Road boundary detection device, road boundary detection method, and road boundary detection program

1 . A road boundary detection device that acquires a set of lines corresponding to a road boundary from point cloud data as road boundary information, the road boundary detection device comprising:

a candidate point detection unit configured to detect each point of road boundary candidates corresponding to candidates of a road boundary from the point cloud data;

a candidate point clustering unit configured to cluster each point of the road boundary candidates;

an adjacent cluster reduction unit configured to reduce a cluster from a distribution of points in clusters in an adjacency relationship by using a predetermined cluster reduction method;

a line fitting unit configured to fit one or more straight lines or curved lines to one or more of the clusters and output fitted lines as road boundary candidates, wherein the fitted lines for the one or more straight lines or the curved lines are determined by calculating a weight value based on a principal component vector formed by line segments connecting the points in clusters;

update a score of the one or more of the clusters based on the calculated weight;

a line connecting unit configured to connect some of the fitted lines by using a predetermined analysis method; and

an information output unit configured to output a calculated line as the road boundary information.

2 . A road boundary detection device that acquires a set of lines corresponding to a road boundary from point cloud data as road boundary information, the road boundary detection device comprising:

a point cloud density reduction unit configured to down-sample the point cloud data to reduce density;

a candidate point detection unit configured to detect each point of road boundary candidates corresponding to candidates of a road boundary from the reduced point cloud data;

a candidate point clustering unit configured to cluster each point of the road boundary candidates;

an adjacent cluster reduction unit configured to reduce a cluster from a distribution of points in clusters in an adjacency relationship by using a predetermined cluster reduction method;

a line fitting unit configured to fit one or more straight lines or curved lines to one or more of the clusters and output fitted lines as road boundary candidates, wherein the fitted lines for the one or more straight lines or the curved lines are determined by calculating a weight value based on a principal component vector formed by line segments connecting the points in clusters;

update a score of the one or more of the clusters based on the calculated weight;

a line connecting unit configured to connect some of the fitted lines by using a predetermined analysis method; and

an information output unit configured to output a calculated line as the road boundary information.

3 . The road boundary detection device according to claim 2 , wherein the point cloud density reduction unit deletes a point of interest in a case where a peripheral point density of each point of the point cloud data after the down-sampling is equal to or less than a threshold value, and excludes the deleted point from processing targets.

4 . The road boundary detection device according to claim 1 , wherein the candidate point clustering unit clusters a set of points arranged in parallel as a different cluster by calculating a principal component vector of a point of interest from a positional relationship between each point of the road boundary candidates and a neighboring point and performing clustering on the basis of an angle formed by the principal component vector and a line segment connecting the point of interest and a peripheral point.

5 . The road boundary detection device according to claim 1 , wherein, in the cluster reduction method, the adjacent cluster reduction unit calculates a principal component vector of a cluster of interest from a set of points included in the cluster, determines two end points from a direction of the principal component vector of the cluster of interest, determines whether or not the clusters are adjacent clusters on the basis of whether or not a foot of a perpendicular line drawn from a midpoint of a peripheral cluster to a straight line connecting the two end points of the cluster of interest is included between the two end points, and

determines a cluster to be reduced on the basis of a score calculated from a length and a width of the cluster between the adjacent clusters.

6 . The road boundary detection device according to claim 1 , wherein the line fitting unit clusters a position trajectory of a measuring device when the point cloud data is measured on the basis of a traveling direction, and determines whether to perform straight line fitting or curved line fitting on a subset of point clouds on the basis of a position of a cluster switching point.

7 . The road boundary detection device according to claim 1 , wherein, in the analysis method, the line connecting unit determines continuity of the fitted line and a line existing in a periphery of the fitted line on the basis of any one of a distance between end points, a difference in a direction of the line at the end points, intersection of the lines, a length of a perpendicular line drawn from an end point of one line to another line, and a length of the line, or a plurality of conditions, and determines whether or not to connect the fitted line and the line existing in the periphery of the fitted line.

8 . The road boundary detection device according to claim 1 , wherein the line connecting unit excludes a short line having a predetermined length or less among the connected lines from the road boundary candidates.

