compares the second position coordinate point sequence calculated by the lane unit statistical processing unit 3 with the fourth position coordinate point sequence that is the closest in distance. When the curvature and position error between the two are larger than a predetermined threshold value, the approximate line reliability calculation unit 6 subtracts a fixed value from the reliability of the second position coordinate point calculated by the statistical processing reliability calculation unit 5. That is, the approximate line reliability calculation unit 6 calculates the reliability of the approximate line based on the difference between the approximate line of the second position coordinate point of the source and the reliability of the second position coordinate point of the source, in addition to the similarity between the second position coordinate point sequence of the source and the fourth position coordinate point sequence.

[0097] <E-2. Effect> As described above, the road map generation device according to the present embodiment acquires a reference image that is a satellite image or an aerial image obtained by photographing a moving body at the same time as the acquisition date and time of the first position coordinate point sequence by the moving body sensor, and generates a fourth position coordinate point sequence representing the position coordinates of the contour points that constitute the contour of the lane in the reference image. Based on the comparison between the second position coordinate point sequence and the fourth position coordinate point sequence, it includes an image shape comparison unit 15 that changes the reliability of the second position coordinate point. Therefore, the road map generation device according to the present embodiment can accurately correct the road map data of the adjacent lane group based on the reference image that is the actual photographed image.

[0098] <F. Embodiment 6> <F-1. Configuration> FIG. 20 shows the configuration of the moving body 102F in the road map generation system according to Embodiment 6. In Embodiments 1-5, as shown in FIG. 1, the road map generation system is configured by the server 101 and the moving body 102. In contrast, in this embodiment, the road map generation system is configured only by the moving body 102F. The moving body in this embodiment is denoted as the moving body 102F to distinguish it from the moving body 102 in Embodiments 1-5. In this embodiment, the same components as those in Embodiments 1 to 5 are denoted by the same reference numerals, and descriptions overlapping with those in Embodiments 1 to 5 are omitted as appropriate.

[0099] As shown in FIG. 20, the moving body 102F includes a storage device 40, a moving body sensor interface 50, a moving body sensor 60, and a processor 100. The configurations of the storage device 40 and the processor 100 in the moving body 102F are the same as those of the storage device 40 and the processor 100 in the server 101A of Embodiment 1. Therefore, in Embodiment 6, the road map generation device is mounted on the moving body 102F. The storage device 40 receives GNSS sensor information and camera sensor information from the moving body sensor 60 via the moving body sensor interface 50, and stores them in the moving body sensor information management table 9.

[0100] <F-2. Effect> The road map generation device according to Embodiment 6 is mounted on the moving body 102F. Therefore, the road map generation device according to Embodiment 6 can generate a road map by utilizing the moving body sensor information acquired by the running of the moving body 102F. As the moving body 102F accumulates the number of running times, an accurate map is generated and can be utilized as a map.

[0101] Although the preferred embodiments and the like have been described in detail above, the present invention is not limited to the above embodiments and the like, and various modifications and substitutions can be made to the above embodiments and the like without departing from the scope described in the claims.

[0102] Hereinafter, aspects of the present disclosure will be summarized and described as appendices.

