JP2007122665A - System and method for recognizing road paint and method of creating road paint database - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To recognize road paint with accuracy by appropriately extracting image patterns that form the same road paint from the original road image information. <P>SOLUTION: A system for recognizing road paint includes: an image information acquisition means 6 for acquiring road image information including images of road paint; a contour line information extracting means 7 for extracting information on the contour lines of areas that are likely to be the road paint included in the road image information acquired; a polygon dividing means 8 for dividing a closed area included in the contour line information into a plurality of projecting area polygons for computing main axes; a main axis computing means 16 for computing the main axis on which the cross-sectional secondary moment of each of the projecting area polygons is at its minimum; and an interconnecting means 11 by which the plurality of projecting area polygons joined to one another via sides crossing the main axis, with differences in the direction of the main axis being within a predetermined tolerance, are interconnected as an assembly constituting the same road paint. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a road paint recognition system, a road paint recognition method, and a road paint database creation method for performing processing for road paint recognition based on road image information including road paint images.

  As a technique for determining the type of road paint based on image information when generating an electronic map database, for example, the following patent document 1 discloses an electronic map data generation system as described below. The configuration of this system is shown in FIG. This system has a map data generation apparatus 101 configured by installing a predetermined program in a general-purpose computer as a core, a road database 102 defined by a link as a line segment and a node that is an intersection of the link, an aerial photograph 103, the attribute information is set for the link of the road database 102, and a new map database is generated.

  The map data generation apparatus 101 includes a command input unit 104, a database reference unit 105, a polygon generation unit 106, a pattern database 107, a pattern analysis unit 108, an attribute setting unit 109, and a result output unit 110. The command input unit 104 inputs a command from an operator through operations such as a keyboard and a mouse. The database reference unit 105 refers to the road database 102 and the aerial photograph 103 based on the input command. The polygon generation unit 106 generates a closed figure having a predetermined area as a processing target area based on each link defined in the road database 102. The pattern database 107 is a database that stores various image patterns included in the aerial photograph 103 and their attributes in association with each other. The pattern analysis unit 108 reads the aerial photograph 103 and analyzes the image pattern included therein based on the pattern database 107. The attribute setting unit 109 sets attribute information for the link of the road database 102 based on the analysis result of the pattern analysis unit 108. The result output unit 110 receives the setting result by the attribute setting unit 109, associates attribute information with each link, and outputs a new map database.

  Therefore, in this system, in order to determine the type of road paint, the pattern analysis unit 108 performs pattern matching between the image pattern included in the image information such as the aerial photograph 103 and the image pattern stored in the pattern database 107. Will do. That is, in the pattern database 107, for example, a pattern in which short line segments constituting a car frame are arranged at regular intervals like a parking lot and an attribute “parking lot” are associated with each other, or a center line or the like A yellow line segment and attribute information “prohibition of protrusion” are stored in association with each other. Then, the pattern analysis unit 108 analyzes the image pattern included in the image information such as the aerial photograph 103, and determines whether the image pattern matches the image pattern stored in the pattern database 107. The road paint type such as is determined.

JP 2003-195747 A (pages 5-6, 10-11, FIG. 3, FIG. 17)

  By the way, in order to actually recognize road paint by pattern matching or the like, it is important to appropriately extract image patterns that are likely to constitute road paint from image information such as the original aerial photograph 103. It becomes. In addition, since the image patterns extracted from the image information include image patterns for a plurality of road paints without being distinguished from each other, a unit that is a pattern matching recognition target among these image patterns. It is important to select appropriately. Specifically, it is necessary to select an image pattern considered to constitute the same road paint and use it as a unit of recognition.

  However, Patent Document 1 discloses a method for appropriately extracting an image pattern that is likely to constitute road paint, and a method for appropriately selecting a unit to be recognized by pattern matching among the extracted image patterns. Is not specifically disclosed. Therefore, an image pattern including an image other than road paint is extracted, or an image pattern in a state where a plurality of road paints are not separated is set as a recognition target. As a result, the number of image patterns to be stored in the pattern database 107 is inevitably large, and there is a problem that it is difficult to accurately recognize road paint. Further, in order to accurately recognize the road paint while avoiding storing such a large number of image patterns in the pattern database 107, it is necessary to perform pattern matching manually.

  The present invention has been made in view of the above problems, and its purpose is to accurately recognize road paint by appropriately extracting image patterns constituting the same road paint from the original road image information. A road paint recognition system and a road paint recognition method, and a road paint database creation method using the method.

  In order to achieve the above object, the road paint recognition system according to the present invention is characterized by image information acquisition means for acquiring road image information including a road paint image, and roads included in the acquired road image information. Outline information extracting means for extracting outline information of a paintable area, polygon dividing means for dividing a closed area included in the outline information into a plurality of area convex polygons for spindle calculation, For a region convex polygon, a plurality of main axis calculation means for calculating a main axis that minimizes the moment of inertia of the cross section, and a plurality of articulations that are within a predetermined allowable range in the direction of the main axis and that are connected via sides intersecting the main axis The above-mentioned area convex polygon is combined with a combination means for combining the region convex polygons as an aggregate constituting the same road paint.

  According to this characteristic configuration, among the plurality of area convex polygons obtained by dividing the closed area included in the contour line information extracted from the road image information, the longitudinal direction is substantially the same direction, and the area is directly or other area. A plurality of area convex polygons arranged so as to be in contact with each other via the convex polygons can be combined along the longitudinal direction thereof to generate an aggregate of area convex polygons. Therefore, it is possible to combine the region convex polygons constituting the road paint formed in a single linear band as one aggregate. Therefore, by using this aggregate as a recognition unit for road paint recognition processing, accurate road paint recognition is possible.

  Here, road network information acquisition means for acquiring information on a road network having a plurality of nodes having position information and a link connecting two nodes, and a centroid operation for calculating a centroid for each of the convex polygons in each region Based on the relationship between the means, the arrangement of the links included in the road network information, the direction of the principal axis of the region convex polygon and the position of the centroid, one or more candidates for each region convex polygon A preliminary determination unit that determines a road paint type and assigns candidate type information indicating a determination result, and the combination unit uses the same plurality of area convex polygons to which the common candidate type information is added. It is preferable that the road paint is combined as an aggregate constituting the road paint.

  According to this configuration, one or more roads that are candidates for each region convex polygon based on the relationship between the arrangement of links included in the road network information and the direction of the principal axis of the region convex polygon and the position of the centroid. The paint type can be determined. Only a plurality of area convex polygons to which common candidate type information is assigned are combined as an aggregate, and area convex polygons that are likely to be different road paint types can be prevented from being combined as an aggregate. Therefore, it is possible to more appropriately extract only region convex polygons that are likely to constitute the same road paint and combine them as an aggregate. As a result, more accurate road paint recognition is possible.

  A further characteristic configuration of the road paint recognition system according to the present invention includes image information acquisition means for acquiring road image information including a road paint image, a plurality of nodes having position information, and a link connecting the two nodes. Road network information acquisition means for acquiring road network information, outline information extraction means for extracting outline information of an area that may be road paint included in the acquired road image information, and the outline Polygon dividing means for dividing the closed area included in the line information into a plurality of area convex polygons for main axis calculation; for each area convex polygon, main axis calculation means for calculating a main axis that minimizes the sectional moment of inertia; and For each region convex polygon, centroid calculating means for calculating a centroid, the arrangement of links included in the road network information, and Based on the relationship between the direction of the principal axis of the region convex polygon and the position of the centroid, one or more candidate road paint types are determined for each region convex polygon, and candidate type information indicating the determination result is given. Pre-determining means for combining, and combining means for combining the plurality of area convex polygons, which are given the common candidate type information and are connected to form one closed area, as an aggregate constituting the same road paint, It is in the point to prepare.

  According to this configuration, one or more roads that are candidates for each region convex polygon based on the relationship between the arrangement of links included in the road network information and the direction of the principal axis of the region convex polygon and the position of the centroid. The paint type can be determined. Then, among the plurality of area convex polygons obtained by dividing the closed area included in the contour line information extracted from the road image information, only the plurality of area convex polygons to which common candidate type information is assigned are combined as an aggregate. can do. In other words, it is possible to prevent area convex polygons that are likely to be of different road paint types from being combined as an aggregate, and more appropriately extract only the area convex polygons that are likely to constitute the same road paint. Can be combined as Therefore, by using this aggregate as a recognition unit for road paint recognition processing, accurate road paint recognition is possible.

  Here, the preliminary determination means is configured to determine each region convexity based on one or more of an angle between a principal axis of the region convex polygon and the link and a distance between a centroid of the region convex polygon and the link. It is preferable that the polygon is configured to determine one or more candidate road paint types.

