Self-position estimation device

Abstract

To provide a self-position estimation device that can improve the accuracy of self-position estimation for moving objects. [Solution] The in-vehicle device 2 is represented by dividing the markings drawn on the road surface into multiple cells C, and the control unit 21 acquires map data in which the position 132 of each of the multiple cells C is set, and acquires point cloud information obtained from the rider 3a as the vehicle moves. The control unit 21 then recognizes point cloud information corresponding to the markings drawn on the road surface from the acquired point cloud information, and estimates the vehicle's own position by matching the recognized point cloud information corresponding to the markings drawn on the road surface with the map data.

Description

Technical Field

[0001] The present invention relates to a self-position estimation device that estimates the self-position of a moving object.

Background Art

[0002] In a moving object such as an autonomous vehicle, it is necessary to match the position of a ground object measured by a sensor such as a LiDAR (Light Detection and Ranging) with the position of the ground object in the map information for autonomous driving to accurately estimate the current position. Patent Document 1 describes an example of a method for estimating the current position using the position of a ground object as a landmark detected by a LiDAR and the ground object in the map information.

[0003] Further, Patent Document 2 describes detecting a white line using a LiDAR and accurately detecting the relative position of the white line in the lateral direction with respect to the vehicle or the direction in which the vehicle is facing the white line.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] By the method described in Patent Document 2, it is possible to estimate the self-position of a vehicle or the like using a white line (lane line). However, for example, when there are markings other than white lines on the road surface such as a traffic lane division indicating a right turn or a left turn, if the marking is erroneously detected as a white line, the accuracy of self-position estimation will decrease.

[0006] One example of a problem that this invention aims to solve is improving the accuracy of self-position estimation for moving objects. [Means for solving the problem]

[0007] To solve the above problems, the invention described in claim 1 is characterized in that the markings drawn on the road surface are divided into a plurality of cells of a certain size and each of the plurality of cells is represented by map data acquisition unit which acquires map data in which information about the position of the cell is set; sensor information acquisition unit which acquires sensor information obtained from a sensor mounted on a moving body; recognition unit which recognizes the sensor information corresponding to the markings from the sensor information acquired by the sensor information acquisition unit; and estimation unit which estimates the self position of the moving body by matching the sensor information corresponding to the markings recognized by the recognition unit with the map data, and each of the cells has attribute information including information about the reflectance intensity of the point cloud corresponding to the cell.

[0008] The invention described in claim 2 is a self-position estimation method performed by a self-position estimation device for estimating the self-position of a moving body, comprising: a map data acquisition step of acquiring map data in which a mark drawn on a road surface is divided into a plurality of cells of a certain size, and each of the plurality of cells has information regarding the position of the cell set; a sensor information acquisition step of acquiring sensor information obtained from a sensor mounted on the moving body; a recognition step of recognizing the sensor information corresponding to the mark from the sensor information acquired in the sensor information acquisition step; and an estimation step of estimating the self-position of the moving body by matching the sensor information corresponding to the mark recognized in the recognition step with the map data, wherein each cell has attribute information including information regarding the reflectance intensity of the point cloud corresponding to the cell.

[0009] The invention described in claim 3 is characterized in that the self-localization method described in claim 2 is performed by a computer.

[0010] The invention described in claim 4 is characterized by storing the self-localization program described in claim 3. [Brief explanation of the drawing]

[0011] [Figure 1] This is a schematic diagram of a system consisting of a server device and an in-vehicle device according to a first embodiment of the present invention. [Figure 2] Figure 1 is a functional configuration diagram of the server device shown. [Figure 3] Figure 1 is a functional configuration diagram of the in-vehicle device. [Figure 4] Figure 2 is a flowchart illustrating the map data generation method in the server device shown. [Figure 5] This is an example of a binarized image. [Figure 6] This is the image obtained by removing unnecessary pixels from the binarized image shown in Figure 5. [Figure 7] This is an explanatory diagram of the cell in this embodiment. [Figure 8] This is an explanatory diagram of a map data structure according to one embodiment of the present invention. [Figure 9] This is a modified version of the flowchart shown in Figure 4. [Figure 10] This is an explanatory diagram showing the case where the white lines are represented as a sequence of points. [Figure 11] This is a flowchart of a self-localization method according to one embodiment of the present invention. [Figure 12] A flowchart of a map data generation method according to a second embodiment of the present invention is shown. [Figure 13] This is a modified version of the flowchart shown in Figure 12. [Modes for carrying out the invention]

