JP6687568B2 - Boundary estimation device, boundary estimation method, and boundary estimation program - Google Patents

Description

  The present disclosure relates to a boundary estimation device, a method, and a program for estimating a lane boundary line from data of a point cloud indicating the surface of a road.

  As in Patent Document 1, there is known a mobile mapping system that measures coordinate point cloud data of a target road by traveling on the target road with a measurement vehicle equipped with a laser scanner. In recent years, with the spread of devices that generate three-dimensional point cloud data, such as 3D laser scanners, UAVs, and mobile mapping systems, the situation of using a large amount of point cloud data is increasing. UAV is an abbreviation for Unmanned Aerial Vehicle.

  Conventionally, in order to create a highly accurate three-dimensional road map, a map creator specifies a lane boundary line by referring to a three-dimensional point cloud, and maps the specified lane boundary line.

JP, 2013-232241, A

As described above, since the map worker creates the lane boundary line by referring to the three-dimensional point cloud, there is a problem that the efficiency of creating the three-dimensional road map is low.
The present disclosure aims to improve the efficiency of creating a three-dimensional road map.

  One aspect of the present disclosure is a boundary estimation device that estimates a lane boundary line from data of a coordinate point group indicating a surface of a target road by a set of coordinate points represented by three-dimensional coordinates, and includes a region setting unit and a boundary. The determination unit, the column cluster setting unit, and the row cluster setting unit are provided. Boundary line determination information for determining whether or not a lane boundary line is located at a point corresponding to the coordinate point is added to the coordinate point data.

  The area setting unit divides the surface of the target road into a plurality of extension direction dividing lines along the road extension direction in which the target road extends and a plurality of vertical direction dividing lines along a direction perpendicular to the road extension direction. Then, each of the plurality of divided areas is set as a divided area.

  The boundary judgment unit arranges the lane boundary line in each of the plurality of divided areas set by the area setting unit based on the boundary judgment information added to the data of the coordinate points included in the divided area. Configured to determine whether or not it has been done.

  The column cluster setting unit is configured to set, for each of the plurality of divided region columns, a set of effective divided regions in which the effective divided regions are continuously arranged along the column direction as one column cluster. The effective division area is a division area determined by the boundary determination unit to have a lane boundary line. The column direction is a direction perpendicular to the road extension direction. The divided area row is a set of divided areas in one row along the row direction.

  The row cluster setting unit is configured to select a column cluster included in the divided area column for each of the plurality of divided area columns. Then, the row cluster setting unit determines that the selected column cluster that is the selected column cluster and the adjacent column cluster that is the column cluster included in the divided region column that is adjacent to the selected column cluster can be combined as a single lane boundary line. It is configured to determine whether or not the preset cluster connection condition shown is satisfied. When the row cluster setting unit determines that the cluster joining condition is met, the row cluster setting unit joins the selected column cluster and the adjacent column cluster, and the column clusters are arranged continuously along the road extension direction. Is configured as a row cluster.

  The boundary estimation device of the present disclosure configured in this manner determines whether or not the lane boundary is the divided area (that is, the effective divided area) based on the boundary determination information, and determines the effective divided area. A lane boundary line is estimated by connecting the lane boundaries along the row direction and the road extension direction. Thereby, the boundary estimation device of the present disclosure can be automated by a computer in order to estimate the lane boundary line. Therefore, the boundary estimation apparatus of the present disclosure does not require the map creation operator to map the lane boundaries while referring to the coordinate point group, and can improve the creation efficiency of the three-dimensional road map.

  In one aspect of the present disclosure, a cluster coupling unit may be included. The cluster joining unit has a plurality of row clusters located on the same road extension line extending along the road extension direction, and a plurality of row clusters being less than a preset joining determination distance along the road extension line. It is configured to combine a plurality of row clusters as one lane boundary when they are spaced apart. Thereby, the boundary estimation device of the present disclosure can set a dashed lane boundary line formed by a plurality of row clusters as one lane boundary line.

  In one aspect of the present disclosure, a column cluster exclusion unit may be included. The column cluster excluding unit is configured to exclude a column cluster whose length along the direction perpendicular to the road extension direction does not match the preset border line determination width. Thereby, the boundary estimation device of the present disclosure can suppress the occurrence of a situation in which a column cluster that is not a lane boundary line is determined to be a lane boundary line.

  In one aspect of the present disclosure, the boundary estimation device of the present disclosure may include a continuous line setting unit. The continuous line setting unit calculates a center position in a direction perpendicular to the road extension direction for each of the plurality of column clusters forming the row cluster, and connects the calculated plurality of center positions along the road extension direction. Then, the continuous line is set. Thereby, the boundary estimation device of the present disclosure can set a continuous line passing through the center of the lane boundary line.

  Another aspect of the present disclosure is a boundary estimation method for estimating a lane boundary line from data of a coordinate point group indicating the surface of a target road by a set of coordinate points represented by three-dimensional coordinates, and an area setting procedure, A boundary determination procedure, a column cluster setting procedure, and a row cluster setting procedure are provided.