  Here, the link included in the road network information is arranged near the center in the width direction of the road according to the road shape. Further, for each road paint type, the feature amount of the angle and distance with respect to the link of the road paint is determined in advance. Further, the direction of the main axis of the region convex polygon substantially coincides with the longitudinal direction of the road paint to which the region convex polygon belongs. That is, the angle between the main axis of the region convex polygon and the link is highly likely to be substantially equal to the angle with respect to the link of the road paint to which each region convex polygon belongs. In addition, the distance between the centroid of the area convex polygon and the link is highly likely to be substantially equal to the distance between the road paint to which each area convex polygon belongs and the link. Therefore, according to this configuration, one or two or more candidate road paint types for each region convex polygon are determined to a certain degree of accuracy for each region convex polygon based on a predetermined angle and distance feature with respect to the link for each road paint type. This makes it possible to make a determination.

  Here, based on aggregate information including shape and arrangement information about the aggregate of the region convex polygons coupled by the coupling means, or based on the aggregate information and the road network information, each of the aggregates It is preferable that the apparatus further includes a determination unit that determines the road paint type of the body.

  According to this configuration, an aggregate formed by combining region convex polygons constituting a road paint formed in a single linear strip is used as a recognition unit for road paint recognition processing, and the aggregate information and the road network Based on the information, it is possible to appropriately determine the road paint type of each aggregate. Therefore, it is possible to accurately recognize the road paint type for the road paint image included in the road image information.

  The characteristic configuration of the road paint recognition method according to the present invention includes an image information acquisition step of acquiring road image information including a road paint image, and an area having a possibility of road paint included in the acquired road image information. A contour information extracting step for extracting the contour information of the region, a polygon dividing step for dividing the closed region included in the contour line information into a plurality of convex polygons for principal axis calculation, A spindle calculation step for calculating a spindle that minimizes the next moment, and a plurality of the region convex polygons that are connected to each other via a side that intersects the spindle with a difference in direction of the spindle within a predetermined allowable range. And a coupling step for coupling as a collection constituting the same road paint.

  According to this characteristic configuration, among the plurality of area convex polygons obtained by dividing the closed area included in the contour line information extracted from the road image information, the longitudinal direction is substantially the same direction, and the area is directly or other area. A plurality of area convex polygons arranged so as to be in contact with each other via the convex polygons can be combined along the longitudinal direction thereof to generate an aggregate of area convex polygons. Therefore, it is possible to combine the region convex polygons constituting the road paint formed in a single linear band as one aggregate. Therefore, by using this aggregate as a recognition unit for road paint recognition processing, accurate road paint recognition is possible.

  A further characteristic configuration of the road paint recognition method according to the present invention includes an image information acquisition step for acquiring road image information including a road paint image, a plurality of nodes having position information, and a link connecting the two nodes. A road network information acquisition step for acquiring road network information, a contour line information extraction step for extracting contour line information of an area with a possibility of road paint included in the acquired road image information, and the contour A polygon dividing step for dividing the closed region included in the line information into a plurality of region convex polygons for main axis calculation, a main axis calculating step for calculating a main axis that minimizes the sectional moment of inertia for each region convex polygon, A centroid calculation step for calculating a centroid for each region convex polygon and included in the road network information One or more candidate road paint types are determined for each region convex polygon based on the relationship between the position of the link and the direction of the principal axis of the region convex polygon and the position of the centroid. Preliminary determination step for assigning candidate type information to be represented, and combining the plurality of region convex polygons that are connected to form one closed region by being shared with the common candidate type information as an aggregate constituting the same road paint And a coupling step.

  According to this configuration, one or more roads that are candidates for each region convex polygon based on the relationship between the arrangement of links included in the road network information and the direction of the principal axis of the region convex polygon and the position of the centroid. The paint type can be determined. Then, among the plurality of area convex polygons obtained by dividing the closed area included in the contour line information extracted from the road image information, only the plurality of area convex polygons to which common candidate type information is assigned are combined as an aggregate. can do. In other words, it is possible to prevent area convex polygons that are likely to be of different road paint types from being combined as an aggregate, and more appropriately extract only the area convex polygons that are likely to constitute the same road paint. Can be combined as Therefore, by using this aggregate as a recognition unit for road paint recognition processing, accurate road paint recognition is possible.

  The characteristic configuration of the road paint database creation method according to the present invention includes an image information acquisition step of acquiring road image information including a road paint image, and a possibility of road paint included in the acquired road image information. Contour line information extracting step for extracting contour information of a certain region, polygon dividing step for dividing a closed region included in the contour line information into a plurality of region convex polygons for main axis calculation, and each region convex polygon, A main axis calculation step for calculating a main axis that minimizes the moment of inertia of the cross section, and a plurality of the convex polygons connected to each other through sides intersecting with the main axis, the difference in direction of the main axis being within a predetermined allowable range Are connected as an aggregate constituting the same road paint, and a plurality of nodes having position information and two nodes are connected. Based on aggregate information including a road network information acquisition step for acquiring information on a road network having links, and information on the shape and arrangement of the aggregate of the region convex polygons combined by the combining step, or the aggregate Based on the body information and the road network information, a determination step for determining a road paint type of each aggregate, and the aggregate information for which the road paint type has been determined are stored in a database as road paint information of the road paint type. And a road paint information registration step for registration.

  According to this characteristic configuration, among the plurality of area convex polygons obtained by dividing the closed area included in the contour line information extracted from the road image information, the longitudinal direction is substantially the same direction, and the area is directly or other area. A plurality of area convex polygons arranged so as to be in contact with each other via the convex polygons can be combined along the longitudinal direction thereof to generate an aggregate of area convex polygons. Therefore, it is possible to combine the region convex polygons constituting the road paint formed in a single linear band as one aggregate. Then, this aggregate can be used as a recognition unit for road paint recognition processing, and the road paint type of each aggregate can be appropriately determined based on the aggregate information and the road network information. And since the aggregate | assembly information by which the road paint classification was determined is registered into a database as road paint information, the database provided with exact road paint information can be created.

  Embodiments of the present invention will be described below with reference to the drawings. The road paint recognition system according to the present embodiment is based on road image information including a road paint image, and among various road paints included in the road image information, in particular, the type of belt-like (linear) road paint. Is recognized and recognized as road paint information. Here, for simplification of description, the types of strip-shaped road paint to be determined are limited. Specifically, the solid line and broken line of the roadway center line, the solid line and broken line of the lane boundary line, the side line (roadway) Outside lines or roadside belts), stop lines, pedestrian crossings, and zebra zones are subject to judgment. However, the present invention can also be applied to determination of other road paint types.

  FIG. 1 is an explanatory diagram showing a schematic configuration of a road paint recognition system according to the present embodiment. As shown in this figure, the road paint recognition system will be described later on a road image information database 2, a road network database 3, a road paint information database 4, and a general-purpose computer 5 connected so as to be accessible thereto. And a system main body 1 configured by installing a program for realizing the function of each means. Here, the system main body 1 includes an image information acquisition unit 6, an outline information extraction unit 7, a polygon division unit 8, a preliminary feature amount extraction unit 9, a preliminary determination unit 10, a combination unit 11, a feature amount extraction unit 12, and a type determination. Means 13, road paint information generation means 14, centroid calculation means 15, main axis calculation means 16, road network information acquisition means 17, representative axis calculation means 18, and grouping means 19 are provided. The preliminary determination unit 10 includes a preliminary determination table 20, and the type determination unit 13 includes a type determination table 21. In the present embodiment, the type determination unit 13 corresponds to the “determination unit” in the present invention. Hereinafter, each configuration will be described in detail.

1. Road image information database 2
The road image information database 2 is a database in which road image information including road paint images is stored. Here, the road image information is raster data. The road image information is associated with coordinate information indicating the coordinates of the area appearing in the image. Such road image information includes, for example, aerial photographs, satellite photographs, actual images such as videos and still images captured by survey vehicles, as well as image information of drawings such as construction drawings, city plans, and housing maps. included.

2. Road network database 3
The road network database 3 is a database storing road network information. The road network includes a large number of nodes N having position information on absolute coordinates expressed by latitude and longitude, and a large number of links L connecting the two nodes N (see FIG. 5). . Normally, the node N is arranged at the center of the intersection, and the link L is arranged near the center in the width direction of the road according to the road shape. Each link L is associated with information such as road width, number of lanes, road type (type of highway, toll road, national road, prefectural road, etc.), link length, and the like.

3. Road paint information database 4
The road paint information database 4 is a database that stores road paint information as a recognition result of the road paint recognition system according to the present embodiment.

4). Image information acquisition means 6
The image information acquisition unit 6 acquires road image information from the road image information database 2 in accordance with an instruction input from the operation unit 5a. Here, for example, image information such as a single aerial photograph associated with the coordinate information is acquired.

5. Outline information extraction means 7
The contour line information extracting unit 7 extracts contour line information of a region that may be road painted, which is included in the acquired road image information. Here, since the road paint is mostly white or yellow, there is a high possibility that a region having a low color density in the road image information including aerial photographs corresponds to the road paint. Therefore, the contour line information extracting means 7 specifically generates binarized road image information to reveal a low density area, and then generates vector data representing the outline of the low density area. Perform raster / vector conversion processing. At this time, if the contour line of the low-density region is used as it is as vector data, fine irregularities due to the variation in dot units are reflected. Therefore, it is preferable to perform noise removal processing and bring the contour line close to a straight line. As a result, as shown in FIG. 2, the contour line information representing the contour line of the area where there is a possibility of road painting is extracted.