[0012] Hereinafter, a self-position estimation device according to an embodiment of the present invention will be described. In the self-position estimation device according to an embodiment of the present invention, a marking drawn on a road surface is represented by being divided into a plurality of cells, and for each of the plurality of cells, map data in which information regarding the position of the cell is set is acquired by a map data acquisition unit, and a sensor information acquisition unit acquires sensor information obtained from a sensor by vehicle travel. Then, a recognition unit recognizes sensor information corresponding to a marking drawn on the road surface from the sensor information acquired by the sensor information acquisition unit, and an estimation unit estimates the self-position of the vehicle by performing matching between the sensor information corresponding to the marking drawn on the road surface recognized by the recognition unit and the map data. By doing so, self-position estimation can be performed using the marking drawn on the road surface, and the accuracy of self-position estimation of the moving object can be improved.

[0013] Also, each cell may have predetermined attribute information about the cell. By doing so, for example, if the dispersion of the point cloud and the reflection intensity in the cell are used as the attribute information, it becomes possible to grasp the distribution and ratio of the paint indicating the marking in the cell, that is, to identify the cell including the contour of the marking.

[0014] Also, the marking drawn on the road surface targeted by the present self-position estimation device may be a marking other than the lane line. By doing so, for example, when the lane line such as a white line has already been digitized by another method, it becomes possible to include in the map data road markings such as traffic lanes for different travel directions indicating right turns and left turns and the maximum speed, excluding the lane line such as a white line.

[0015] In addition, the self-position estimation method according to an embodiment of the present invention is such that a sign drawn on a road surface is represented by being divided into a plurality of cells, and map data in which information regarding the position of each cell is set is acquired in a map data acquisition step, and sensor information obtained from a sensor by vehicle travel is acquired in a sensor information acquisition step. Then, in a recognition step, sensor information corresponding to a sign drawn on the road surface is recognized from the sensor information acquired in the sensor information acquisition step, and the self-position of the vehicle is estimated by performing matching between the sensor information corresponding to the sign drawn on the road surface recognized in the recognition step and the map data. By doing so, a sign drawn on the road surface can be included in the map data, and the accuracy of self-position estimation of the moving object can be improved.

[0016] In addition, the self-position estimation program according to an embodiment of the present invention causes a computer to execute the above-described self-position estimation method. By doing so, a sign drawn on the road surface can be included in the map data by using a computer, and the accuracy of self-position estimation of the moving object can be improved.

[0017] Further, the above-described self-position estimation program may be stored in a computer-readable recording medium. By doing so, the program can be distributed not only by being incorporated into a device but also alone, and version updates and the like can be easily performed.

Example

[0018] The map data generation device and map data structure according to the first embodiment of the present invention will be described with reference to FIGS. The map data generation device according to this embodiment is configured as a server device 1 in FIG. The server device 1 acquires information collected by an in-vehicle device 2 mounted on a vehicle V as a moving object via a network N such as the Internet, and generates map data having a map data structure described later based on the acquired information.

[0019] Figure 2 shows the functional configuration of server device 1. As shown in Figure 2, server device 1 comprises a control unit 11, a communication unit 12, and a storage unit 13. The control unit 11 functions as the CPU (Central Processing Unit) of server device 1 and is responsible for the overall control of server device 1. The control unit 11 generates map data based on, for example, point cloud information transmitted from the in-vehicle device 2.