  The area setting procedure divides the surface of the target road into a plurality of extension direction dividing lines along the road extension direction in which the target road extends and a plurality of vertical direction dividing lines along the direction perpendicular to the road extension direction. Then, each of the plurality of divided areas is set as a divided area.

  For the boundary judgment procedure, for each of the plurality of divided areas set by the area setting procedure, the lane boundary line is arranged in the divided area based on the boundary judgment information added to the data of the coordinate points included in the divided area. It is determined whether or not it has been done.

  The column cluster setting procedure sets, for each of the plurality of divided region columns, a set of effective divided regions in which effective divided regions are continuously arranged along the column direction as one column cluster.

  The row cluster setting procedure selects a column cluster included in the divided area column for each of the plurality of divided area columns. After that, the row cluster setting procedure confirms that the selected column cluster that is the selected column cluster and the adjacent column cluster that is the column cluster included in the divided area column that is adjacent to the selected column cluster can be combined as a single lane boundary line. It is determined whether or not the preset cluster connection condition shown is satisfied. Then, when it is determined that the cluster joining condition is satisfied, the row cluster setting procedure joins the selected column cluster and the adjacent column cluster, and the column clusters are arranged continuously along the road extension direction. Is set as one row cluster.

The boundary estimation method of the present disclosure is a method executed by the boundary estimation device of the present disclosure, and by executing the method, the same effect as the boundary estimation device of the present disclosure can be obtained.
Yet another aspect of the present disclosure is a boundary estimation program that causes a computer to function as an area setting unit, a boundary determination unit, a column cluster setting unit, and a row cluster setting unit.

  The computer controlled by the boundary estimation program of the present disclosure can configure a part of the boundary estimation device of the present disclosure, and can obtain the same effect as the boundary estimation device of the present disclosure.

It is a block diagram showing a configuration of a boundary estimation apparatus 1. It is a flowchart which shows a boundary estimation process. It is a flowchart which shows a characteristic matrix creation process. It is a figure explaining the method of setting a local feature area | region in a linear target road. It is a figure explaining the method of setting a local feature area | region in an S-shaped target road. It is a flow chart which shows local shape extraction processing. It is a flowchart which shows a clustering process. It is a flowchart which shows a column direction clustering process. It is a flowchart which shows a row direction clustering process. It is a figure which shows the specific example of a clustering process. It is a flow chart which shows local shape combination processing. It is a figure explaining the method of joining a row level cluster. It is a flowchart which shows a continuous line calculation process.

Embodiments of the present disclosure will be described below with reference to the drawings.
As shown in FIG. 1, the boundary estimation apparatus 1 of this embodiment includes a display unit 11, an operation input unit 12, a data storage unit 13, a data input / output unit 14, and a control unit 15.

The display unit 11 includes a display device (not shown) and displays various images on the display screen of the display device.
The operation input unit 12 outputs input operation information for specifying an input operation performed by the user via a keyboard and a mouse (not shown).

The data storage unit 13 is a storage device for storing various data.
The data input / output unit 14 inputs / outputs data to / from an external device connected by wire or wirelessly.

  The control unit 15 is mainly composed of a well-known microcomputer including a CPU, a ROM, a RAM, an I / O, and a bus line connecting these components. Various functions of the microcomputer are realized by the CPU executing the programs stored in the non-transitional substantive recording medium. In this example, the ROM corresponds to the non-transitional substantive recording medium storing the program. Further, by executing this program, the method corresponding to the program is executed. Then, the control unit 15 executes various processes based on the inputs from the operation input unit 12 and the data input / output unit 14, and controls the display unit 11, the data storage unit 13, and the data input / output unit 14.

  In the boundary estimation device 1 configured in this way, the control unit 15 executes a boundary estimation process. It should be noted that some or all of the functions executed by the control unit 15 may be configured as hardware by one or a plurality of ICs or the like.

  Here, the procedure of the boundary estimation process executed by the control unit 15 will be described. The boundary estimation process is executed by activating the boundary estimation program 20 stored in the control unit 15 in order to execute the boundary estimation process by a user's input operation. The boundary estimation program 20 may be installed in the boundary estimation apparatus 1 in advance, or may be installed via a recording medium or a network. Examples of the recording medium include an optical disc, a magnetic disc, and a semiconductor memory.

  When the boundary estimation process is executed, the control unit 15 first, as shown in FIG. 2, in S10, an image for selecting a road (hereinafter referred to as a target road) for which road map data is to be created (hereinafter referred to as a target road). Hereinafter, the target road selection image) is displayed on the display screen of the display unit 11.

  After that, in S20, it is determined whether or not the target road specifying information for specifying the target road selected by the user is input from the operation input unit 12. If the target road identification information has not been input, the processing in S20 is repeated to wait until the target road identification information is input. When the target road identification information is input, the coordinate point group data of the target road identified by the target road identification information is acquired from the data storage unit 13 in S30. The coordinate point cloud data of the target road is stored in the data storage unit 13 in advance before starting the boundary estimation process.