6). Polygon dividing means 8
As shown in FIG. 3, the polygon dividing means 8 divides the closed area included in the extracted contour information into a plurality of area convex polygons P for principal axis calculation. FIG. 3 is an enlarged view of a portion of FIG. Here, the region convex polygon P is a polygon in which the inner angles of all the vertices are less than 180 [°], that is, a convex polygon, and is a polygon that is subject to calculation of the main axis Mp by the main axis calculation means 16 described later. The area convex polygon P has a main axis Mp along the longitudinal direction of the road paint formed in a linear belt shape so that the number of divisions is as small as possible, that is, the area of each area convex polygon P is as large as possible. It is desirable to divide so that Therefore, the polygon dividing unit 8 performs polygon dividing processing according to the following dividing conditions, for example.

Region convex polygon division condition (1) Two polygons (closed region) having n vertices (n is a natural number of 3 or more) included in the contour line information extracted by the contour line information extracting means 7 A diagonal line connecting vertices is drawn and divided into n-2 triangles. At this time, the diagonal line can be, for example, n-3 diagonal lines connecting any one vertex and another vertex.
(2) When dividing (1) above, three adjacent vertices Vi, Vi + 1, Vi + 2 of the polygon are selected, and a triangle connecting the three vertices Vi, Vi + 1, Vi + 2 is outside the polygon. In the case of the above, the triangle connecting the three vertices Vi, Vi + 1, Vi + 2 is not adopted.
(3) When dividing (1) above, three adjacent vertices Vi, Vi + 1, Vi + 2 of the polygon are selected, and the polygon is placed inside a triangle connecting the three vertices Vi, Vi + 1, Vi + 2. If there are other vertices, the triangle connecting the three vertices Vi, Vi + 1, Vi + 2 is further divided. This subdivision can be performed, for example, by drawing a diagonal line that connects the other vertexes to the existing vertexes so as not to intersect the already drawn diagonal lines.
(4) After the division in (1) above, if a vertex with an internal angle of 180 [deg.] Or more does not appear after removal for each of the diagonal lines drawn for division, the diagonal line is removed. Join triangles. As a result, all the diagonal lines that can be removed are removed.
(5) The combined figure in the above (4) is defined as a region convex polygon P.

  By dividing under the above dividing conditions, the closed area (polygon) included in the contour line information can be divided into a plurality of area convex polygons P so that the number of divisions is substantially minimized. In addition, by dividing in this way, when the closed region related to the road paint formed in a linear belt shape is divided, the region is divided so that the convex polygons P having a long shape in the longitudinal direction of the road paint occupy a large number. can do. Therefore, as a result, each region convex polygon P has a main axis in a direction close to the longitudinal direction of the road paint. The polygon dividing unit 8 preferably divides the closed region included in the extracted outline information so that the number of divisions is minimum (the area of each region convex polygon P is maximum). It is not limited. That is, the polygon dividing means 8 may divide the closed region so that many of the divided region convex polygons P have the main axis Mp along the longitudinal direction of the road paint formed in a linear strip shape. What is necessary is just to divide | segment into the convex polygon which has a certain amount of size at the limit. In some cases, the closed region is not divided and one closed region is composed of one region convex polygon P. In this case, the one area convex polygon P is handled in the same manner as the aggregate A described later.

  As described above, the area convex polygon information including the shape and arrangement information of the area convex polygon P is acquired by the image information acquisition unit 6, the contour line information extraction unit 7, and the polygon dividing unit 8. That is, the image information acquisition unit 6, the contour line information extraction unit 7, and the polygon division unit 8 constitute a polygon information acquisition unit 22.

7). Centroid calculation means 15
The centroid calculating means 15 calculates the centroid Op for each region convex polygon P as shown in FIG. Further, as shown in FIG. 7, the centroid calculating means 15 also calculates the centroid Oa of the aggregate A formed by combining a plurality of region convex polygons P by the combining means 11 described later. As a method for calculating these centroids Op and Oa, known methods can be used.

8). Spindle calculation means 16
As shown in FIG. 4, the main axis calculation means 16 calculates a main axis Mp that minimizes the sectional moment of inertia for each region convex polygon P as an axis representing the longitudinal direction. Here, the main axis calculation means 16 calculates a unit vector representing one of the main axes Mp in the axial direction with the centroid Op as the origin. Here, there are two vectors having different 180 ° directions as unit vectors representing the axial direction of the main axis Mp, but here, they are unified to only one of them by limiting the effective angle range or the like. I am going to do that. A known method can be used as a method of calculating the main axis Mp.

9. Road network information acquisition means 17
The road network information acquisition means 17 acquires road network information having a link L connecting the plurality of nodes N and the two nodes N from the road network database 3. At this time, as shown in FIG. 5, the road network information acquisition unit 17 includes at least a node N and a link L in the area represented by the road image information acquired by the image information acquisition unit 6. Get network information. As a result, it is possible to obtain the arrangement relationship such as the distance between the region convex polygon P generated from the road image information, the later-described aggregate A, and the nodes N and links L of the road network.

10. Preliminary feature amount extraction means 9
The preliminary feature amount extraction unit 9 extracts a predetermined preliminary feature amount related to the region convex polygon P based on the region convex polygon information and the road network information. In the present embodiment, the preliminary feature quantity extraction unit 9 uses, as a preliminary feature quantity, the relationship between the arrangement of the link L and the direction of the principal axis Mp of the region convex polygon P and the position of the centroid Op, more specifically, For each region convex polygon P, (i) “angle between main axis Mp and link L” and (ii) “distance between centroid Op and link L” are extracted. As described above, the spindle Mp is calculated by the spindle calculating means 16, and the centroid Op is calculated by the centroid calculating means 15. Here, as the link L from which the preliminary feature quantity is extracted, the link L at the closest position on the road where the convex polygons P of the respective regions exist is employed. (I) “An angle between the main axis Mp and the link L” is calculated by regarding the link L as a vector in one direction. Further, (ii) “Distance between centroid Op and link L” calculates the shortest distance.

11. Preliminary determination means 10
The preliminary determination means 10 includes a preliminary determination table 20 in which a preliminary determination condition that defines a preliminary feature amount condition for each road paint type is stored. As the preliminary feature quantity used as the basis for the determination at this time, the one extracted by the preliminary feature quantity extraction unit 9 is used. Based on the preliminary determination condition, one or more candidate road paint types are determined for each region convex polygon P, and candidate type information indicating the determination result is given. Specifically, the candidate type information to be given here includes “roadway centerline solid line”, “roadway centerline broken line”, “lane boundary solid line”, “lane boundary broken line”, “side line”, “stop line” ”,“ Pedestrian crossing ”, and“ zebra zone ”. In the present application, the “side line” is used as a generic name for the roadway outer line and the roadside belt. Further, in the following description, the term “roadway centerline” includes both “roadway centerline solid line” and “roadway centerline dashed line”, and the term “lane boundary line” simply refers to “lane boundary line solid line” and It includes both “lane boundary broken lines”. The determination process by the preliminary determination means 10 will be described later in detail based on the flowchart shown in FIG. 10, and here, the preliminary determination conditions stored in the preliminary determination table 20 will be described.

  In this embodiment, (a) an angle condition between the main axis Mp and the link L and (b) a distance condition between the centroid Op and the link L are set as the preliminary determination conditions. (A) The conditions of the angle between the main axis Mp and the link L are “roadway center line”, “lane boundary line”, “side line”, “pedestrian crossing”, “stop line”, and “zebra zone”. This is a condition for distinguishing and adding candidate type information. This condition corresponds to the determination conditions of steps # 22 and # 24 in the flowchart of FIG. Specifically, (a) the condition of the angle between the main axis Mp and the link L is defined by the following three conditions: (i) “parallel”, (ii) “vertical”, (iii) “other” It is set. The “parallel” and “vertical” conditions should not be strict conditions but should be substantially parallel and substantially vertical conditions, and it is desirable to set a certain angle range. In this case, the angle range is that in the direction of the main axis Mp of the region convex polygon P obtained by dividing the contour information about the linear belt-like road paint arranged substantially parallel or substantially perpendicular to the link L as described above. It is preferable to experimentally obtain a range in which variation can occur and set a range wider than the range. Here, for example, ± 15 [°] or ± 10 [°] is set. As a result of setting the conditions (i) to (iii) as described above, the condition (i) indicates that the road paint satisfying the conditions is “roadway center line”, “lane boundary line”, “side line”, and “pedestrian crossing”. Included and excludes “stop lines” and “zebra zones”. On the other hand, condition (ii) includes “stop line” as a road paint satisfying this condition, and “roadway center line”, “lane boundary line”, “side line”, “pedestrian crossing”, and “zebra zone” Excluded. Therefore, “zebra zone” is included as road paint corresponding to the condition (iii).

