JP6871806B2 - Road edge extractor - Google Patents

Description

The present invention relates to a road edge extraction device that detects a road edge forming a step based on three-dimensional coordinate data of a point cloud extracted from the surface of a feature.

The following Patent Document 1 discloses a technique for detecting a step on the ground based on point cloud data representing a three-dimensional shape of a feature surface. When the technology detects a boundary (edge) formed by a step in a subspace that divides the target space, the tracking space is sequentially set in the direction of the extension line of the edge, and the edge is searched for in the tracking space. To do.

Japanese Unexamined Patent Publication No. 2012-37490

When the edges are sequentially tracked to extract the road edge, if the tracking fails at a certain point, the road edge beyond that point is not extracted, and there is a problem that the robustness of the extraction is poor.

The present invention has been made to solve the above problems, and even if there is a part where edge extraction is difficult on the road edge, the influence can be locally suppressed and a wide range of road edges can be detected. An object of the present invention is to provide a road edge extraction device.

(1) The road edge extraction device according to the present invention is a device that extracts the road edge forming a step in the target space based on the three-dimensional coordinate data of the point group extracted from the surface of the feature, and the road is a plane thereof. It is a means for dividing a plurality of search unit intervals along a line segment and searching for a candidate segment of the road edge in a subspace corresponding to each search unit interval in the target space, and is a horizontal plane at the road edge. The portion sandwiched between the vertical planes provided at both ends of the search unit interval based on the fact that the density of the point group projected on is increased and the point group has a height difference equal to or higher than a preset step threshold. Candidate segment search means for obtaining a line segment to be the candidate segment corresponding to the search unit interval from the distribution of the point group in space and the candidate segment presumed to be lined up along the road edge are sequentially connected to be candidates. It has a segment connection means for generating a segment sequence, and the segment connection means selects the candidate segment according to the order of the search unit interval, and the selected candidate segment of interest is the preceding candidate selected one before. In determining whether or not the segment has a connection relationship, if there is no candidate segment already connected to the preceding candidate segment, 4 between both ends of the preceding candidate segment and each of both ends of the attention candidate segment. The minimum distance of the street distances is equal to or less than a predetermined first distance threshold, and the preceding candidate segment is a vector whose end point corresponds to the minimum distance, and the attention candidate segment corresponds to the minimum distance. If the angle formed by both vectors when the vector starting from the end point is equal to or less than a predetermined first angle threshold, the preceding candidate segment and the attention candidate segment are between the end points where the minimum distance is obtained. It is determined that the candidate segment sequence has the connection relationship.

(2) In the road edge extraction device according to (1) above, when the preceding candidate segment is already a part of the candidate segment sequence, the segment connecting means has the end point of the preceding candidate segment and the attention candidate. If the minimum distance among the two distances between both ends of the segment is equal to or less than the first distance threshold value, and the angle between the preceding candidate segment and the attention candidate segment is equal to or less than the first angle threshold value. It can be determined that the preceding candidate segment and the attention candidate segment have the connection relationship between the end points having the minimum distance and are included in the same candidate segment sequence.

(3) In the road edge extraction device according to (1) or (2) above, the candidate segment sequence is selected according to the order of the search unit sections, and the selected candidate segment sequence of interest and the candidate segment sequence of interest are selected. Further, the road edge segment row generating means for generating the road edge segment row by connecting to the preceding candidate segment row adjacent to the preceding candidate segment row is further provided, and the road edge segment row generating means is the tail of the leading candidate segment row. The length of the connection vector set from the candidate segment toward the first candidate segment of the attention candidate segment string is equal to or less than a predetermined second distance threshold, and the connection vector, the tail candidate segment, and the first candidate segment are each. If the angle formed by is equal to or less than a predetermined second angle threshold value, the preceding candidate segment row and the attention candidate segment row can be connected by the connection vector to form the road edge segment row.

According to the present invention, it is possible to stably extract a wide range of road edges.

It is a block diagram which shows the schematic structure of the road edge extraction device which concerns on embodiment of this invention. It is a schematic flow chart of the road edge extraction processing by the road edge extraction device which concerns on embodiment of this invention. It is a schematic plan view of an example of the target space. It is a schematic flow chart of a segment search process. It is a schematic diagram explaining the feature of the step of the road edge in a unit space. It is a schematic flow chart of a road edge determination process. It is a schematic flow chart of a segment connection process. It is a schematic diagram explaining the distance between the endpoints of a segment of interest and a leading segment when the leading segment does not form a segment sequence. It is a schematic diagram explaining the distance between the endpoints of a segment of interest and a leading segment when the leading segment forms a segment sequence. It is a schematic flow chart of the connectivity check process. It is a schematic diagram explaining the joining form when the end points of the candidate segments are close to each other. It is a schematic diagram explaining the joining form when the candidate segments intersect each other. It is a schematic diagram explaining the joining form when one segment and the extension line of the other segment intersect. It is a schematic diagram explaining the bonding form when the end point of one segment is in the buffer area of the other segment. It is a schematic flow chart of the segment cleansing process. It is a schematic diagram which shows the state which set the connection vector between the candidate segment sequence. It is a schematic diagram explaining the mode of the cleansing process of a segment sequence. It is a schematic diagram explaining another aspect of the cleansing process of a segment sequence. It is a schematic diagram which shows the example of the segment cleansing process. It is a schematic diagram explaining the smoothing process.