  The coordinate point cloud data of the target road is a set of three-dimensional coordinates of a plurality of points representing the three-dimensional shape of the target road, and is acquired by using, for example, a mobile mapping system (hereinafter, MMS). MMS is an abbreviation for Mobile Mapping System. The MMS measures coordinate point cloud data of the target road and its surroundings by running on the target road with a measurement vehicle equipped with a laser scanner. The coordinate point cloud data of the target road and its surroundings is based on the current position of the measurement vehicle, the direction in which the laser scanner radiates the pulse laser light, and the time from the irradiation of the pulse laser light to the detection of the reflected laser light. It is acquired by measuring the three-dimensional coordinates of the point where the laser light is reflected. The data structure of the coordinate point cloud data may be a text format, or a general point cloud format that has already been distributed (for example, LAS format) may be used, and the format is not limited. Hereinafter, each of the plurality of points forming the coordinate point group represented by the coordinate point group data acquired in S30 is referred to as a coordinate point.

In addition, each of the plurality of point data forming the coordinate point group data includes coordinate values represented by three-dimensional coordinates and attached information indicating the detected reflection intensity of the reflected laser light.
Then, in S40, a characteristic matrix creating process is executed. Here, the procedure of the characteristic matrix creating process of S40 will be described.

  When the feature matrix creating process is executed, the control unit 15 first sets a vertical feature region in S110, as shown in FIG. In S110, first, as shown in FIG. 4, data indicating the traveling locus Ps of the target road Ro is acquired from the data storage unit 13. The traveling locus Ps is a locus when the measurement vehicle travels on the target road. The data indicating the traveling locus Ps is stored in the data storage unit 13 in advance before the boundary estimation process is started. Next, in S110, as shown in FIG. 4, two vertical region setting lines L11 and L12 are set. The vertical region setting lines L11 and L12 are set to be parallel to the traveling locus Ps and sandwich the traveling locus Ps and the target road Ro between the vertical region setting line L11 and the vertical region setting line L12. Further, in S110, two vertical area setting lines Ll3 and Ll4 are set. The vertical area setting line L13 is set to be perpendicular to the traveling locus Ps at the start point of the target road Ro. The vertical region setting line L14 is set to be perpendicular to the traveling locus Ps at the end point of the target road Ro. Then, in S110, as shown in FIG. 4, the region surrounded by the vertical region setting lines L11, L12, L13, L14 is set as the vertical characteristic region R11.

  Then, when the process of S110 is completed, as shown in FIG. 3, a crossing characteristic region is set in S120. In S120, first, as shown in FIG. 4, a crossing area setting line Lt perpendicular to the traveling locus Ps is set for each preset division distance Dd1 along the traveling locus Ps. Then, in S120, each of the plurality of regions divided by the transverse region setting line Lt in the vertical characteristic region Rl is set as the transverse characteristic region Rt.

  Then, when the processing of S120 ends, as shown in FIG. 3, a local feature region is set in S130. In S130, first, as shown in FIG. 4, a plurality of local area setting lines Lr are set. The local area setting line Lr is set to be parallel to the traveling locus Ps for each division distance Dd2 set in advance along the vertical area setting line Ll3. Then, in S130, in each of the plurality of transverse characteristic regions Rt, each of the plurality of regions divided by the local region setting line Lr is set as the local characteristic region Rr.

  Then, when the process of S130 ends, as shown in FIG. 3, the feature amount of the local feature region is calculated using the attached information in S140. In S140, specifically, first, for each of the plurality of local feature regions, all coordinate points included in the local feature region are extracted. Then, for each of the plurality of local feature regions, the average value of the reflection intensities at all the extracted coordinate points is calculated as the feature amount of the local feature region.

  Next, in S150, a feature matrix is created, and the feature matrix creation process ends. The feature matrix is created by arranging the feature amounts of the plurality of local feature regions calculated in S140 in a matrix corresponding to the positions of the local feature regions. As shown in FIG. 4, the vertical characteristic region Rl has a plurality of local features in a matrix of m rows × n columns, with rows in the direction along the travel locus Ps and columns in the direction perpendicular to the travel locus Ps. It is divided into areas. Note that m and n are positive integers. Therefore, the feature matrix is a matrix with m rows and n columns. Then, for example, the local characteristic region Rra shown in FIG. 4 is arranged at the position of 5 rows and 3 columns in the vertical characteristic region Rl divided into a matrix of m rows × n columns. Therefore, the feature amount of the local feature region Rra is arranged at the position of 5 rows and 3 columns in the feature matrix.

  Further, as shown in FIG. 5, for example, even when the target road Ro is curved like an S shape, a crossing region setting line Lt perpendicular to the curved traveling locus Ps, The local feature region Rr can be set by setting the local region setting line Lr parallel to the traveling locus Ps.