  Further, as a preliminary determination condition, (b) the distance condition between the centroid Op and the link L distinguishes “roadway center line”, “lane boundary line”, and “side line” and assigns candidate type information. It is a condition for. This condition corresponds to the determination conditions of steps # 27 and # 29 in the flowchart of FIG. Specifically, this condition is: (i) “within ± 2 [m]” (region i in FIG. 5), (ii) “within ± 2 [m] from a position 1/2 of the road width” ( Three conditions of areas ii) and (iii) “other than that” (area iii in FIG. 5) are set. The numerical value “± 2 [m]” in these conditions is a numerical value set as a smaller value based on the fact that the width of one lane of the road is around 3 [m]. Only. That is, by setting the width in the direction orthogonal to the link L of the area related to each condition to be less than the width of one lane, the area related to each condition is the “roadway center line”, “side line”, and “lane boundary line”. Two or more of them are not included. Accordingly, other numerical values such as 1.5 [m] and 2.5 [m] can be set as long as the numerical values are less than the width of one lane. In addition, about distances, such as 2 [m], it determines by converting into real distance based on the reduced scale of road image information. As a result of setting the conditions (i) to (iii) in this way, the condition (i) includes “roadway center line” as a road paint satisfying the condition, and excludes “side line” and “lane boundary line”. The On the other hand, the condition (ii) includes “side line” as road paint satisfying this condition, and excludes “roadway center line” and “lane boundary line”. Therefore, the road paint corresponding to the condition (iii) includes “lane boundary line”. In this preliminary determination condition, “stop line”, “pedestrian crossing”, and “zebra zone” are not determined.

12 Coupling means 11
The combining means 11 combines the plurality of region convex polygons P as an aggregate A that constitutes the same road paint according to a predetermined combining condition. In the present embodiment, the combination conditions are as follows: (i) “difference in the direction of the main axis Mp is within a predetermined allowable range”, (ii) “sides intersecting the main axis Mp” Three conditions are set: “to be connected” and (iii) “common candidate type information is given”.

  Here, regarding the condition (i), the allowable range of the difference in the direction of the main axis Mp is the direction of the main axis Mp of the region convex polygon P obtained by dividing the contour information about the straight belt-like road paint as described above. It is preferable to experimentally obtain a range in which the variation of the above can occur and to set a range wider than the range. Here, for example, it is set to ± 15 [°]. By setting in this way, as a result, most of the principal axes Mp of the plurality of area convex polygons P obtained by dividing the contour information about one straight belt-like road paint are included in the permissible range. It will be. Regarding the condition (ii), the side that intersects with the main axis Mp is the side that intersects with the axis that extends the unit vector representing the main axis Mp calculated by the main axis calculating means 16 on both sides. Here, since the main axis Mp is an axis along the longitudinal direction of the region convex polygon P, the side intersecting with the main axis Mp is the short side of the region convex polygon P. Further, regarding condition (iii), when a plurality of candidate type information is assigned to the region convex polygon P, this condition (iii) is satisfied if at least one of them is common.

  The combination means 11 combines the plurality of region convex polygons P whose longitudinal directions are substantially the same depending on the condition (i). Further, according to the condition (ii), a plurality of area convex polygons P arranged so as to be in contact with each other directly or via another area convex polygon P can be combined along the longitudinal direction thereof. That is, the combination of the condition (i) and the condition (ii) can be combined to form a single closed region, and the region convex polygons P constituting the straight belt-like road paint can be suitably combined. . Furthermore, it is possible to prevent region convex polygons P that are likely to be of different road paint types from being combined according to the condition (iii). Therefore, by these conditions (i) to (iii), as shown in FIG. 6, common candidate type information is given, and a plurality of region convex polygons P constituting one linear belt-like road paint Can be combined to generate an aggregate A. As described above, when one closed region is composed of one region convex polygon P, the single region convex polygon P is handled in the same manner as the aggregate A. In the present embodiment, the aggregate A corresponds to the “polygon element” in the present invention.

  As described above, a plurality of region convex polygons P are assembled by the image information acquisition unit 6, the contour line information extraction unit 7, the polygon division unit 8, the preliminary feature amount extraction unit 9, the preliminary determination unit 10, and the combination unit 11. Polygon element information (aggregate information) including information on the shape and arrangement of the aggregate A as a polygon element is acquired. That is, the image information acquisition unit 6, the contour line information extraction unit 7, the polygon division unit 8, the preliminary feature quantity extraction unit 9, the preliminary determination unit 10, and the combination unit 11 constitute a polygon element information acquisition unit 23.

13. Representative axis calculation means 18
As shown in FIG. 7, the representative axis calculation means 18 calculates a representative axis Ma representing the longitudinal direction of each aggregate A. Here, the representative axis calculation means 18 uses the centroid Oa of the aggregate A calculated by the centroid calculation means 15 as the origin, and sets the principal axis Mp (unit vector) of each region convex polygon P constituting the aggregate A as the origin. A sum vector obtained by multiplying the area of the area convex polygon P by the area A is added to all the area convex polygons P constituting the aggregate A, and the axis represented by this vector is used as the representative axis Ma. In other words, the representative axis Ma is calculated as a vector representing a weighted average of the main axes Mp of the respective region convex polygons P with the area of each region convex polygon P as a weight. Here, there are two vectors having different 180 ° directions as vectors representing the axial direction of the representative axis Ma, but here, only one of them is unified by limiting an effective angle range or the like. I am going to do that.

14 Feature amount extraction means 12
The feature amount extraction unit 12 extracts a predetermined feature amount related to the aggregate A based on the aggregate information about the aggregate A or based on the aggregate information and the road network information. In the present embodiment, the feature quantity extraction unit 12 extracts individual feature quantities for each aggregate A, and also group feature quantities for each group including a plurality of aggregates A grouped by the grouping unit 19 described later. Also extract. Here, the feature quantity extraction means 12 includes (i) “area of the aggregate A”, (ii) “perimeter of the aggregate A”, and (iii) “representative axis Ma of the aggregate A as individual feature quantities. “Angle with link L”, (iv) “Distance between centroid Oa of aggregate A and link L” is extracted. Here, the link L which is the closest position on the road where each aggregate A exists is adopted as the link L from which the individual feature amount is extracted. (Iii) “An angle between the representative axis Ma of the aggregate A and the link L” is calculated by regarding the link L as a vector in one direction. Further, (iv) “Distance between centroid Oa of assembly A and link L” calculates the shortest distance.

  Further, the feature quantity extraction unit 12 uses (i) “an angle between the direction connecting the centroids Oa of the plurality of aggregates A in the group (centroid connection direction) and the link L” as the group feature quantity, (ii) “Distance between centroids Oa of multiple assemblies A in the group (centroid connection line) and link”, (iii) “Angle between representative axes Ma of each assembly A in the group” To do. Here, the link L which is the closest position on the road where the group of the aggregate A exists is adopted as the link L from which the group feature amount is extracted. Then, (i) “the angle between the direction of connecting the centroids Oa of the plurality of aggregates A in the group (centroid connection direction) and the link L”, the centroid connection direction is 3 for the aggregates A in the group. If there are more than one, the average value of the unit vectors representing the directions of the centroid connection lines connecting the centroids Oa is used. The angle is calculated by regarding the link L as a vector in one direction. In addition, (ii) “distance between the line (centroid connection line) connecting the centroids Oa of the plurality of aggregates A in the group and the link” calculates the average distance. When there are three or more assemblies A in the group, the centroid connection line may be a broken line. (Iii) “An angle between the representative axes Ma of the aggregates A in the group” calculates an angle formed by vectors representing the representative axes Ma of the aggregates A.

15. Grouping means 19
As shown in FIG. 8, the grouping means 19 groups a plurality of aggregates A that meet a predetermined grouping condition. The purpose of this grouping is to form a group of road paints by a plurality of aggregates A such as “roadway center line broken line”, “lane boundary line broken line”, “pedestrian crossing”, and “zebra zone”. The possibility is to make one group, and to extract the feature amount and determine the road paint type for each group. Here, as the grouping condition, specifically, each aggregate A is (i) “having the same area and circumference”, and (ii) “adjacently arranged at predetermined intervals” Two conditions are set. When grouping, when three or more assemblies A are sequentially arranged at a predetermined interval defined in the condition (ii), all these assemblies A are grouped together. .

  Here, regarding the condition (i), the range included in the “same area and circumference” is considered in consideration of the influence of noise other than road paint included in the road image information, the accuracy of contour extraction, etc. Set a range that allows some variation. This range is, for example, within ± 30% of the average value of the other aggregates A in area and circumference (the area and circumference of the aggregate A when there is one other aggregate A), etc. be able to. Regarding condition (ii), the “predetermined interval” is set according to the intervals in the “roadway center line broken line”, “lane boundary line broken line”, “crosswalk”, and “zebra zone”. For example, in the “roadway center line broken line” and the “lane boundary line broken line”, the interval between the broken lines is set to about 3 to 20 [m]. In the “pedestrian crossing”, the interval between the parallel lines is about 0.5 [m], and in the “zebra zone”, the interval between the parallel lines is about 1 to 1.5 [m]. Therefore, for example, it is preferable that the interval in the direction parallel to the representative axis Ma of the aggregate A is within 25 [m], and the interval in the direction perpendicular to the representative axis Ma of the aggregate A is within 2 [m]. In addition, about the distance prescribed | regulated in such conditions (ii), it shall convert and determine to an actual distance based on the reduced scale of road image information.