[0103] (Appendix 1) A position information extraction unit that extracts position information representing the position of the moving body measured by a moving body sensor mounted on the moving body as a first position coordinate point sequence composed of a plurality of first position coordinate points; A traveling lane determination unit that determines the lane in which the moving body was traveling at each of the first position coordinate points based on a captured image of a camera mounted on the moving body and assigns it to the first position coordinate point; A lane unit statistical processing unit that performs statistical processing on the first position coordinate point sequence for each lane to generate a second position coordinate point sequence composed of a plurality of second position coordinate points; An approximation line calculation unit that calculates an approximation line for the second sequence of positional coordinate points for each lane, A statistical processing reliability calculation unit calculates the reliability of each of the second position coordinate points based on the number of source first position coordinate points and the accuracy of the source first position coordinate points when acquired by the mobile sensor. An approximation line reliability calculation unit calculates the reliability of the approximation line based on the difference of the approximation line to the second position coordinate point sequence from which the approximation line is generated, and the reliability of each of the second position coordinate points that make up the second position coordinate point sequence from which the approximation line is created. An adjacent lane information generation unit identifies three or more of the aforementioned approximate lines as adjacent lane groups based on the distance between the second position coordinate point sequences that are the source of each approximate line, and the direction of travel of the moving body in each approximate line, The adjacent lane shape correction unit corrects the adjacent lane group by unifying the shape of each adjacent approximation line, which is an approximation line constituting the adjacent lane group, based on the reliability of each adjacent approximation line, and by moving the position of each adjacent approximation line in a direction perpendicular to the orientation of each adjacent approximation line by a distance determined based on the lane width defined by the distance between two adjacent adjacent approximation lines and the reliability of each adjacent approximation line, and stores the corrected adjacent lane group data in a storage device. Road map generation device.

[0104] (Note 2) The aforementioned approximation line is composed of a straight line, a circular arc, or a clothoid curve. The road map generation device described in Appendix 1.

[0105] (Note 3) The position information includes information about the height of the moving body in the first position coordinate sequence. The adjacent lane information generation unit identifies the adjacent lane group based on the height information of the moving body in the first position coordinate point sequence, which is the source of the second position coordinate point sequence from which each of the approximate lines is created. A road map generating device as described in Appendix 1 or Appendix 2.

[0106] (Note 4) The system further includes a road structure information setting unit that sets lane width conditions for each road type, relating to the maximum and minimum values ​​of the lane width, in accordance with the design specifications stipulated in the Road Structure Ordinance. The adjacent lane shape correction unit corrects the group of adjacent lanes so that the lane width satisfies the lane width condition. A road map generating device as described in any one of the items from Appendix 1 to Appendix 3.

[0107] (Note 5) The system further includes a connection information setting unit that sets connection information for the approximation line based on the connection information for the second sequence of position coordinate points, The adjacent lane shape correction unit corrects the position of the adjacent lane group based on the connection information of the approximation line. A road map generating device as described in any one of the items from Appendix 1 to Appendix 4.

[0108] (Note 6) A lane shape estimation unit generates a drawing image of the second sequence of positional coordinates, and uses machine learning with the drawing image and a high-precision map image of the same geographical area as the drawing image as training data to create an estimated lane shape image representing the lane shape corresponding to the second sequence of positional coordinates; The system includes a lane shape extraction unit that generates a third sequence of position coordinate points, including the position coordinates and connection information of contour points that constitute the outline of the lane in the estimated lane shape image, The approximation line calculation unit calculates an approximation line for the third position coordinate point sequence instead of the second position coordinate point sequence. A road map generating device as described in any one of the items from Appendix 1 to Appendix 5.

[0109] (Note 7) The system further includes an image shape comparison unit that acquires a reference image, which is a satellite image or aerial image taken at the same time as the acquisition date and time of the first position coordinate sequence by the mobile sensor and includes the same geographical area as the first position coordinate sequence, generates a fourth position coordinate sequence representing the position coordinates of contour points that constitute the outline of the lane in the reference image, and modifies the reliability of the second position coordinate points based on a comparison between the second position coordinate sequence and the fourth position coordinate sequence. A road map generating device as described in any one of the items from Appendix 1 to Appendix 6.

[0110] (Note 8) Mounted on the aforementioned mobile body, A road map generating device as described in any one of the items from Appendix 1 to Appendix 7.

