Feature data generation system, feature database update system, and feature data generation method
In a feature data generation system, an edge pattern indicating a boundary of a target feature existing around a vehicle is extracted from position data of a large number of measurement points surrounding the vehicle measured using a LIDAR technology, on the basis of image data of surroundings of the vehicle obtained by imaging so as to reduce the amount of information, positions in a terrestrial reference frame are assigned to the extracted edge pattern, and a shape characteristic vector and a feature characteristic vector to be used for generating or updating a feature in a feature database, are generated from the edge pattern to which the positions in the terrestrial reference frame are assigned.
1 . A feature data generation method comprising:
an edge pattern extraction step of extracting an edge pattern indicating a boundary of a target feature existing around a moving body from position data of a large number of measurement points around the moving body measured using a LIDAR technology to reduce the amount of information;
a terrestrial reference frame position assignment step of assigning positions in a terrestrial reference frame to the extracted edge pattern;
a feature data generation step of generating feature data representing a feature from the edge pattern to which the positions in the terrestrial reference frame are assigned;
a shape characteristic vector setting step of, if an aggregate of the position data in the edge pattern to which the positions in the terrestrial reference frame are assigned matches data prepared in advance for describing an existing feature at a predetermined ratio or more, setting the aggregate of the position data as a shape characteristic vector to be used for generating or updating the feature data;
a shape characteristic vector acquiring step of acquiring a plurality of the shape characteristic vectors from position data of surroundings of a plurality of the moving bodies, through the extraction of the edge pattern, the assignment of the positions in the terrestrial reference frame to the edge pattern, and the setting of the shape characteristic vector; and
a feature characteristic vector generating step of performing statistical processing on the plurality of shape characteristic vectors to generate a feature characteristic vector for generating or updating the feature data.
2 . The feature data generation method according to claim 1 , further comprising a removing step of detecting and removing an anomalous measurement value from the shape characteristic vector on the basis of a Mahalanobis' generalized distance.
3 . The feature data generation method according to claim 1 , further comprising a determining step of calculating at least one statistical value in the shape characteristic vector and determining whether the shape characteristic vector includes a temporal change on the basis of the calculated statistical value.
4 . The feature data generation method according to claim 1 , further comprising a crustal movement correction step of performing crustal movement correction on the feature characteristic vector in order to adjust position information included in the feature characteristic vector to a geodetic reference system of a region where the moving bodies exist.
5 . The feature data generation method according to claim 1 , further comprising a crustal movement correction step of performing crustal movement correction in advance on the edge pattern to which the positions in the terrestrial reference frame are assigned in order to adjust position information included in the feature characteristic vector to a geodetic reference system of a region where the moving bodies exists.
6 . The feature data generation method according to claim 1 , further comprising an index calculation step of calculating an index for quality assurance of the feature characteristic vector on the basis of at least one of a time when the feature characteristic vector was generated and a statistical value of the feature characteristic vector.
7 . The feature data generation method according to claim 6 , further comprising a position data re-measurement request step of requesting re-measurement of position data by the LIDAR technology in an area corresponding to the feature characteristic vector on the basis of the index for quality assurance.
8 . The feature data generation method according to claim 1 , further comprising a position data measurement step of measuring a large number of pieces of position data intensively in an area corresponding to a feature included in a standard accuracy map using the LIDAR technology.
9 . The feature data generation method according to claim 1 , further comprising a position data measurement step of measuring a large number of pieces of position data intensively in an area corresponding to a feature recognized by the image data of surroundings of the moving body using the LIDAR technology.
10 . The feature data generation method according to claim 8 , wherein the large number of pieces of position data are obtained by a LIDAR scanner having a laser scanner to which a laser light irradiation direction varying technology is applied in the position data measurement step.
11 . The feature data generation method according to claim 1 , further comprising an edge pattern extraction step of extracting the edge pattern on the basis of the image data of surroundings of the moving body obtained by imaging, from the position data of the large number of measurement points around the moving body measured using the LIDAR technology.
12 . A feature data generation method comprising:
an edge pattern extraction step of extracting an edge pattern indicating a boundary of a target feature existing around a moving body from position data of a large number of measurement points around the moving body measured using a LIDAR technology to reduce the amount of information;
a terrestrial reference frame position assignment step of assigning positions in a terrestrial reference frame to the extracted edge pattern;
a feature data generation step of generating feature data representing a feature from the edge pattern to which the positions in the terrestrial reference frame are assigned;
a shape characteristic vector setting step of, if an aggregate of the position data in the edge pattern to which the positions in the terrestrial reference frame are assigned matches data prepared in advance for describing an existing feature at a predetermined ratio or more, setting the aggregate of the position data as a shape characteristic vector to be used for generating or updating the feature data;
a shape characteristic vector acquiring step of acquiring a plurality of the shape characteristic vectors by repeating the extraction of the edge pattern, the assignment of the positions in the terrestrial reference frame to the edge pattern, and the setting of the shape characteristic vector; and
a feature characteristic vector generating step of performing statistical processing on the plurality of shape characteristic vectors to generate a feature characteristic vector for generating or updating the feature data.