Then, when the feature matrix creating process of S40 is completed, the local shape extracting process is executed in S50 as shown in FIG. Here, the procedure of the local shape extraction processing of S50 will be described.
When the local shape extraction process is executed, the control unit 15 first binarizes the feature amount of the feature matrix in S210, as shown in FIG. Specifically, for each of the m × n feature amounts arranged in the feature matrix, the feature amount is set to 1 when the feature amount is equal to or greater than a preset binarization determination value, and If the amount is less than the binarization determination value, the feature amount is set to 0. The binarization determination value is set so that the feature amount becomes 1 when the local feature region is white or yellow.

  Then, in S220, a standard value that has not been selected in this local shape extraction processing is selected from among a plurality of standard values that have been set in advance. Standards are often defined for the white and yellow lines used to represent lane boundaries. The standard value selected in S220 is the width of the white line and the yellow line defined as the standard at the lane boundary line.

Then, in S230, a clustering process is executed. Here, the procedure of the clustering process of S230 will be described.
When the clustering process is executed, as shown in FIG. 7, the control unit 15 first selects the column on the leftmost side from the columns that have not been selected in the feature matrix in S310. Then, in S320, column-direction clustering processing is executed. Here, the procedure of the column direction clustering process in S320 will be described.

  When the column-direction clustering process is executed, as shown in FIG. 8, the control unit 15 first selects the highest element among the unselected elements in the column selected in S310 in S410. The element is a value arranged in the feature matrix, and is set to 1 or 0 by the process of S210.

  Then, in S420, it is determined whether the element selected in S410 (hereinafter, selected element) is valid. In S420, specifically, when the selection element is 1, it is determined to be valid. Here, if the selected element is valid, the selected element is adopted, and the process proceeds to S460.

  On the other hand, if the selected element is not valid, in S440, the width of the plurality of elements (hereinafter, adopted element group) continuously adopted in S430 is substantially equal to the standard value selected in S220. Judge whether or not. If the width of the adopted element group does not substantially match the standard value, the process proceeds to S460. On the other hand, when the width of the adopted element group is substantially equal to the standard value, the adopted element group is registered as a column level cluster (hereinafter referred to as a column level cluster) in S450, and the process proceeds to S460.

  Then, in S460, it is determined whether or not all elements have been selected in the column selected in S310. Here, if all the elements have not been selected, the process proceeds to S410. On the other hand, when all the elements have been selected, the column direction clustering process ends.

  Then, when the column direction clustering process of S320 ends, as shown in FIG. 7, in S330, it is determined whether or not all columns have been selected in the feature matrix. Here, if all columns are not selected, the process proceeds to S310. On the other hand, if all the columns have been selected, in S340, the column on the leftmost side is selected from the columns not yet selected in the feature matrix. Then, in S350, a row-direction clustering process is executed. Here, the procedure of the row-direction clustering processing in S350 will be described.

  When the row-direction clustering process is executed, the control unit 15 first selects, in S510, a column level cluster that has not been selected in the column selected in S340, as shown in FIG. Next, in S520, the column level cluster closest to the column level cluster selected in S510 is selected from the columns on the right of the column selected in S340.

  Then, in S530, it is determined whether or not the distance between the column level cluster selected in S510 and the column level cluster selected in S520 (hereinafter, the selected cluster distance) is smaller than a preset connection determination value. . The selected inter-cluster distance is, for example, the distance between the center position of the column level cluster selected in S510 and the center position of the column level cluster selected in S520. Here, if the distance between the selected clusters is equal to or greater than the connection determination value, the process proceeds to S550.

  On the other hand, when the distance between the selected clusters is smaller than the connection determination value, in S540, the column level cluster selected in S510 and the column level cluster selected in S520 are combined, and the combined column level cluster is It is registered as a row-level cluster (hereinafter, row-level cluster), and the process proceeds to S550.

  Then, in S550, it is determined whether or not all column level clusters have been selected in the column selected in S340. Here, if all the column level clusters have not been selected, the process proceeds to S510. On the other hand, when all the column level clusters have been selected, the row direction clustering process ends.

  Then, when the row-direction clustering process of S350 is completed, as shown in FIG. 7, in S360, it is determined whether or not all columns are selected in the feature matrix. Here, if all the columns are not selected, the process proceeds to S340. On the other hand, when all columns have been selected, the clustering process ends.

  Here, a specific example of the clustering process will be described. As shown in FIG. 10, for simplification of description, only the portion where the asphalt AP is arranged on both sides of one white line WL is set as the target road Ro. When the column direction clustering process is executed, first, the crossing characteristic region Rt1 set on the leftmost side of the target road Ro is selected. That is, the first column in the feature matrix is selected. The transverse characteristic region Rt1 includes local characteristic regions Rr1, Rr2, Rr3, Rr4, Rr5, Rr6, Rr7, Rr8 in order from the top. Then, it is determined whether or not the elements of the feature matrix corresponding to the local feature regions Rr1 to Rr8 are valid from the top. As a result, the local characteristic regions Rr3, Rr4, Rr5, Rr6 are adopted. This is because the local characteristic regions Rr3 to Rr6 are regions set on the white line. Then, it is assumed that the width from the local characteristic region Rr3 to the local characteristic region Rr6 substantially matches the standard value. As a result, the aggregate of the local characteristic regions Rr3 to Rr6 is registered as the column level cluster Cc1.