[0111] (Note 9) The position information extraction unit extracts position information that represents the position of the mobile body measured by the mobile body sensor mounted on the mobile body as a sequence of first position coordinate points consisting of multiple first position coordinate points. The lane determination unit determines the lane the moving body was traveling in at each of the first position coordinate points based on the image captured by the camera mounted on the moving body, and assigns it to the first position coordinate point. The lane-unit statistical processing unit performs statistical processing on the first position coordinate point sequence for each lane to generate a second position coordinate point sequence consisting of a plurality of second position coordinate points. The approximation line calculation unit calculates an approximation line for the second sequence of position coordinate points for each lane. The statistical processing confidence calculation unit calculates the confidence level of each of the second position coordinate points based on the number of the source first position coordinate points and the accuracy of the source first position coordinate points when acquired by the mobile sensor. The approximation line reliability calculation unit calculates the reliability of the approximation line based on the difference of the approximation line to the second position coordinate point sequence from which the approximation line was generated, and the reliability of each of the second position coordinate points that make up the second position coordinate point sequence from which the approximation line was created. The adjacent lane information generation unit identifies three or more of the adjacent lane groups based on the distance between the second position coordinate point sequences that are the source of each of the approximate lines, and the direction of travel of the moving body in each of the approximate lines. The adjacent lane shape correction unit unifies the shape of each adjacent approximation line, which constitutes the group of adjacent lanes, based on the reliability of each adjacent approximation line. The position of each adjacent approximation line is moved in a direction perpendicular to the orientation of each adjacent approximation line by a distance determined based on the lane width defined by the distance between two adjacent approximation lines and the reliability of each adjacent approximation line. The group of adjacent lanes is then corrected, and the data of the corrected group of adjacent lanes is stored in a storage device as road map data. Road map generation method.

[0112] (Note 10) Computers, A position information extraction unit extracts position information that represents the position of a mobile object measured by a mobile object sensor mounted on the mobile object as a sequence of first position coordinate points consisting of multiple first position coordinate points, A lane determination unit determines the lane the moving body was traveling in at each of the first position coordinate points based on the image captured by the camera mounted on the moving body, and assigns it to the first position coordinate point. A lane-unit statistical processing unit performs statistical processing on the first position coordinate point sequence for each lane to generate a second position coordinate point sequence consisting of a plurality of second position coordinate points, An approximation line calculation unit that calculates an approximation line for the second sequence of positional coordinate points for each lane, A statistical processing reliability calculation unit calculates the reliability of each of the second position coordinate points based on the number of source first position coordinate points and the accuracy of the source first position coordinate points when acquired by the mobile sensor. An approximation line reliability calculation unit calculates the reliability of the approximation line based on the difference of the approximation line to the second position coordinate point sequence from which the approximation line is generated, and the reliability of each of the second position coordinate points that make up the second position coordinate point sequence from which the approximation line is created. An adjacent lane information generation unit identifies the group of adjacent lanes to which the approximation line belongs, based on the distance between the second position coordinate point sequences that are the source of each approximation line, and the direction of travel of the moving body in each approximation line. To operate as an adjacent lane shape correction unit, it unifies the shape of each adjacent approximation line, which is an approximation line constituting the group of adjacent lanes, based on the reliability of each adjacent approximation line, corrects the group of adjacent lanes by moving the position of each adjacent approximation line in a direction perpendicular to the orientation of each adjacent approximation line by a distance determined based on the lane width defined by the distance between two adjacent approximation lines and the reliability of each adjacent approximation line, and stores the corrected data of the group of adjacent lanes as road map data in a storage device, Road map generation program. [Explanation of symbols]

[0113] 1 Location information extraction unit, 2 Driving lane determination unit, 3 Lane unit statistical processing unit, 4 Approximate line calculation unit, 5 Statistical processing reliability calculation unit, 6 Approximate line reliability calculation unit, 7 Adjacent lane information generation unit, 8 Adjacent lane shape correction unit, 9 Mobile sensor information management table, 10 Road map data, 11 Road structure information setting unit, 12 Connection information setting unit, 13 Lane shape estimation unit, 14 Lane shape extraction unit, 15 Image shape comparison unit, 20 Communication interface, 30 Communicator, 40 Storage device, 50 Mobile sensor interface, 60 Mobile sensor, 100 Processor, 101 Server, 102 Mobile object, 1000 Road map generation system.