[0020] The communication unit 12 functions as the network interface of the server device 1 and receives, for example, point cloud information transmitted by the in-vehicle device 2. The communication unit 12 also distributes (transmits) the map data generated by the control unit 11 to the in-vehicle device 2 of the vehicle V. In other words, the server device 1 functions as a storage device where the map data is stored.

[0021] The storage unit 13 functions as a storage medium such as a hard disk of the server device 1 and stores map data and the like generated by the control unit 11. The storage medium may also be an optical disc or a memory card.

[0022] Figure 3 shows the functional configuration of the in-vehicle device 2. As shown in Figure 3, the in-vehicle device 2 comprises a control unit 21, a communication unit 22, and a storage unit 23. In addition, as shown in Figure 3, the in-vehicle device 2 is connected to sensors 3 mounted on the vehicle V, including a lidar 3a and a GPS (Global Positioning System) receiver 3b.

[0023] The control unit 21 is composed of a microcomputer, for example, having a CPU and memory, and is responsible for the overall control of the in-vehicle device 2. The control unit 21 transmits point cloud information detected by the sensor 3 from the communication unit 12 to the server device 1. The control unit 21 also estimates its own position based on information indicating features such as markings drawn on the road surface recognized using the lidar 3a, and information such as markings drawn on the road surface included in map data having a map data structure described later, generated by the server device 1. The control unit 21 may also function as the control unit of a guidance device that guides (directs) the vehicle V along a pre-searched route based on its estimated own position. This guidance device includes at least one of a so-called automatic driving function that makes the vehicle V drive autonomously along the route, and a navigation function that presents guidance information along the route to the driver, etc.

[0024] The communication unit 22 transmits point cloud information and other data acquired by the control unit 21 to the server device 1. The communication unit 22 also receives map data and other data distributed from the server device 1.

[0025] The memory unit 23 stores programs and various data that operate in the control unit 21. The memory unit 23 also stores map data distributed from the server device 1.

[0026] Sensor 3 comprises a rider 3a and a GPS receiver 3b. In addition to these sensors, Sensor 3 may also include other sensors, such as a speed sensor. Furthermore, an on-board camera that captures images of the surroundings, such as the front of the vehicle, may be included in Sensor 3.

[0027] The lidar 3a discretely measures the distance to objects in the external environment by emitting laser light as an electromagnetic wave. It outputs a pulsed laser while changing the output direction within a predetermined detection area, and generates point cloud information by receiving the reflected waves of that laser. The lidar 3a outputs multiple pulsed lasers within the detection area and generates point cloud information based on the reflected waves of these multiple pulsed lasers. Each piece of information constituting the point cloud information indicates the output direction of the laser, the distance to the object that reflected the laser, and the intensity of the reflected wave (reflection intensity). In this embodiment, the lidar 3a mainly irradiates the road surface with the laser, and the road surface is used as the detection area. Therefore, the point cloud information includes information indicating the distance to the road surface as the target object and information on the reflection intensity. Of course, the laser may be emitted to areas other than the road surface to acquire surrounding information other than road surface information.

[0028] The GPS receiver 3b periodically receives radio waves transmitted from multiple GPS satellites, as is well known, to detect the current location and time.

[0029] Next, the method for generating map data in the server device 1 (map data generation device) with the above configuration will be explained with reference to Figures 4 to 7. Figure 4 is a flowchart executed by the server device 1. This flowchart is executed by the control unit 11 of the server device 1. In other words, it is configured as a computer program (map data generation program) executed by the CPU of the control unit 11. Note that this program is not limited to being stored in the storage unit 13, but may also be stored on an optical disc, memory card, etc.

[0030] First, in step S11, the control unit 11 acquires point cloud information from the in-vehicle device 2. In this step, the distance to each point in the point cloud information collected from the in-vehicle device 2 is converted into absolute coordinates based on the current position detected by the GPS receiver 3b. Alternatively, the data may be converted into absolute coordinates in the in-vehicle device 2 beforehand. In other words, the control unit 11 functions as an acquisition unit that acquires point cloud information (road surface information) from the lidar 3a (a predetermined sensor).