  Next, the transverse characteristic region Rt2 set to the right of the transverse characteristic region Rt1 is selected. That is, the second column in the feature matrix is selected. The transverse characteristic region Rt2 includes local characteristic regions Rr11, Rr12, Rr13, Rr14, Rr15, Rr16, Rr17, Rr18 in order from the top. Then, it is determined whether or not the elements of the feature matrix corresponding to the local feature regions Rr11 to Rr18 are valid from the top. As a result, the local characteristic regions Rr13, Rr14, Rr15, Rr16 are adopted. This is because the local characteristic regions Rr13 to Rr16 are regions set on the white line. Then, it is assumed that the width from the local characteristic region Rr13 to the local characteristic region Rr16 substantially matches the standard value. As a result, the aggregate of the local characteristic regions Rr13 to Rr16 is registered as the column level cluster Cc2.

  Then, column level clusters are similarly registered up to the rightmost crossing characteristic region Rt6 on the target road Ro. As a result, the column level clusters Cc1, Cc2, Cc3, Cc4, Cc5, Cc6 are registered.

  Next, when the row-direction clustering process is executed, first, the column level cluster Cc1 included in the transverse characteristic region Rt1 is selected. Then, the column level cluster Cc2 is selected in the transverse characteristic region Rt2 located on the right of the transverse characteristic region Rt1. Here, since the distance between the column level cluster Cc1 and the column level cluster Cc2 becomes the connection determination value, the column level cluster Cc1 and the column level cluster Cc2 are combined and registered as a row level cluster.

  Then, the row level clusters are similarly registered up to the rightmost crossing characteristic region Rt6 on the target road Ro. As a result, a set of column level clusters Cc1, Cc2, Cc3, Cc4, Cc5, Cc6 is registered as one row level cluster Cr1.

  Then, when the clustering process of S230 is completed, as shown in FIG. 6, it is determined in S240 whether all standard values have been selected. If all standard values have not been selected, the process proceeds to S220. On the other hand, when all standard values are selected, the local shape combination process is executed in S250. Here, the procedure of the local shape combination processing of S250 will be described.

  When the local shape combination process is executed, the control unit 15 first selects all row level clusters having the same standard value in S610, as shown in FIG. Then, in S620, among the row level clusters selected in S610, between the row level clusters adjacent to each other, the parallel distance Dp along the direction parallel to the traveling locus Ps and the direction perpendicular to the traveling locus Ps. The vertical distance Dv along is calculated. For example, as shown in FIG. 12, assume that row level clusters Cr11, Cr12, Cr13, Cr14, Cr15, Cr16, Cr17, Cr18 are registered for the target road Ro.

  For example, the row level cluster Cr11 calculates the parallel distance Dp and the vertical distance Dv with the row level clusters Cr13, Cr15, Cr17. The row level cluster Cr13 calculates the parallel distance Dp and the vertical distance Dv with the row level clusters Cr11, Cr14, Cr15, Cr16. The row level cluster Cr16 calculates the parallel distance Dp and the vertical distance Dv with the row level clusters Cr12, Cr13, Cr14, Cr15, Cr16, Cr17, Cr18.

  Then, in S630, by using the parallel distance Dp and the vertical distance Dv calculated in S620, the parallel distance Dp is less than a preset parallel connection determination value and the vertical distance Dv is preset. Join row-level clusters that are less than the value. For example, in FIG. 12, the parallel distance Dp and the vertical distance Dv between the row level cluster Cr13 and the row level cluster Cr15 are less than the parallel connection determination value and less than the vertical connection determination value, respectively. Further, the parallel distance Dp and the vertical distance Dv between the row level cluster Cr15 and the row level cluster Cr17 are less than the parallel connection determination value and less than the vertical connection determination value, respectively. Therefore, the row level clusters Cr13, Cr15, Cr17 are combined. Similarly, row level clusters Cr14, Cr16, Cr18 are combined.

  Then, when the processing of S630 ends, as shown in FIG. 11, in S640, it is determined whether or not the row level cluster has been selected with all the standard values. If no row level cluster is selected for all standard values, the process proceeds to S610. On the other hand, when the row level clusters are selected with all the standard values, the local shape combination processing ends.

Then, when the local shape combination processing of S250 ends, the local shape extraction processing ends, as shown in FIG.
Then, when the local shape extraction processing of S50 is completed, a continuous line calculation processing is executed in S60, as shown in FIG. Here, the procedure of the continuous line calculation processing in S60 will be described.

  When the continuous line calculation process is executed, the control unit 15 first selects one row level cluster in S710, as shown in FIG. However, a set of a plurality of row level clusters combined in S630 is selected as one row level cluster.