Claims

1. A position information extraction unit extracts position information that represents the position of the mobile body measured by a mobile body sensor mounted on the mobile body as a sequence of first position coordinate points consisting of a plurality of first position coordinate points, A lane determination unit that determines the lane the moving body was traveling in at each of the first position coordinate points based on the image captured by the camera mounted on the moving body and assigns it to the first position coordinate point, A lane-unit statistical processing unit performs statistical processing on the first position coordinate point sequence for each lane to generate a second position coordinate point sequence consisting of a plurality of second position coordinate points, An approximation line calculation unit that calculates an approximation line for the second sequence of positional coordinate points for each lane, A statistical processing reliability calculation unit calculates the reliability of each of the second position coordinate points based on the number of the first position coordinate points that were generated and the accuracy of the first position coordinate points that were generated when acquired by the mobile sensor. An approximation line reliability calculation unit calculates the reliability of the approximation line based on the difference of the approximation line to the second position coordinate point sequence from which the approximation line is generated, and the reliability of each of the second position coordinate points that make up the second position coordinate point sequence from which the approximation line is created. An adjacent lane information generation unit identifies the group of adjacent lanes to which the approximation line belongs, based on the distance between the second position coordinate point sequences that are the source of each approximation line, and the direction of travel of the moving body in each approximation line, The adjacent lane shape correction unit corrects the adjacent lane group by unifying the shape of each adjacent approximation line, which is an approximation line constituting the adjacent lane group, based on the reliability of each adjacent approximation line, and by moving the position of each adjacent approximation line in a direction perpendicular to the orientation of each adjacent approximation line by a distance determined based on the lane width defined by the distance between two adjacent adjacent approximation lines and the reliability of each adjacent approximation line, and stores the corrected adjacent lane group data in a storage device as road map data. Road map generation device.

2. The aforementioned approximation line is composed of a straight line, a circular arc, or a clothoid curve. The road map generation device according to claim 1.

3. The position information includes information about the height of the moving body in the first position coordinate sequence. The adjacent lane information generation unit identifies the adjacent lane group based on the height information of the moving body in the first position coordinate point sequence, which is the source of the second position coordinate point sequence from which each of the approximate lines is created. The road map generation device according to claim 1.

4. The system further includes a road structure information setting unit that sets lane width conditions for each road type, relating to the maximum and minimum values ​​of the lane width, in accordance with the design specifications stipulated in the Road Structure Ordinance. The adjacent lane shape correction unit corrects the group of adjacent lanes so that the lane width satisfies the lane width condition. The road map generation device according to claim 1.

5. The system further includes a connection information setting unit that sets connection information for the approximation line based on the connection information for the second sequence of position coordinate points, The adjacent lane shape correction unit corrects the position of the adjacent lane group based on the connection information of the approximation line. The road map generation device according to claim 1.

6. A lane shape estimation unit generates a drawing image of the second sequence of positional coordinates, and uses machine learning with the drawing image and a high-precision map image of the same geographical area as the drawing image as training data to create an estimated lane shape image representing the lane shape corresponding to the second sequence of positional coordinates; The system includes a lane shape extraction unit that generates a third sequence of position coordinate points, including the position coordinates and connection information of contour points that constitute the outline of the lane in the estimated lane shape image, The approximation line calculation unit calculates an approximation line for the third position coordinate point sequence instead of the second position coordinate point sequence. The road map generation device according to claim 1.

7. The system further includes an image shape comparison unit that acquires a reference image, which is a satellite image or aerial image taken at the same time as the acquisition of the first position coordinate sequence by the mobile sensor, and which includes the same geographical area as the first position coordinate sequence; generates a fourth position coordinate sequence representing the position coordinates of contour points that constitute the outline of the lane in the reference image; and modifies the reliability of the second position coordinate points based on a comparison between the second position coordinate sequence and the fourth position coordinate sequence. The road map generation device according to claim 1.