[0031] Next, in step S12, the control unit 11 generates a binarized image based on the point cloud information acquired in step S11. The binarized image in this embodiment is an image obtained by converting a point cloud with three-dimensional coordinates into a two-dimensional image and binarizing it based on a predetermined threshold value according to the reflection intensity of each point. In the binarized image thus generated, points with high reflection intensity are represented as pixels with high brightness. The pixels included in this binarized image also inherit the coordinate information of the point cloud before conversion.

[0032] In other words, the control unit 11 functions as a marking information generation unit that generates a binarized image (information showing markings drawn on the road surface) based on the point cloud information (road surface information) acquired by the acquisition unit.

[0033] Next, in step S13, the control unit 11 removes pixels with high brightness outside the road surface from the binarized image generated in step S12. An example of removal will be explained with reference to Figures 5 and 6. Figure 5 is a binarized image generated based on point cloud information of a certain road. A white line (lane marking) W, a traffic division R indicating a right turn (hereinafter referred to as the right-turn arrow R), and a traffic division S indicating going straight (hereinafter referred to as the straight-ahead arrow S) are drawn on this binarized image. In addition, this binarized image may also contain unnecessary pixels U outside the road surface (pixels with high brightness outside the road surface).

[0034] Then, in such a binarized image, unnecessary pixels U other than road markings such as white lines W and lane markings for each direction of travel are removed. One method of removal is to separately obtain map data of the area from which point cloud information was acquired, and compare the binarized image with the passable area (road surface area) set in that map data to identify pixels with high brightness outside the road surface. Alternatively, road and lane width information may be obtained from the road and lane network data included in the map data to identify pixels with high brightness outside the road surface.

[0035] Figure 6 shows the binarized image after removing unnecessary pixels U. Through the process described above, unnecessary pixels U are removed, leaving only the markings painted on the road surface, such as the white lines W, the right-turn arrow R, and the straight-ahead arrow S.

[0036] Next, in step S14, the control unit 11 groups the pixels corresponding to the markings drawn on the road surface into cells of a fixed size in the binarized image processed in step S13 (generates cells). The generated cells are assigned coordinate information based on the pixels contained within them. In other words, the markings are divided into multiple cells based on the binarized image (information indicating the markings) generated by the marking information generation unit.

[0037] Here, the cells of this embodiment will be explained with reference to Figure 7. Figure 7 shows a road surface with a white line W and a right-turn arrow R drawn on it. In Figure 7, both the white line W and the right-turn arrow R are represented as pixels with high brightness within the dashed lines in step S13. In other words, the area within the dashed lines is the region drawn on the road surface as the white line W and the right-turn arrow R.

[0038] As shown in Figure 7, by grouping pixels corresponding to markings drawn on the road surface into cells C of a fixed size, white lines W and right-turn arrows R are divided into multiple cells C. The size of these cells C can be appropriately determined so as to be large enough to divide the markings drawn on the road surface into multiple parts. Furthermore, the shape of cells C is not limited to squares; it can also be rectangles or other rectangular shapes, or polygons such as triangles or hexagons.

[0039] Furthermore, in the example in Figure 7, only the parts that include at least some of the markings painted on the road surface are grouped into cells, but the entire road surface could also be divided into multiple cells in a grid (mesh) structure. However, dividing only the markings painted on the road surface into cells can reduce the amount of data in the map data. Also, if the entire road surface is divided into multiple cells, it is advisable to add information such as a flag to cells that contain markings painted on the road surface to indicate this.

[0040] Returning to the explanation of Figure 4, in step S15, the control unit 11 refers to the point cloud obtained in step S11 from the coordinates of cell C generated in step S14 in order to assign statistical information, which will be described later, and associates cell C with the point cloud. That is, each cell C is assigned coordinate information as position information (information about position).