  Next, in S720, the position of the center point is calculated for each column for the row level cluster selected in S710. For example, as shown in FIG. 10, the row level cluster Cr1 is a set of column level clusters Cc1, Cc2, Cc3, Cc4, Cc5 and Cc6. Therefore, the position of the center point is calculated for each of the column level clusters Cc1, Cc2, Cc3, Cc4, Cc5, Cc6. In FIG. 10, points Pc1, Pc2, Pc3, Pc4, Pc5, Pc6 are shown as the center points of the column level clusters Cc1, Cc2, Cc3, Cc4, Cc5, Cc6.

  Then, in S730, the center points calculated in S720 are connected, and the continuous line obtained by the connection is registered. In FIG. 10, a continuous line is a line in which points Pc1, Pc2, Pc3, Pc4, Pc5, and Pc6 are sequentially connected by a straight line.

  Then, in S740, it is determined whether all row level clusters have been selected. Here, when all the row level clusters are not selected, the process proceeds to S710. On the other hand, when all the row level clusters have been selected, the continuous line calculation process ends.

Then, when the continuous line calculation process of S60 ends, as shown in FIG. 2, the data of the continuous line registered in S60 is stored in the data storage unit 13 in S70, and the boundary estimation process ends.
The boundary estimation device 1 configured in this way estimates a lane boundary line from the data of the coordinate point group indicating the surface of the target road by a set of coordinate points represented by three-dimensional coordinates. The coordinate point data is provided with accompanying information indicating the detected reflection intensity of the reflected laser light in order to determine whether or not the lane boundary line is arranged at the point corresponding to the coordinate point.

  The boundary estimation apparatus 1 extends on the surface of the target road Ro along a plurality of local region setting lines Lr along the traveling locus Ps along which the target road Ro extends and along a direction (hereinafter, column direction) perpendicular to the traveling locus Ps. Further, the plurality of divided regions are divided by the plurality of crossing region setting lines Lt, and each of the divided regions is set as the local characteristic region Rr.

  For each of the set local feature regions Rr, the boundary estimation device 1 determines the lane boundary line in the local feature region Rr based on the attached information added to the data of the coordinate points included in the local feature region Rr. It is determined whether or not it is placed. Hereinafter, the local feature region Rr determined to have the lane boundary line is referred to as an effective local region.

  The boundary estimation apparatus 1 includes an effective local area in which the effective local areas are continuously arranged along the column direction for each of a set of local feature areas Rr in one row along the column direction (hereinafter, local area row). Set as one column level cluster.

  The boundary estimation device 1 selects a column level cluster included in the local region sequence for each of the plurality of local region sequences. After that, the boundary estimation apparatus 1 sets between the selected column level cluster (hereinafter, selected column level cluster) and the column level cluster (hereinafter, adjacent column level cluster) included in the local region column adjacent to the selected column level cluster. The distance (that is, the distance between selected clusters) is smaller than a preset connection determination value. The determination condition that the distance between selected clusters is smaller than the connection determination value is a cluster connection condition that indicates that the selected column level cluster and the adjacent column level cluster can be combined as a single lane boundary line.

  When the boundary estimation device 1 determines that the distance between the selected clusters is smaller than the connection determination value, the boundary estimation device 1 joins the selected column level cluster and the adjacent column level cluster, and the column level clusters continue along the traveling locus Ps. A set of column-level clusters arranged in a row is set as one row-level cluster.

  As described above, the boundary estimation device 1 determines whether or not the lane boundary is the local characteristic region Rr (that is, the effective local region) based on the attached information, and the effective local region is arranged in the row direction and the traveling direction. The lane boundary line is estimated by connecting along the locus Ps. Thereby, the boundary estimation device 1 can be automated by a computer in order to estimate the lane boundary line. For this reason, the boundary estimation apparatus 1 does not need to map the lane boundary line by referring to the coordinate point group by the map creator, and can improve the efficiency of creating the three-dimensional road map.

  Further, the boundary estimation apparatus 1 determines that, in a plurality of row level clusters, the parallel distance Dp is less than a preset parallel connection determination value and the vertical distance Dv is less than a preset vertical connection determination value, Combine multiple row level clusters into one lane boundary. The vertical distance Dv being less than the preset vertical connection determination value indicates that the plurality of row level clusters are located on the same road extension line extending along the traveling locus Ps. As a result, the boundary estimation apparatus 1 can set a broken lane boundary line formed by a plurality of row level clusters as one lane boundary line.

  In addition, the boundary estimation device 1 determines a column level cluster whose length (that is, width) along the direction (that is, column direction) perpendicular to the traveling locus Ps does not match a preset standard value. exclude. Thereby, the boundary estimation apparatus 1 can suppress the occurrence of a situation in which a column level cluster that is not a lane boundary line is determined to be a lane boundary line.

  Further, the boundary estimation apparatus 1 calculates the position of the center point in the column direction for each of the plurality of column level clusters forming the row level cluster, and connects the calculated positions of the plurality of center points along the travel locus Ps. And set a continuous line. Thereby, the boundary estimation device 1 can set a continuous line passing through the center of the lane boundary line.