8. Mounted on the aforementioned mobile body, The road map generation device according to claim 1.

9. A road map generation method using a road map generation device comprising a location information extraction unit, a driving lane determination unit, a lane unit statistical processing unit, an approximation line calculation unit, a statistical processing reliability calculation unit, an approximation line reliability calculation unit, an adjacent lane information generation unit, and an adjacent lane shape correction unit, The position information extraction unit extracts position information that represents the position of the mobile body measured by a mobile body sensor mounted on the mobile body as a sequence of first position coordinate points consisting of a plurality of first position coordinate points. The lane determination unit determines the lane the moving body was traveling in at each of the first position coordinate points based on the image captured by the camera mounted on the moving body, and assigns it to the first position coordinate point. The lane-unit statistical processing unit performs statistical processing on the first position coordinate point sequence for each lane to generate a second position coordinate point sequence consisting of a plurality of second position coordinate points. The approximation line calculation unit calculates an approximation line for the second sequence of position coordinate points for each lane. The statistical processing reliability calculation unit calculates the reliability of each of the second position coordinate points based on the number of the first position coordinate points that were generated and the accuracy of the first position coordinate points that were generated when acquired by the moving sensor. The approximation line reliability calculation unit calculates the reliability of the approximation line based on the difference between the approximation line and the second position coordinate point sequence from which the approximation line was generated, and the reliability of each of the second position coordinate points that make up the second position coordinate point sequence from which the approximation line was created. The adjacent lane information generation unit identifies three or more of the adjacent lane groups based on the distance between the second position coordinate point sequences that are the source of each of the approximate lines, and the direction of travel of the moving body in each of the approximate lines. The adjacent lane shape correction unit unifies the shape of each adjacent approximation line, which constitutes the group of adjacent lanes, based on the reliability of each adjacent approximation line, and corrects the group of adjacent lanes by moving the position of each adjacent approximation line in a direction perpendicular to the direction of each adjacent approximation line by a distance determined based on the lane width defined by the distance between two adjacent approximation lines and the reliability of each adjacent approximation line, and stores the corrected data of the group of adjacent lanes as road map data in a storage device. Road map generation method.

10. Computers, A position information extraction unit extracts position information that represents the position of a mobile object measured by a mobile object sensor mounted on the mobile object as a sequence of first position coordinate points consisting of multiple first position coordinate points, A lane determination unit that determines the lane the moving body was traveling in at each of the first position coordinate points based on the image captured by the camera mounted on the moving body and assigns it to the first position coordinate point, A lane-unit statistical processing unit performs statistical processing on the first position coordinate point sequence for each lane to generate a second position coordinate point sequence consisting of a plurality of second position coordinate points, An approximation line calculation unit that calculates an approximation line for the second sequence of positional coordinate points for each lane, A statistical processing reliability calculation unit calculates the reliability of each of the second position coordinate points based on the number of the first position coordinate points that were generated and the accuracy of the first position coordinate points that were generated when acquired by the mobile sensor. An approximation line reliability calculation unit calculates the reliability of the approximation line based on the difference of the approximation line to the second position coordinate point sequence from which the approximation line is generated, and the reliability of each of the second position coordinate points that make up the second position coordinate point sequence from which the approximation line is created. An adjacent lane information generation unit identifies the group of adjacent lanes to which the approximation line belongs, based on the distance between the second position coordinate point sequences that are the source of each approximation line, and the direction of travel of the moving body in each approximation line, To operate as an adjacent lane shape correction unit, it unifies the shape of each adjacent approximation line, which constitutes the group of adjacent lanes, based on the reliability of each adjacent approximation line, corrects the group of adjacent lanes by moving the position of each adjacent approximation line in a direction perpendicular to the direction of each adjacent approximation line by a distance determined based on the lane width defined by the distance between two adjacent approximation lines and the reliability of each adjacent approximation line, and stores the corrected data of the group of adjacent lanes as road map data in a storage device, Road map generation program.

Images