[0041] Then, in step S16, the control unit 11 assigns statistical information as attribute information to each cell C. Examples of statistical information include the average, variance, or histogram of the reflectance intensity of the point cloud corresponding to the cell C. Map data is generated by including the markings drawn on the road surface, which consist of multiple cells C to which statistical information has been assigned, into the map data. In other words, the control unit 11 functions as a map data generation unit that generates map data including information about the position of each of the multiple cells C based on point cloud information (road surface information).

[0042] As is clear from the above explanation, step S11 functions as an acquisition step, step S12 as a display information generation step, and steps S14 to S16 as map data generation steps.

[0043] Furthermore, if an on-board camera is mounted on the vehicle V as part of sensor 3, a colored point cloud may be created by adding a color based on the image captured by the camera to each point, and the mean or variance of the color may be used instead of the reflectance intensity. In addition, the mean or covariance of the coordinate values ​​of the point cloud corresponding to the cell C can also be added as statistical information. In this case, it is necessary to perform a filtering process by reflectance intensity or color to extract the coordinate values ​​of the point cloud containing only the markings drawn on the road surface.

[0044] Figure 8 shows an example of the map data structure of map data generated by the flowchart in Figure 4. As shown in Figure 8, the map data structure includes cell ID 131, location 132, and statistical information 133. The cell ID is an ID that identifies the cell. Location 132 is information indicating the location of the cell. This location 132 is not limited to setting an absolute position (latitude, longitude) for each cell C, but may also be set as a relative position from a reference cell C. For example, for the right-turn arrow R in Figure 7, the cell C in the left corner may be used as the reference cell to set the absolute position, and the other cells C may be set as relative positions from the reference cell. Statistical information 133 contains information such as the reflectance intensity and the mean and variance of the coordinate values ​​mentioned above. In addition to the items shown in Figure 8, information indicating the content of the marking (white line, right-turn arrow, maximum speed, etc.) may also be included.

[0045] In other words, the map data structure shown in Figure 8 represents markings drawn on the road surface, divided into multiple cells C, and each of the multiple cells C is further set with information about its location (location 132).

[0046] Map data with such a map data structure can, for example, when an in-vehicle device 2 functions as a guidance device that guides (directs) a vehicle V, such as for navigation or autonomous driving, refer to this map data to recognize and guide the vehicle, including markings other than white lines drawn on the road surface.

[0047] By the way, in the flowchart shown in Figure 4, all markings drawn on the road surface are divided into cells C. However, since lane lines (e.g., white lines) are simple lines, it is possible to further reduce the amount of data by storing, for example, the centerlines of the lane lines as a sequence of points in the map data. For this reason, lane lines may be excluded from the objects to be divided into cells C. Figure 9 shows a method for generating map data that divides into cells C while excluding lane lines.

[0048] First, in step S21, the control unit 11 acquires point cloud information and the like from the in-vehicle device 2, similar to step S11 in Figure 4.

[0049] Next, in step S22, the control unit 11 generates a binarized image based on the point cloud information acquired in step S21, similar to step S12 in Figure 4.

[0050] Next, in step S23, the control unit 11 extracts white lines (boundary lines) from the binarized image. An example of the method for extracting boundary lines in this step will be described. First, image processing is performed on the binarized image to detect, for example, line segments (straight lines). Then, the detected line segments (straight lines) are grouped together to extract the contour of one boundary line. Then, using known methods, the extracted boundary line is represented by a sequence of points. Furthermore, the ends and non-ends of the white lines may be recognized from the extracted boundary line, and a sequence of points may be interpolated between the recognized ends and non-ends (i.e., a sequence of points may be interpolated into the continuous portion between the ends).

[0051] In this way, the white line can be represented by a sequence of dots (see Figure 10). Figure 10 shows the portion of the white line W in Figure 7 as a sequence of dots. The white line W is represented by multiple dots P instead of cells C through the process in step S23 described above.