  In the embodiment described above, S110 to S150 correspond to the processing as the area setting unit and the area setting procedure, S210, S220, S410, and S420 correspond to the processing as the boundary determining unit and the boundary determining procedure, and S430 to S450. Corresponds to the processing as the column cluster setting unit and the column cluster setting procedure, and S510 to S550 correspond to the processing as the row cluster setting unit and the row cluster setting procedure.

  Further, the attached information corresponds to the boundary line determination information, the traveling locus Ps corresponds to the road extension direction, the local area setting line Lr corresponds to the extension direction dividing line, and the crossing area setting line Lt corresponds to the vertical direction dividing line. However, the local characteristic region Rr corresponds to the divided region.

Further, the column level cluster corresponds to the column cluster, the determination condition of S530 corresponds to the cluster combination condition, and the row level cluster corresponds to the row cluster.
Further, S610 to S640 correspond to the process as the cluster connecting unit, the parallel connection determination value corresponds to the connecting determination distance, S440 corresponds to the process as the column cluster excluding unit, and the standard value corresponds to the boundary line determination width. However, S710 to S740 correspond to the processing as the continuous line setting unit.

Although one embodiment of the present disclosure has been described above, the present disclosure is not limited to the above embodiment, and various modifications can be implemented.
For example, in the above-described embodiment, the mode in which the average value of the reflection intensities is calculated as the feature amount of the local feature region in S140 has been described. However, the feature amount of the local feature area is not limited to the average value of the reflection intensity, and any information that can determine whether or not the lane boundary line is arranged in the local feature area may be used. For example, the plurality of point data forming the coordinate point group data may include the information indicating the altitude of the coordinate point instead of the reflection intensity as the additional information. Then, as the feature amount of the local feature area, the minimum value of the altitudes of all the coordinate points included in the local feature area may be adopted. Since the white line or the yellow line indicating the lane boundary line is slightly raised with respect to the surrounding asphalt, it is possible to determine whether or not the lane boundary line is arranged in the local feature area based on the altitude.

  Further, in the above-described embodiment, the form in which the data indicating the traveling locus Ps of the target road Ro is acquired from the data storage unit 13 in S110 is described, but the data indicating the line along the road extension direction in which the target road extends is acquired. I wish I could. For example, the user of the boundary estimation device 1 may set a line along the road extension direction on the display screen of the display unit 11.

  Further, in the above-described embodiment, the form in which the average value of the reflection intensity is calculated as the feature amount of the local feature region is shown. However, as the feature amount of the local feature region, the color of the coordinate points included in the local feature region, the average value of RGB values, the maximum value of the altitude, the variance of the altitude, or the like may be adopted.

  In the above embodiment, the feature amount is set to 1 when the feature amount is equal to or greater than the preset binarization determination value. As described above, in the above-described embodiment, the binarization determination value is preset. However, the feature amount of the feature matrix may be overlooked, and the binarization determination value capable of appropriately determining whether or not it is the lane boundary line may be dynamically determined.

  Further, in the above-described embodiment, the lane boundary line is estimated from the coordinate point group data of the target road specified by the target road specifying information, but the data storage is performed without performing the process of causing the user to select the target road. The lane boundary line may be estimated from all coordinate point group data stored in the unit 13.

  The function of one component in each of the above embodiments may be shared by a plurality of components, or the function of a plurality of components may be performed by one component. In addition, part of the configuration of each of the above embodiments may be omitted as long as the problem can be solved. Further, at least a part of the configuration of each of the above-described embodiments may be added to or replaced with the configuration of the other above-described embodiments. It should be noted that all aspects included in the technical idea specified by the wording recited in the claims are embodiments of the present disclosure.

  In addition to the boundary estimation apparatus 1 described above, a system including the boundary estimation apparatus 1 as a component, a program for causing a computer to function as the boundary estimation apparatus 1, and a non-transitional actual state of a semiconductor memory or the like in which the program is recorded. The present disclosure can be realized in various forms such as a dynamic recording medium and a boundary estimation method.

  DESCRIPTION OF SYMBOLS 1 ... Boundary estimation device, 15 ... Control part, 20 ... Boundary estimation program

Claims (6)