16. Type determination means 13
The type determination unit 13 includes a type determination table 21 in which a type determination condition that defines the condition of the feature amount (individual feature amount and group feature amount) for each road paint type is stored. As the feature quantity that is the basis of the determination at this time, the feature quantity extracted by the feature quantity extraction means 12 is used. Then, based on the type determination condition, the road paint type is determined for each aggregate A and each group of the aggregate A. Specifically, the road paint types to be determined here are “roadway centerline solid line”, “roadway centerline broken line”, “lane boundary solid line”, “lane boundary broken line”, “side line”, “stop line” ”,“ Pedestrian crossing ”, and“ zebra zone ”. The determination processing by the type determination unit 13 will be described in detail later based on the flowcharts shown in FIGS. 11 and 12, and the type determination conditions stored in the type determination table 21 will be described here.

  In the present embodiment, the type determination condition includes an individual condition that is a condition related to an individual feature quantity and a group condition that is a condition related to a group feature quantity. The individual conditions include (a) conditions for the area and circumference of the aggregate A, (b) conditions for the angle between the representative axis Ma of the aggregate A and the link L, and (c) the centroid Oa of the aggregate A. And the distance condition between the link L are set. As group conditions, (d) the condition of the angle between the link L and the direction connecting the centroids Oa of the plurality of assemblies A in the group, and (e) the line connecting the centroids Oa of the plurality of assemblies A in the group. (F) The condition of the angle between the representative axes Ma of the aggregates A in the group is set. Each will be described below.

First, individual conditions in the type determination conditions will be described. (A) The condition of the area and circumference of the assembly A is a continuous long solid road paint such as “solid line of the roadway center line”, “solid line of the lane boundary line”, “side line”, etc. Road paint) and "dashed road dashed line", "lane border dashed line", "stop line", "pedestrian crossing", and "zebra zone" etc. Paint) and short solid line road paint, and noise such as vehicles and traffic lights, and road paint that is not subject to discrimination. This condition corresponds to the determination conditions of steps # 43 and # 50 in the flowchart of FIG. Specifically, this condition includes three conditions: (i) “area ≧ Pa and circumferential length ≧ Qa”, (ii) “Pb ≦ area ≦ Pc and Qb ≦ circumferential length ≦ Qc”, and (iii) “others”. Is set. Here, the condition (i) includes continuous long solid line road paint (solid line road paint) such as “solid line of roadway center line”, “solid line of lane boundary line”, and “side line”. Set the conditions to be met. Here, for example, Pa = 1.5 [m ² ] and Qa = 20 [m] are set. On the other hand, the condition (ii) is that a road paint with a broken line such as “road line middle line broken line”, “lane boundary broken line”, “stop line”, “pedestrian crossing”, “zebra zone”, etc. Of road paint) and short solid line road paint. Here, for example, Pb = 0.75 [m ² ], Pc = 1.4 [m ² ], Qb = 4.5 [m], and Qc = 18 [m] are set. Here, since the shape and dimensions of the road paint are defined by laws and regulations, it is preferable that the numerical ranges defined by these conditions (i) and (ii) are determined based on the laws and regulations. As a result of setting the conditions (i) and (ii) in this way, the condition (iii) includes noise such as vehicles and traffic lights, road paint that is not a discrimination target, and the like.

  (B) The conditions of the angle between the representative axis Ma of the assembly A and the link L are “roadway center line”, “lane boundary line”, “side line”, “crosswalk”, “stop line”, “ This is a condition for distinguishing from the “zebra zone”. This condition corresponds to the determination conditions of steps # 44 and # 52 in the flowchart of FIG. 11 and step # 55 in the flowchart of FIG. Specifically, three conditions of (i) “parallel”, (ii) “vertical”, and (iii) “other” are set as the conditions. The “parallel” and “vertical” conditions should not be strict conditions but should be substantially parallel and substantially vertical conditions, and it is desirable to set a certain angle range. The angular range at this time is obtained experimentally by determining the range in which the direction of the representative axis Ma of the assembly A can vary with respect to the linear strip-like road paint arranged substantially parallel to or substantially perpendicular to the link L. It is preferable to set a wider range. This range is preferably a range narrower than the angle condition between the main axis Mp and the link L in the preliminary determination condition described above in order to improve the accuracy of road paint type determination. Here, for example, ± 10 [°] is set. As a result of setting the conditions (i) to (iii) as described above, the condition (i) indicates that the road paint satisfying the conditions is “roadway center line”, “lane boundary line”, “side line”, and “pedestrian crossing”. Included and excludes “stop lines” and “zebra zones”. On the other hand, condition (ii) includes “stop line” as a road paint satisfying this condition, and “roadway center line”, “lane boundary line”, “side line”, “pedestrian crossing”, and “zebra zone” Excluded. Therefore, “zebra zone” is included as road paint corresponding to the condition (iii).

  (C) The condition of the distance between the centroid Oa of the aggregate A and the link L is a condition for distinguishing and determining “roadway center line”, “side line”, and “lane boundary line”. This condition corresponds to the determination conditions of steps # 45 and # 47 in the flowchart of FIG. Specifically, these conditions are: (i) “within ± 2 [m]” (region i in FIG. 7), (ii) “within ± 2 [m] from a position 1/2 of the road width” (FIG. 7 conditions ii) and (iii) “other than that” (area iii in FIG. 7) are set. The numerical value “± 2 [m]” in these conditions is a numerical value set as a smaller value based on the fact that the width of one lane of the road is around 3 [m]. Only. That is, by setting the width in the direction orthogonal to the link L of the area related to each condition to be less than the width of one lane, the area related to each condition is the “roadway center line”, “side line”, and “lane boundary line”. Two or more of them are not included. Accordingly, other numerical values such as 1.5 [m] and 2.5 [m] can be set as long as the numerical values are less than the width of one lane. In addition, about distances, such as 2 [m], it determines by converting into real distance based on the reduced scale of road image information. As a result of setting the conditions (i) to (iii) in this way, the condition (i) includes “roadway center line” as a road paint satisfying the condition, and excludes “side line” and “lane boundary line”. The On the other hand, the condition (ii) includes “side line” as road paint satisfying this condition, and excludes “roadway center line” and “lane boundary line”. Therefore, the road paint corresponding to the condition (iii) includes “lane boundary line”. In this determination condition, “stop line”, “pedestrian crossing”, and “zebra zone” are not determined.

  Next, the group condition in the type determination condition will be described. Under this group condition, road paint composed of a group of a plurality of aggregates A grouped by the grouping means 19, specifically, “roadway center line broken line”, “lane boundary line broken line”, “crosswalk” , And “zebra zone” are subject to road paint type determination.

  (D) The conditions of the angle between the direction connecting the centroids Oa of the plurality of aggregates A in the group (centroid connection direction) and the link L are “roadway center line broken line”, “lane boundary line broken line”, “ This is a condition for making a distinction between the “pedestrian crossing” and the “zebra zone”. This condition corresponds to the determination conditions of steps # 60 and # 61 in the flowchart of FIG. Specifically, three conditions of (i) “parallel”, (ii) “vertical”, and (iii) “other” are set as the conditions. The “parallel” and “vertical” conditions should not be strict conditions but should be substantially parallel and substantially vertical conditions, and it is desirable to set a certain angle range. The angle range in this case is a plurality of aggregates A constituting “roadway center line broken line” and “lane boundary line broken line” which are road paints in which the direction connecting the centroids Oa of the aggregate A is substantially parallel to the link L. Variation of the direction connecting the centroids Oa, and the centroids Oa of the plurality of aggregates A constituting the “pedestrian crossing” which is road paint in which the direction connecting the centroids Oa of the aggregate A is substantially perpendicular to the link L. It is preferable to experimentally obtain a range in which variations in the connecting direction can occur and to set a range wider than the range. Here, for example, it is set to ± 15 [°]. As a result of setting the conditions (i) to (iii) in this way, the condition (i) includes “roadway center line broken line” and “lane boundary line broken line” as road paint satisfying this, and “crosswalk” And the “zebra zone” is excluded. On the other hand, in the condition (ii), “pedestrian crossing” is included as road paint satisfying this, and “zebra zone” is excluded. Therefore, “zebra zone” is included as road paint corresponding to the condition (iii).