[0052] Returning to the explanation of Figure 9, in step S24, the control unit 11 removes pixels with high brightness outside the road surface and pixels representing the white lines extracted in step S23 from the binarized image generated in step S22. Pixels with high brightness outside the road surface can be removed in the same manner as in step S13. The white lines can be removed based on the contours extracted by the processing in step S23.

[0053] Next, in step S25, the control unit 11 groups the pixels corresponding to the markings drawn on the road surface into cells C of a certain size in the binarized image processed in step S24 (generates cells C). In this step, since the white lines have been removed in step S24, the markings drawn on the road surface, excluding the white lines, are divided into cells C.

[0054] Next, in step S26, the control unit 11, similar to step S15 in Figure 4, associates cell C with the point cloud obtained in step S21 by referring to the coordinates of cell C generated in step S25 in order to assign statistical information.

[0055] Then, in step S27, the control unit 11 assigns statistical information to each cell C, similar to step S16 in Figure 4. Map data is generated by including the markings drawn on the road surface, which consist of multiple cells C to which statistical information has been assigned, and the white lines represented as a sequence of points in step S23, into the map data.

[0056] Next, a self-position estimation method performed by the in-vehicle device 2 using map data having the map data structure generated by the method described above will be explained with reference to the flowchart in Figure 11. This flowchart is executed by the control unit 21 of the in-vehicle device 2. In other words, it is configured as a computer program (map data generation program) executed by the CPU of the control unit 21. Note that this program is not limited to being stored in the storage unit 23, but may also be stored on an optical disc, memory card, etc.

[0057] First, in step S31, the control unit 21 acquires point cloud information from the rider 3a. This point cloud information is information acquired by the rider 3a in real time. In other words, the control unit 21 functions as a sensor information acquisition unit that acquires point cloud information (sensor information) obtained from the rider 3a (sensor) through vehicle movement.

[0058] Next, in step S32, the control unit 21 recognizes the markings drawn on the road surface from the point cloud information acquired in step S31, for example by filtering the ground surface point cloud. That is, the control unit 21 functions as a recognition unit that recognizes point cloud information (sensor information) that corresponds to the markings drawn on the road surface from the point cloud information (sensor information) acquired by the sensor information acquisition unit.

[0059] Next, in step S33, the control unit 21 acquires map data from the storage unit 23. The map data acquired in this step has the map data structure described in Figure 8. That is, the control unit 21 functions as a map data acquisition unit that acquires map data in which markings drawn on the road surface are divided into a plurality of cells C, and each of the plurality of cells C has information about the location of the cell C set. Note that the map data may also be acquired directly from the server device 1.

[0060] Then, in step S34, the control unit 21 estimates its own position by matching the point cloud information processed in step S32 with the markings included in the map data. Specifically, it estimates its own position by matching the point cloud information with the statistical information of each cell on the map data side. As for the matching method, a well-known method of matching cells C with point clouds, such as NDT (Normal Distributions Transform), can be used. In other words, the control unit 21 functions as an estimation unit that estimates the vehicle's own position by matching the point cloud information (sensor information) corresponding to the markings recognized by the recognition unit with the map data.

[0061] Alternatively, the matching may be performed after generating statistical information by dividing the point cloud information acquired in real time from the RIDA3a into cells, similar to map data.

[0062] As is clear from the above explanation, step S31 functions as a sensor information acquisition step, step S32 as a recognition step, step S33 as a map data acquisition step, and step S34 as an estimation step.

[0063] Furthermore, while the above explanation described the generation of map data on server device 1, it is also possible to generate the map data on in-vehicle device 2 and upload it to server device 1. In other words, the flowcharts shown in Figures 4 and 9 are executed on in-vehicle device 2. In this case, server device 1 performs processing such as combining and updating the partial map data uploaded from each in-vehicle device 2.

[0064] Furthermore, in the above explanation, the vehicle that collects point cloud data and the vehicle that performs self-localization were the same vehicle V, but they may be different vehicles. For example, a measurement vehicle equipped with dedicated measuring equipment could collect point cloud data, and a general vehicle could receive the map data generated using the measurement results and perform self-localization.