A boundary estimation device for estimating a lane boundary line from data of a coordinate point group indicating a surface of a target road by a set of coordinate points represented by three-dimensional coordinates,
The coordinate point data is provided with boundary line determination information for determining whether or not the lane boundary line is arranged at a point corresponding to the coordinate point,
The surface of the target road is divided by a plurality of extension direction division lines along the road extension direction in which the target road extends and a plurality of vertical direction division lines along a direction perpendicular to the road extension direction. An area setting unit configured to set each of the plurality of divided areas as a divided area,
For each of the plurality of divided areas set by the area setting unit, based on the boundary line determination information added to the data of the coordinate points included in the divided area, the lane boundary line in the divided area. A boundary determination unit configured to determine whether or not is arranged,
The divided area judged by the boundary judgment unit that the lane boundary line is arranged is an effective divided area, and a direction perpendicular to the road extension direction is a row direction, and one row along the row direction. The divided area set is defined as a divided area column, and the effective divided area set in which the effective divided areas are continuously arranged in the row direction is a single row for each of the plurality of divided area rows. A column cluster setting section configured to be set as a cluster,
For each of the plurality of divided area columns, the column clusters included in the divided area column are selected, and then the selected column cluster that is the selected column cluster and the divided area column adjacent to the selected column cluster are selected. It is determined whether or not a preset cluster combination condition that indicates that the adjacent column cluster that is the included column cluster can be combined as a single lane boundary line is satisfied, and that the cluster combination condition is satisfied. In this case, the selected column cluster and the adjacent column cluster are combined, and a set of the column clusters in which the column clusters are continuously arranged along the road extension direction is set as one row cluster. And a line cluster setting unit configured as described above. The boundary estimation apparatus according to claim 1, wherein
The plurality of row clusters are located on the same road extension line extending along the road extension direction, and the plurality of row clusters are separated along the road extension line by a distance less than a preset combination determination distance. Boundary arranging unit configured to combine a plurality of the row clusters as one lane boundary line when the plurality of row clusters are arranged. The boundary estimation device according to claim 1 or 2, wherein
Boundary estimation provided with a column cluster exclusion unit configured to exclude the column cluster whose length along the direction perpendicular to the road extension direction does not match a preset boundary line determination width apparatus. The boundary estimation device according to any one of claims 1 to 3,
For each of the plurality of column clusters forming the row cluster, a center position in a direction perpendicular to the road extension direction is calculated, and the calculated plurality of center positions are connected along the road extension direction. Boundary estimation apparatus comprising a continuous line setting unit configured to set continuous lines. A boundary estimation method for estimating a lane boundary line from data of a coordinate point group indicating a surface of a target road by a set of coordinate points represented by three-dimensional coordinates,
The coordinate point data is provided with boundary line determination information for determining whether or not the lane boundary line is arranged at a point corresponding to the coordinate point,
The surface of the target road is divided by a plurality of extension direction division lines along the road extension direction in which the target road extends and a plurality of vertical direction division lines along a direction perpendicular to the road extension direction. , An area setting procedure for setting each of a plurality of divided areas as a divided area,
For each of the plurality of divided areas set by the area setting procedure, based on the boundary line determination information added to the data of the coordinate points included in the divided area, the lane boundary line in the divided area. A boundary determination procedure for determining whether or not is placed,
The divided area determined by the boundary determination procedure as the lane boundary line being arranged is an effective divided area, and a direction perpendicular to the road extension direction is a row direction, and one row along the row direction The divided area set is a divided area column, and each of the divided area rows is a set of the effective divided areas in which the effective divided areas are continuously arranged in the row direction. Column cluster setting procedure to set as a cluster,
For each of the plurality of divided area columns, select the column cluster included in the divided area column, and then select the selected column cluster that is the selected column cluster and the divided area column adjacent to the selected column cluster. It is determined whether or not a preset cluster combination condition that indicates that the included column cluster and an adjacent column cluster can be combined as a single lane boundary line is satisfied, and it is determined that the cluster combination condition is satisfied. In this case, the selected column cluster and the adjacent column cluster are combined, and a set of the column clusters in which the column clusters are continuously arranged along the road extension direction is set as one row cluster. A boundary estimation method comprising a row cluster setting procedure. A boundary estimation program for estimating a lane boundary line from data of a coordinate point group indicating the surface of a target road by a set of coordinate points represented by three-dimensional coordinates,
The coordinate point data is provided with boundary line determination information for determining whether or not the lane boundary line is arranged at a point corresponding to the coordinate point,
Computer,
The surface of the target road is divided by a plurality of extension direction division lines along the road extension direction in which the target road extends and a plurality of vertical direction division lines along a direction perpendicular to the road extension direction. An area setting unit configured to set each of the divided areas as a divided area,
For each of the plurality of divided areas set by the area setting unit, based on the boundary line determination information added to the data of the coordinate points included in the divided area, the lane boundary line in the divided area. A boundary determination unit configured to determine whether or not
The divided area determined by the boundary determining unit as the lane boundary line being arranged is an effective divided area, and a direction perpendicular to the road extension direction is a row direction, and one row along the row direction The divided area set is a divided area column, and each of the divided area rows is a set of the effective divided areas in which the effective divided areas are continuously arranged in the row direction. A column cluster setting unit configured to be set as a cluster, and
For each of the plurality of divided area columns, select the column cluster included in the divided area column, and then select the selected column cluster that is the selected column cluster and the divided area column adjacent to the selected column cluster. It is determined whether or not a preset cluster combination condition that indicates that the included column cluster and an adjacent column cluster can be combined as a single lane boundary line is satisfied, and it is determined that the cluster combination condition is satisfied. In this case, the selected column cluster and the adjacent column cluster are combined, and a set of the column clusters in which the column clusters are continuously arranged along the road extension direction is set as one row cluster. A boundary estimation program that functions as a line cluster setting unit configured as above.