  (E) The condition of the distance between the link (centroid connection line) connecting the centroids Oa of a plurality of aggregates A in the group and the link is as follows: “roadway center line broken line” and “lane boundary line broken line” This is a condition for determining separately. This condition corresponds to the determination condition of step # 64 in the flowchart of FIG. Specifically, two conditions are set: (i) “within ± 2 [m]” (region i in FIG. 8) and (ii) “other” (regions ii and iii in FIG. 7). is doing. The numerical value “± 2 [m]” in these conditions is a numerical value set as a smaller value based on the fact that the width of one lane of the road is around 3 [m]. Only. In other words, by setting the width in the direction perpendicular to the link L of the area related to each condition to be less than the width of one lane, the area related to each condition is both “roadway centerline broken line” and “lane boundary line broken line”. It is not included. Accordingly, other numerical values such as 1.5 [m] and 2.5 [m] can be set as long as the numerical values are less than the width of one lane. In addition, about distances, such as 2 [m], it determines by converting into real distance based on the reduced scale of road image information. As a result of setting the conditions (i) and (ii) as described above, the condition (i) includes “roadway center line broken line” as road paint satisfying the condition (i) and excludes “lane boundary line broken line”. Therefore, the road paint corresponding to the condition (ii) includes “lane boundary broken line”. In this determination condition, “pedestrian crossing” and “zebra zone” are not determined.

  (F) Regarding the condition of the angle between the representative axes Ma of the aggregates A in the group, “zebra zone” is distinguished from noise such as vehicles and traffic lights, road paint that is not a discrimination target, and the like. It is a condition for. This condition corresponds to the determination condition of step # 56 in the flowchart of FIG. Specifically, two conditions of (i) “parallel” and (ii) “other” are set as the conditions. The “parallel” condition should be a substantially parallel condition, not a strict condition, and it is desirable to set a certain angle range. In this case, the angular range is determined by experimentally determining a range in which the angle variation between the representative axes Ma of the plurality of aggregates A constituting the “zebra zone” to be discriminated may occur, and setting the range wider than the range. Is preferred. Here, for example, it is set to ± 15 [°]. As a result of setting the conditions (i) and (ii) in this way, the condition (i) includes a “zebra zone” as a road paint that satisfies this condition, and is not subject to noise or discrimination of vehicles or traffic lights. Road paint etc. are excluded. Therefore, what corresponds to the condition (ii) includes noise such as vehicles and traffic lights, road paint that is not a discrimination target, and the like. The road paint corresponding to the condition (i) includes “pedestrian crossing”, but since the road paint corresponding to the condition (d) (ii) has already been determined, the determination is made here. Not included.

17. Road paint information generation means 14
The road paint information generation unit 14 generates road paint information of the road paint type using the information of the assembly A for which the road paint type is determined by the type determination unit 13 and stores the road paint information in the road paint information database 4. The road paint information includes information indicating the road paint type and information on the shape and arrangement (coordinates) of the road paint.

18. Recognition process by road paint recognition system (overall process)
Next, road paint recognition processing by the road paint recognition system according to the present embodiment will be described in detail based on the flowcharts shown in FIGS. FIG. 9 is a flowchart showing the entire road paint recognition processing method according to the present embodiment. FIG. 10 is a flowchart showing details of the “preliminary determination” process in step # 08 of the flowchart shown in FIG. 11 and 12 are flowcharts showing details of the “type determination” process in step # 11 of the flowchart shown in FIG.

  As shown in FIG. 9, in the road paint recognition system according to the present embodiment, first, road image information is acquired from the road image information database 2 by the image information acquisition means 6 in accordance with an instruction input from the operation unit 5a ( Step # 01). Next, the acquired road image information is binarized by the contour line information extraction means 7 (step # 02). This is because road paints are mostly white or yellow, and areas with low color density in the road image information are likely to fall under road paint, so that such areas are revealed. . After that, the contour line information extracting means 7 extracts contour line information representing the contour line of the light-density region (step # 03). Specifically, the extraction of the contour line information is performed by raster / vector conversion processing for generating vector data representing the contour line of the low-density region. Then, if the contour line of the low-density region is used as it is as vector data, fine irregularities due to the dot unit variation are reflected, so that the contour line information extracting means 7 removes noise for making the contour line close to a straight line. Processing is performed (step # 04). As a result, as shown in FIG. 2, the contour line information representing the contour line of the area where there is a possibility of road painting is extracted.

  Next, as shown in FIG. 3, the polygon dividing means 8 divides the closed area included in the extracted contour information into a plurality of area convex polygons P for principal axis calculation (step # 05). The specific method of this area convex polygon division processing has already been described. Thereafter, as shown in FIG. 4, the main axis calculation means 16 and the centroid calculation means 15 calculate the main axis Mp and the centroid Op of each region convex polygon P divided in step # 05 (step # 06). Further, the road network information acquisition means 17 acquires road network information from the road network database 3 (step # 07). As a result, as shown in FIG. 5, it is possible to obtain the positional relationship between the plurality of region convex polygons P, the nodes N, and the links L. Thereafter, a preliminary feature amount extraction unit 9 extracts a predetermined preliminary feature amount relating to each region convex polygon P, and a preliminary determination unit 10 determines one or more road paint types as candidates for each region convex polygon P. Preliminary determination processing for assigning candidate type information indicating the determination result is performed (step # 08). The preliminary determination process will be described in detail later based on the flowchart shown in FIG.

  Next, as shown in FIG. 6, the plurality of area convex polygons P are combined by the combining means 11 as an aggregate A constituting the same road paint as shown in FIG. 6 (step # 09). This coupling condition is as already described. Then, the representative axis calculating means 18 and the centroid calculating means 15 calculate the representative axis Ma and the centroid Oa of each aggregate A as shown in FIG. 7 (step # 10). After that, the feature amount extraction unit 12 extracts a predetermined feature amount relating to the aggregate A, and the type determination unit 13 performs type determination processing for determining the road paint type for each aggregate A and each group of the aggregate A (step # 11). This type determination process will be described later in detail based on the flowcharts shown in FIGS. Then, road paint information is generated by the road paint information generating means 14 (step # 12) and stored in the road paint information database 4 (step # 13). This completes the road paint recognition process.

19. “Preliminary Determination” Processing Next, details of the “preliminary determination” processing in step # 08 of the flowchart shown in FIG. 9 will be described based on the flowchart shown in FIG. As shown in this figure, in the “preliminary determination” process, first, the preliminary feature amount extraction means 9 acquires the information on the principal axis Mp of each region convex polygon P calculated in step # 06 and the step # 07. Based on the link L information included in the road network information, the angle between the main axis Mp and the link L is calculated for each region convex polygon P (step # 21). Then, the preliminary determination means 10 determines whether or not the angle between the spindle Mp and the link L is vertical (step # 22). If the angle between the main axis Mp and the link L is vertical (step # 22: YES), “stop line” candidate type information is given (step # 23). In the example shown in FIG. 5, candidate type information of “stop line” is given to the area convex polygon P to which the reference symbol P1 is attached. If the angle between the main axis Mp and the link L is not vertical (step # 22: NO), it is next determined whether or not the angle between the main axis Mp and the link L is parallel (step # 24). If the angle between the spindle Mp and the link L is not parallel (step # 24: NO), candidate type information of “zebra zone” is given (step # 25). In the example illustrated in FIG. 5, candidate type information “zebra zone” is assigned to the region convex polygon P to which the symbol P <b> 2 is attached.

  On the other hand, if the angle between the main axis Mp and the link L is parallel (step # 24: YES), then information on the centroid Op of each region convex polygon P calculated in step # 06, and step # Based on the link L information included in the road network information acquired in 07, the distance between the centroid Op and the link L is calculated for each region convex polygon P (step # 26). Then, the preliminary determination means 10 determines whether or not the distance between the centroid Op and the link L is within ± 2 [m] (step # 27). If the distance between the centroid Op and the link L is within ± 2 [m] (step # 27: YES), candidates for “roadway centerline solid line”, “roadway centerline dashed line”, and “pedestrian crossing” Type information is assigned (step # 28). In the example illustrated in FIG. 5, candidate type information of “roadway centerline solid line”, “roadway centerline broken line”, and “pedestrian crossing” is assigned to the area convex polygon P to which symbols P3 and P4 are attached. On the other hand, if the distance between the centroid Op and the link L is not within ± 2 [m] (step # 27: NO), then the distance between the centroid Op and the link L is 1/2 the road width. It is determined whether it is within ± 2 [m] from the determined position (step # 29). When the distance between the centroid Op and the link L is within ± 2 [m] from the position at which the road width is 1/2 (step # 29: YES), the “side line” and the “crosswalk” Candidate type information is assigned (step # 30). In the example shown in FIG. 5, candidate type information of “side line” and “pedestrian crossing” is given to the area convex polygon P to which the symbols P3 and P5 are attached. Further, when the distance between the centroid Op and the link L is not within ± 2 [m] from the position at which the road width is 1/2 (step # 29: NO), the “lane boundary solid line”, “lane boundary” Candidate type information of “dashed line” and “pedestrian crossing” is given (step # 31). In the example shown in FIG. 5, candidate type information of “lane boundary solid line”, “lane boundary broken line”, and “pedestrian crossing” is assigned to the area convex polygon P to which the symbols P3 and P6 are attached. This completes the “preliminary determination” process.