[0065] Furthermore, in the above-described embodiment, the point cloud was acquired by a lidar 3a mounted on a vehicle. However, instead of a vehicle, a lidar may be mounted on a flying object such as an aircraft, and map data may be generated by irradiating the point cloud with laser light from above toward the ground and performing the binarization process described above on the obtained point cloud.

[0066] In this embodiment, the server device 1 has a communication unit 12 that acquires point cloud information detected by the lidar 3a, and a control unit 11 that generates a binarized image based on the point cloud information acquired by the communication unit 12. The portion of the binarized image that shows markings drawn on the road surface is then divided into multiple cells C, and map data is generated for each of the multiple cells C, including position 132 and statistical information 133. In this way, markings drawn on the road surface can be included in the map data, improving the accuracy of the vehicle V's self-position estimation. Furthermore, because the markings drawn on the road surface are represented by multiple cells C, complex processing such as extracting and plotting the contours of markings with complex shapes is not required, and since there is no need to reproduce the contours, the amount of information can be reduced.

[0067] Furthermore, since the control unit 11 sets statistical information 133 for each cell C based on the point cloud information, for example, by including the dispersion and reflectance of the point cloud within cell C as statistical information, it becomes possible to grasp the distribution and proportion of the paint indicating the markings within that cell C, in other words, it becomes possible to identify the cell C that contains the outline of the markings.

[0068] Furthermore, markings painted on the road surface may include markings other than lane markings. By doing so, it becomes possible to include road markings other than lane markings such as white lines, such as lane markings, as well as lane markings indicating direction of travel such as right turns and left turns, and speed limits, in the map data.

[0069] Furthermore, in this embodiment, the map data structure represents markings drawn on the road surface by dividing them into multiple cells C, and each of the multiple cells C has a position 132 assigned to it. The in-vehicle device 2, which utilizes this map data structure, recognizes the markings drawn on the road surface by the position 132 of the cell C and performs self-position estimation processing for the vehicle V. By doing so, markings drawn on the road surface can be included in the map data, and the accuracy of self-position estimation for the vehicle V can be improved. In addition, since the markings are represented by multiple cells C, complex processing such as extracting and plotting the contours of markings with complex shapes is not required, and since there is no need to reproduce the contours, the amount of information can be reduced.

[0070] Furthermore, the map data structure contains statistical information for each cell C. By doing so, it becomes possible to store statistical information such as the coordinates of the point cloud and the variance of the reflectivity within cell C, thereby understanding the distribution and proportion of paint indicating markings within that cell C, and in other words, identifying cells that contain the outlines of markings.

[0071] Furthermore, since the server device 1 stores map data having the map data structure described above, it can distribute map data to multiple vehicles.

[0072] Furthermore, the in-vehicle device 2 represents markings drawn on the road surface by dividing them into multiple cells C, and the control unit 21 acquires map data in which the position 132 of each cell C is set, and acquires point cloud information obtained from the rider 3a as the vehicle moves. The control unit 21 then recognizes point cloud information corresponding to the markings drawn on the road surface from the acquired point cloud information, and estimates the vehicle's own position by matching the recognized point cloud information corresponding to the markings drawn on the road surface with the map data. In this way, the vehicle's own position can be estimated using markings drawn on the road surface, and the accuracy of the vehicle V's own position estimation can be improved. [Examples]

[0073] Next, a map data generation device according to a second embodiment of the present invention will be described with reference to Figures 12 to 13. Note that parts identical to those in the first embodiment described above are denoted by the same reference numerals and their descriptions are omitted.

[0074] This embodiment has the same configuration as Figure 1, but the method of generating map data is different. Figure 12 shows a flowchart of the map data generation method according to this embodiment. This flowchart is executed by the control unit 11 of the server device 1.

[0075] First, in step S41, the control unit 11 acquires point cloud information and the like from the in-vehicle device 2.