9 . A road boundary detection method for acquiring, by a computer, a set of lines corresponding to a road boundary from point cloud data as road boundary information, the road boundary detection method causing the computer to execute processing of:

detecting each point of road boundary candidates corresponding to candidates of a road boundary from the point cloud data;

clustering each point of the road boundary candidates;

reducing a cluster from a distribution of points in clusters in an adjacency relationship by using a predetermined cluster reduction method;

fitting one or more straight lines or curved lines to one or more of the clusters and outputting fitted lines, wherein the fitted lines for the one or more straight lines or the curved lines are determined by calculating a weight value based on a principal component vector formed by line segments connecting the points in clusters;

updating a score of the one or more of the clusters based on the calculated weight;

connecting some of the fitted lines; and

outputting a calculated line as the road boundary information.

10 . A road boundary detection program for acquiring a set of lines corresponding to a road boundary from point cloud data as road boundary information, the road boundary detection program causing a computer to execute processing of:

detecting each point of road boundary candidates corresponding to candidates of a road boundary from the point cloud data;

clustering each point of the road boundary candidates;

reducing a cluster from a distribution of points in clusters in an adjacency relationship by using a predetermined cluster reduction method;

fitting one or more straight lines or curved lines to one or more of the clusters and outputting fitted lines, wherein the fitted lines for the one or more straight lines or the curved lines are determined by calculating a weight value based on a principal component vector formed by line segments connecting the points in clusters;

updating a score of the one or more of the clusters based on the calculated weight;

connecting some of the fitted lines; and

outputting a calculated line as the road boundary information.

11 . The road boundary detection method according to claim 9 , wherein a point of interest is deleted in a case where a peripheral point density of each point of the point cloud data after the down-sampling is equal to or less than a threshold value, and excludes the deleted point from processing targets.

12 . The road boundary detection method according to claim 9 , wherein a set of points is clustered and arranged in parallel as a different cluster by calculating a principal component vector of a point of interest from a positional relationship between each point of the road boundary candidates and a neighboring point and performing clustering on the basis of an angle formed by the principal component vector and a line segment connecting the point of interest and a peripheral point.

13 . The road boundary detection method according to claim 9 , wherein, in the cluster reduction method, a principal component vector of a cluster of interest is calculated from a set of points included in the cluster, determines two end points from a direction of the principal component vector of the cluster of interest, determines whether or not the clusters are adjacent clusters on the basis of whether or not a foot of a perpendicular line drawn from a midpoint of a peripheral cluster to a straight line connecting the two end points of the cluster of interest is included between the two end points, and determines a cluster to be reduced on the basis of a score calculated from a length and a width of the cluster between the adjacent clusters.

14 . The road boundary detection method according to claim 9 , wherein a position trajectory is clustered when the point cloud data is measured on the basis of a traveling direction, and determines whether to perform straight line fitting or curved line fitting on a subset of point clouds on the basis of a position of a cluster switching point.

15 . The road boundary detection method according to claim 9 , wherein, in the analysis method, continuity of the fitted line and a line existing in a periphery of the fitted line are determined on the basis of any one of a distance between end points, a difference in a direction of the line at the end points, intersection of the lines, a length of a perpendicular line drawn from an end point of one line to another line, and a length of the line, or a plurality of conditions, and determines whether or not to connect the fitted line and the line existing in the periphery of the fitted line.

16 . The road boundary detection device according to claim 9 , wherein a short line having a predetermined length or less among the connected lines is excluded from the road boundary candidates.

17 . The road boundary detection program according to claim 10 , wherein a point of interest is deleted in a case where a peripheral point density of each point of the point cloud data after the down-sampling is equal to or less than a threshold value, and excludes the deleted point from processing targets.

18 . The road boundary detection program according to claim 10 , wherein a set of points is clustered and arranged in parallel as a different cluster by calculating a principal component vector of a point of interest from a positional relationship between each point of the road boundary candidates and a neighboring point and performing clustering on the basis of an angle formed by the principal component vector and a line segment connecting the point of interest and a peripheral point.

19 . The road boundary detection program according to claim 10 , wherein, in the cluster reduction method, a principal component vector of a cluster of interest is calculated from a set of points included in the cluster, determines two end points from a direction of the principal component vector of the cluster of interest, determines whether or not the clusters are adjacent clusters on the basis of whether or not a foot of a perpendicular line drawn from a midpoint of a peripheral cluster to a straight line connecting the two end points of the cluster of interest is included between the two end points, and determines a cluster to be reduced on the basis of a score calculated from a length and a width of the cluster between the adjacent clusters.

20 . The road boundary detection program according to claim 10 , wherein a position trajectory is clustered when the point cloud data is measured on the basis of a traveling direction, and determines whether to perform straight line fitting or curved line fitting on a subset of point clouds on the basis of a position of a cluster switching point.