20. “Type Determination” Process Next, details of the “type determination” process in step # 11 of the flowchart shown in FIG. 9 will be described based on the flowcharts shown in FIGS. Of these figures, FIG. 11 mainly shows determination processing based on individual conditions regarding individual feature amounts, and FIG. 12 mainly shows determination processing based on group conditions regarding group feature amounts.

  As shown in FIG. 11, in the “type determination” process, first, the feature amount extraction unit 12 calculates the area and circumference of the aggregate A obtained in step # 09 (step # 41). Further, based on the information on the representative axis Ma of each aggregate A calculated in step # 10 and the information on the link L included in the road network information acquired in step # 07, the representative axis Ma for each aggregate A. And the link L are calculated (step # 42). Then, the type determination means 13 determines whether or not the area of the aggregate A is Pa or more (area ≧ Pa) and the circumference is Qa or more (circumference ≧ Qa) (step # 43). Here, the setting criteria for the values of Pa and Qa are as described above. If the area is equal to or greater than Pa and the circumference is equal to or greater than Qa (step # 43: YES), the type determining unit 13 determines the representative axis Ma and the link L of the aggregate A based on the calculation result of step # 42. It is determined whether the angles are parallel (step # 44). When the angle between the representative axis Ma of the aggregate A and the link L is parallel (step # 44: YES), the aggregate A is a continuous long solid line road paint (the road paint of the solid line segment line). ). Therefore, the distance between the centroid Oa of the aggregate A and the link L is calculated (step # 45).

  Then, the type determining means 13 determines whether or not the distance between the centroid Oa of the aggregate A and the link L is within ± 2 [m] (step # 46). When the distance between the centroid Oa and the link L is within ± 2 [m] (step # 46: YES), the aggregate A is likely to exist near the center in the width direction of the road. It is determined as “center line solid line” (step # 47). In the example shown in FIG. 7, the assembly A to which the reference numeral A <b> 1 is attached is determined as the “roadway center line solid line”. On the other hand, when the distance between the centroid Oa and the link L is not within ± 2 [m] (step # 46: NO), the distance between the centroid Oa and the link L is from a position that is 1/2 of the road width. It is determined whether it is within ± 2 [m] (step # 48). When the distance between the centroid Oa and the link L is within ± 2 [m] from the position at which the road width is 1/2 (step # 48: YES), the aggregate A is in the width direction of the road. Since there is a high possibility that it exists near both sides, it is determined as a “side line” (step # 49). In the example illustrated in FIG. 7, the assembly A to which the reference numeral A2 is attached is determined as the “side line”. When the distance between the centroid Oa and the link L is not within ± 2 [m] from the position at which the road width is 1/2 (step # 48: NO), the aggregate A is located near the center in the width direction of the road and Since there is a high possibility that it is not in any of the vicinity of both sides, it is determined as a “lane boundary solid line” (step # 50). In the example illustrated in FIG. 7, there is no road paint that is determined as a “lane boundary solid line”. In FIG. 7, when a continuous long solid road paint exists at a position where the assembly A to which A3 is attached exists, it is determined as a “lane boundary solid line”.

  On the other hand, when the area ≧ Pa and the circumference ≧ Qa are not satisfied (step # 43: NO), or when the angle between the representative axis Ma of the aggregate A and the link L is not parallel (step # 44: NO), the type determination is performed. In the means 13, it is determined whether or not the area of the assembly A is Pb or more and Pc or less (Pb ≦ area ≦ Pc) and the circumference is Qb or more and Qc or less (Qb ≦ circumference length ≦ Qc) (step #). 51). Here, the setting criteria for the values of Pb, Pc, Qb, and Qc are as described above. If Pb ≦ area ≦ Pc and Qb ≦ peripheral length ≦ Qc are not satisfied (step # 51: NO), “other” is determined (step # 52). Here, “others” includes noise such as vehicles and traffic lights, road paint that is not a discrimination target, and the like. When Pb ≦ area ≦ Pc and Qb ≦ peripheral length ≦ Qc (step # 51: YES), the aggregate A has a broken-line road paint (broken line-shaped road paint) or a short solid line shape. There is a high possibility that it is the assembly A constituting the road paint. Therefore, next, the type determination means 13 determines whether or not the angle between the representative axis Ma of the aggregate A and the link L is vertical based on the calculation result of step # 42 (step # 53). When the angle between the representative axis Ma of the aggregate A and the link L is vertical (step # 53: YES), there is a high possibility that the aggregate A is a short solid road paint perpendicular to the link L. Therefore, it is determined as a “stop line” (step # 54). In the example illustrated in FIG. 7, the aggregate A to which A4 is attached is determined as the “stop line”.

  On the other hand, when the angle between the representative axis Ma of the aggregate A and the link L is not vertical (step # 53: NO), the flow proceeds to FIG. 12, and the grouping means 19 sets the same area and circumference. A plurality of aggregates A that are arranged close to each other at a predetermined interval are grouped together (step # 54). As a result, as shown in FIG. 8, a set of road paints is constituted by a plurality of aggregates A such as “roadway center line broken line”, “lane boundary line broken line”, “crosswalk”, and “zebra zone”. Potential things can be grouped together. Then, the type determination means 13 determines whether or not the representative axis Ma of each aggregate A in the group is parallel to the link L based on the calculation result of step # 42 (step # 55). If the representative axis Ma of each assembly A in the group is not parallel to the link L (step # 55: NO), then the direction of the representative axis Ma of each assembly A in the group is parallel to each other. It is determined whether or not (step # 56). When the directions of the representative axes Ma of the aggregates A in the group are parallel to each other (step # 56: YES), the group is inclined at the same angle with respect to the link L, and a plurality of aggregates A Is determined to be a “zebra zone” (step # 57). In the example illustrated in FIG. 8, the assembly A to which the symbol A5 is attached is determined as the “zebra zone”. If the directions of the representative axes Ma of the aggregates A in the group are not parallel to each other (step # 56: NO), regularity is not seen between the aggregates A in the group. (Step # 58). The contents of “Others” are as described above.

  On the other hand, when the representative axis Ma of each assembly A in the group is parallel to the link L (step # 55: YES), the feature amount extraction unit 12 then performs a plurality of grouped assemblies A. The angle between the direction connecting the centroids Oa (the centroid connection direction) and the link L is calculated (step # 59). Then, the type determining means 13 determines whether or not the angle between the centroid connection direction of the plurality of grouped assemblies A and the link L is parallel (step # 60). When the angle between the centroid connection direction of the plurality of grouped assemblies A and the link L is not parallel (step # 60: NO), the centroid connection direction and the link of the plurality of grouped assemblies A are linked. It is determined whether or not the angle with L is vertical (step # 61). If it is vertical (step # 61: YES), the group may be a group in which a plurality of aggregates A parallel to the link L are spaced apart from each other in a direction orthogonal to the link L. Since it is highly probable, it is determined as “pedestrian crossing” (step # 62). In the example illustrated in FIG. 8, the assembly A to which A6 is attached is determined as a “pedestrian crossing”. If it is not vertical (step # 61: NO), it is determined as a “zebra zone” (step # 57).

  On the other hand, when the angle between the centroid connection direction of the plurality of aggregates A and the link L is parallel (step # 60: YES), the group of the aggregates A is a road paint with a broken line shape ( There is a high possibility that it is a road paint with a lane marking in broken lines. Then, the feature amount extraction means 12 calculates the distance between the link (centroid connection line) and the link connecting the centroids Oa of the plurality of aggregates A grouped (step # 63). Then, the type determining means 13 determines whether or not the distance between the centroid connection lines and the links of the plurality of grouped assemblies A is within ± 2 [m] (step # 64). If the distance between the centroid connection line of the plurality of grouped assemblies A and the link is within ± 2 [m] (step # 64: YES), the group of the assembly A is a road. Since there is a high possibility that it exists in the vicinity of the center in the width direction, it is determined that the roadway center line is broken (step # 65). In the example illustrated in FIG. 8, there is no road paint determined as “roadway centerline broken line”. In addition, in FIG. 7, when the broken-line road paint exists in the position where the aggregate | assembly A to which the code | symbol A1 was attached | subjected exists, it determines with it being a "roadway centerline broken line". On the other hand, when the distance between the centroid connection line of the plurality of grouped assemblies A and the link is not within ± 2 [m] (step # 64: NO), the group of the assembly A is the width of the road. There is a high possibility that it is not near the center of the direction, and here, since the broken “side line” is not a recognition target, it is determined as “lane boundary broken line” (step # 65). In the example illustrated in FIG. 8, the assembly A to which the symbol A3 is attached is determined as the “lane boundary broken line”. This completes the “type determination” process.