[0076] Next, in step S42, the control unit 11 filters the point cloud acquired in step S41 based on the reflectance intensity information and height information (distance information) that each point cloud possesses. Filtering by reflectance intensity information allows for the extraction of markings drawn on the road surface, such as road markings made of retroreflective material, which have high reflectance. Filtering by height information allows for the exclusion of points with high reflectance other than the road surface, such as signs.

[0077] Next, in step S43, the point cloud information processed in step S42 is grouped into cells C of a fixed size, similar to step S14 (cell C is generated). At this step, each cell C is assigned positional information based on the absolute coordinates of each point in the point cloud information.

[0078] Then, in step S44, the control unit 11 assigns statistical information as attribute information to each cell C, similar to step S16 and so on. Map data is generated by including the markings drawn on the road surface, which consist of multiple cells C to which statistical information has been assigned, into the map data.

[0079] Here, as with the first embodiment, the boundary lines may be excluded from the object to be divided into cells C. Figure 13 shows a method for generating map data that divides into cells C while excluding the boundary lines.

[0080] First, in step S51, the control unit 11 acquires point cloud information and the like from the in-vehicle device 2, similar to step S21 described above.

[0081] Next, in step S52, the control unit 11 generates a binarized image based on the point cloud information acquired in step S51, similar to step S22 described above.

[0082] Next, in step S53, the control unit 11 extracts white lines (section lines) from the binarized image.

[0083] Next, in step S54, the control unit 11 filters the point cloud information acquired in step S51 for the point cloud corresponding to the white line. The control unit 11 also filters based on the reflection intensity information and height information (distance information) that each point cloud possesses, similar to step S42.

[0084] Next, in step S55, the control unit 11 groups the point cloud information processed in step S54 into cells C of a fixed size, similar to step S43 (generates cells C).

[0085] Next, in step S56, the control unit 11 assigns statistical information as attribute information to each cell C, similar to step S44. Map data is generated by including the markings drawn on the road surface, which consist of multiple cells C to which statistical information has been assigned, and the white lines represented as a sequence of points in step S53, into the map data.

[0086] In this embodiment, the server device 1 has a communication unit 12 that acquires point cloud information detected by the lidar 3a, and a control unit 11 that filters the point cloud information acquired by the communication unit 12. The filtered point cloud then divides the portion of the markings drawn on the road surface into multiple cells C, and generates map data for each of the multiple cells C, including position 132 and statistical information 133. In this way, markings drawn on the road surface can be directly converted into cells from the point cloud information.

[0087] In the two embodiments described above, cell C was planar, but it may also be divided into three-dimensional structures such as voxels or prisms.

[0088] Furthermore, the present invention is not limited to the embodiments described above. That is, those skilled in the art can implement the invention in various ways, without departing from the core principles, in accordance with prior art knowledge. Such modifications, as long as they still incorporate the map data generation device, map data structure, and self-position estimation device of the present invention, are of course included within the scope of the present invention. [Explanation of Symbols]

[0089] 1. Server equipment (storage device, map data generation device) 2 In-vehicle device (self-position estimation device) 3a LiDAR (sensor) 11 Communications Department (Acquisition Department) 12 Control Unit (Signage Information Generation Unit, Map Data Generation Unit) 13 Storage unit (storage medium) 21 Control Unit (Map data acquisition unit, sensor information acquisition unit, recognition unit, estimation unit) 23 Storage unit (storage medium) 132 Position (information about the cell's location) 133 Statistical information (attribute information)

Claims

[Claim 1] A map data acquisition unit acquires map data in which markings drawn on the road surface are divided into multiple cells, and each of the multiple cells has information about the location of that cell set. A sensor information acquisition unit that acquires sensor information obtained from sensors during vehicle operation, A recognition unit recognizes the sensor information corresponding to the marking from the sensor information acquired by the sensor information acquisition unit, An estimation unit estimates the vehicle's own position by matching the sensor information corresponding to the marking recognized by the recognition unit with the map data. A self-position estimation device characterized by comprising the following features.

Images