[Other Embodiments]
(1) In the above embodiment, as the preliminary determination condition regarding the distance between the centroid Op and the link L and the type determination condition regarding the distance between the centroid Oa and the link L, “within ± 2 [m]” and “ The case where the condition of “within ± 2 [m] from a position 1/2 of the width of the road” is used has been described. However, such constant condition setting is merely an example. Therefore, it is also one of preferred embodiments that these conditions are functions of attributes relating to road width and the number of lanes as link information for determining the width of one lane, for example.

(2) In the above-described embodiment, the combining unit 11 is configured to (i) “difference in the direction of the main axis Mp is within a predetermined allowable range” for the plurality of region convex polygons P, (ii) “with the main axis Mp” Aggregates that form the same road paint with a plurality of region convex polygons P that meet the three conditions of “joining through intersecting edges” and (iii) “common candidate type information is given” The case of combining as A has been described. However, the coupling condition of the coupling means 11 is not limited to this. Therefore, for example, the combining means 11 is configured to combine a plurality of region convex polygons P that are given common candidate type information and are connected to form one closed region as an aggregate A that forms the same road paint. This is also a preferred embodiment. The plurality of region convex polygons P connected to form one closed region are a plurality of regions arranged so as to be in contact with each other directly or via another region convex polygon P so as to form one closed region. It refers to the area convex polygon P.

(3) In the above embodiment, the case where the representative axis Ma of the aggregate A is calculated as a vector representing the weighted average of the main axes Mp of each region convex polygon P with the area of each region convex polygon P as a weight. explained. However, the method of calculating the representative axis Ma of the aggregate A is not limited to this. Therefore, for example, it is also a preferred embodiment to calculate with the principal axis having the smallest cross-sectional secondary moment of the assembly A as the representative axis.

(4) The preliminary feature amount for the region convex polygon P used in the above embodiment and the feature amount (individual feature amount and group feature amount) for the aggregate A are examples, and all these feature amounts are not necessarily included. It is not necessary to extract, and depending on the number of types of road paint to be determined, extracting only a part of these is also one preferred embodiment. It is also a preferred embodiment to extract feature quantities other than these feature quantities.

The block diagram which shows schematic structure of the road paint recognition system which concerns on embodiment of this invention. The figure which shows the state after extracting outline information from the road image information which concerns on embodiment of this invention The figure which shows the state after convex polygon division | segmentation in a part of FIG. The figure which shows the centroid and principal axis of the area | region convex polygon shown in FIG. The figure which shows the state after convex polygon division | segmentation from the outline information of FIG. 2, and road network information acquisition The figure which shows the state after combining the area | region convex polygon of FIG. 5, and producing | generating an aggregate | assembly. The figure which shows the centroid and representative axis of the aggregate | assembly shown in FIG. The figure which shows the state after grouping of the aggregate | assembly shown in FIG. The flowchart which shows the whole recognition processing method of the road paint which concerns on embodiment of this invention. A flowchart showing details of the “preliminary determination” process in step # 08 of FIG. The flowchart which shows the detail of the first half part of the "type determination" process of step # 11 of FIG. The flowchart which shows the detail of the second half part of the "type determination" process of step # 11 of FIG. The block diagram which shows the structure of the electronic map data generation system which concerns on background art

Explanation of symbols

1: System main body 2: Road image information database 3: Road network database 4: Road paint information database 6: Image information acquisition means 7: Outline line information extraction means 8: Polygon dividing means 9: Preliminary feature quantity extraction means 10: Preliminary determination Means 11: Combining means 12: Feature amount extracting means 13: Type determining means 14: Road paint information generating means 15: Centroid calculating means 16: Main axis calculating means 17: Road network information acquiring means 18: Representative axis calculating means 19: Group Converting means 20: preliminary determination table 21: type determination table 22: polygon information acquisition means 23: polygon element information acquisition means N: node L: link P: area convex polygon Op: centroid of area convex polygon Mp: area convex polygon Main axis A: aggregate Oa: centroid of aggregate Ma: representative axis of aggregate

Claims (8)

Image information acquisition means for acquiring road image information including an image of road paint;
Contour information extracting means for extracting contour information of an area with a possibility of road paint included in the acquired road image information;
Polygon dividing means for dividing the closed area included in the contour line information into a plurality of area convex polygons for principal axis calculation;
For each region convex polygon, a main axis calculating means for calculating a main axis that minimizes the sectional moment of inertia;
A coupling means for coupling the plurality of region convex polygons connected through the sides intersecting the main axis as an aggregate constituting the same road paint, the difference in the direction of the main axis is within a predetermined allowable range;
Road paint recognition system with. Road network information acquisition means for acquiring information of a road network having a plurality of nodes having position information and a link connecting the two nodes;
A centroid calculating means for calculating a centroid for each region convex polygon,
One or more candidate road paint types for each region convex polygon based on the relationship between the link arrangement included in the road network information and the direction of the principal axis of the region convex polygon and the position of the centroid. And preliminary determination means for assigning candidate type information representing a determination result,
2. The road paint recognition system according to claim 1, wherein the combining unit combines the plurality of area convex polygons to which the common candidate type information is assigned as an aggregate constituting the same road paint. Image information acquisition means for acquiring road image information including an image of road paint;
Road network information acquisition means for acquiring information of a road network having a plurality of nodes having position information and a link connecting the two nodes;
Contour information extracting means for extracting contour information of an area with a possibility of road paint included in the acquired road image information;
Polygon dividing means for dividing the closed area included in the contour line information into a plurality of area convex polygons for principal axis calculation;
For each region convex polygon, a main axis calculating means for calculating a main axis that minimizes the sectional moment of inertia;
A centroid calculating means for calculating a centroid for each region convex polygon,
One or more candidate road paint types for each region convex polygon based on the relationship between the link arrangement included in the road network information and the direction of the principal axis of the region convex polygon and the position of the centroid. And preliminary determination means for assigning candidate type information representing a determination result;
Combining means for combining the plurality of region convex polygons, which are provided with the common candidate type information and are connected to form one closed region, as an aggregate constituting the same road paint;
Road paint recognition system with.   The preliminary determination means, based on any one or more of the angle between the principal axis of the region convex polygon and the link, and the distance between the centroid of the region convex polygon and the link, for each region convex polygon, The road paint recognition system according to claim 2 or 3, wherein one or more candidate road paint types are determined.   The roads of the respective aggregates based on aggregate information including shape and arrangement information about the aggregates of the region convex polygons coupled by the coupling means, or based on the aggregate information and the road network information The road paint recognition system according to any one of claims 1 to 4, further comprising determination means for determining a paint type. An image information acquisition step of acquiring road image information including an image of road paint;
A contour line information extracting step for extracting contour line information of a region that is likely to be road paint included in the acquired road image information;
A polygon dividing step for dividing the closed region included in the contour line information into a plurality of region convex polygons for principal axis calculation;
For each region convex polygon, a main axis calculation step for calculating a main axis that minimizes the sectional moment of inertia;
A joining step of joining the plurality of region convex polygons connected through the sides intersecting with the main axis as an aggregate constituting the same road paint, the difference in the direction of the main axis is within a predetermined allowable range;
A road paint recognition method comprising: An image information acquisition step of acquiring road image information including an image of road paint;
A road network information acquisition step of acquiring information of a road network having a plurality of nodes having position information and a link connecting the two nodes;
A contour line information extracting step for extracting contour line information of a region that is likely to be road paint included in the acquired road image information;
A polygon dividing step for dividing the closed region included in the contour line information into a plurality of region convex polygons for principal axis calculation;
For each region convex polygon, a main axis calculation step for calculating a main axis that minimizes the sectional moment of inertia;
For each region convex polygon, a centroid calculating step for calculating a centroid;
One or more candidate road paint types for each region convex polygon based on the relationship between the link arrangement included in the road network information and the direction of the principal axis of the region convex polygon and the position of the centroid. And a preliminary determination step of assigning candidate type information representing a determination result;
A combination step of combining the plurality of area convex polygons, which are provided with the common candidate type information and are connected to form one closed area, as an aggregate constituting the same road paint;
A road paint recognition method comprising: An image information acquisition step of acquiring road image information including an image of road paint;
A contour line information extracting step for extracting contour line information of a region that is likely to be road paint included in the acquired road image information;
A polygon dividing step for dividing the closed region included in the contour line information into a plurality of region convex polygons for principal axis calculation;
For each region convex polygon, a main axis calculation step for calculating a main axis that minimizes the sectional moment of inertia;
A joining step of joining the plurality of region convex polygons connected through the sides intersecting with the main axis as an aggregate constituting the same road paint, the difference in the direction of the main axis is within a predetermined allowable range;
A road network information acquisition step of acquiring information of a road network having a plurality of nodes having position information and a link connecting the two nodes;
The road of each aggregate based on aggregate information including shape and arrangement information about the aggregate of the region convex polygons combined in the combining step, or based on the aggregate information and the road network information A determination step of determining a paint type;
A road paint information registration step of registering the aggregate information for which the road paint type has been determined in a database as road paint information of the road paint type;
To create a road paint database.

Images