Hereinafter, the road edge extraction device 2 according to the embodiment of the present invention (hereinafter referred to as the embodiment) will be described with reference to the drawings. The road edge extraction device 2 extracts a road edge having a step based on the point cloud data representing the three-dimensional shape of the feature surface. The point cloud data is acquired by, for example, a mobile mapping system. The mobile mapping system is equipped with a laser scanner in an automobile and irradiates the laser diagonally downward from the upper part of the vehicle body. The optical axis of the laser is scanned laterally, and laser pulses are emitted at every minute angle within the scanning angle range. The distance is measured based on the time from the laser emission to the reception of the reflected light, and at that time, the laser emission direction, the time, the position and orientation of the vehicle body, and the like are measured. From these measurement data, point cloud data representing the three-dimensional coordinates of the point where the laser pulse is reflected can be obtained.

FIG. 1 is a block diagram showing a schematic configuration of the road edge extraction device 2. The road edge extraction device 2 is a system including an arithmetic processing unit 4, a storage device 6, an input device 8, and an output device 10. As the arithmetic processing unit 4, it is possible to create dedicated hardware for performing various arithmetic processing of the present device, but in the present embodiment, the arithmetic processing unit 4 uses a computer and a program executed on the computer. Is built.

The CPU (Central Processing Unit) of the computer constitutes the arithmetic processing unit 4, and functions as the candidate segment search means 20, the segment connection means 22, and the road edge segment sequence generation means 24, which will be described later.

The candidate segment search means 20 is a means for dividing a road into a plurality of search unit sections along its horizontal alignment and searching for a road edge candidate segment in a subspace corresponding to each search unit section in the target space. The candidate segment is a candidate for a fragment existing in a subspace of the edge of the road edge.

The segment connection means 22 is a means for sequentially connecting candidate segments presumed to be lined up along the road edge to generate a candidate segment sequence.

The road edge segment row generation means 24 selects a candidate segment row according to the order of the search unit sections, and selects the selected attention candidate segment row and the preceding candidate segment row adjacent to the attention candidate segment row and selected earlier. It is a means of connecting and generating a row of road edge segments.

The storage device 6 is composed of a hard disk or the like built in the computer. The storage device 6 stores a program and other programs for making the arithmetic processing unit 4 function as the candidate segment search means 20, the segment connection means 22, and the road edge segment sequence generation means 24, and various data necessary for processing of this system. To do. For example, the storage device 6 stores point cloud data of a processing target space for extracting a road edge as processing target data.

The input device 8 is a keyboard, a mouse, or the like, and is used by the user to operate the system.

The output device 10 is a display, a printer, or the like, and is used to show the user the edge of the road edge obtained by this system by screen display, printing, or the like.

FIG. 2 is a schematic flow chart of the road edge extraction process by the road edge extraction device 2. The candidate segment search means 20 sets the road reference line along the horizontal alignment of the road, that is, along the longitudinal direction of the road (step S2), and sets the search unit section on the road reference line (step S4). FIG. 3 is a schematic plan view of an example of the target space. The mobile mapping system records the trajectory of the vehicle equipped with the system, and the road reference line 30 can be set based on the trajectory. In the present embodiment, the road reference line 30 is set in the road, that is, in the road width sandwiched between the road edges 32L and 32R. The search unit section _IS may be, for example, a road reference line 30 divided at regular intervals. You can also set to have a overlap search unit interval I _S adjacent to each other, it may be subjected to a set spaced.

A subspace 36 is set in the target space corresponding to each search unit interval. The subspace 36 is a space sandwiched between vertical planes 34 provided at both ends of the search unit interval. In the present embodiment, the two vertical planes 34 at both ends of the search unit section are set orthogonal to the line segment connecting both ends of the search unit section. Therefore, the two vertical planes 34 are parallel to each other even in the curved portion of the road reference line 30.

Candidate segment searching means 20 repeats the process of searching the road edge of the edge for each search unit interval I _S (step S6). Specifically, the edges are searched on both sides of the road reference line 30 in response to the setting of the road reference line 30 in the road. For example, a process (step S8) in which one direction along the road is set as the traveling direction and a candidate segment is searched for in the subspace on the left side of the traveling direction from the road reference line 30, and the traveling direction from the road reference line 30. The process of searching for a candidate segment in the subspace on the right side (step S10) is performed in the loop process of step S6. In the example of FIG. 3, when the traveling direction is set from the lower side to the upper side, the subspace on the road edge 32L side from the road reference line 30 becomes the subspace on the left side in the traveling direction, and the subspace on the road edge 32R side advances. It is a subspace on the right side of the direction.

If the candidate segment searching means 20 searches the candidate segments per search unit interval I _S ends, the road edge determination processing is performed by the segment connecting means 22 and the road edge segment sequence generating unit 24, the road edge is determined (step S12 ).

FIG. 4 is a schematic flow chart of the segment search process (steps S8 and S10 in FIG. 2). The candidate segment search means 20 divides the subspaces 36 on the left side and the right side in the traveling direction into a plurality of unit spaces 38 arranged in the road crossing direction within a certain distance from the road reference line 30 (step S20). Specifically, the subspace 36 is divided by a vertical plane orthogonal to the vertical plane 34, and a unit space 38 sequentially shifted from the road reference line 30 to the road end side is set (step S22).

The unit space 38 set in this way is a quadrangular prism-shaped space in which the target space is divided into a two-dimensional orthogonal lattice (mesh) in a horizontal plane. For example, the length of the search unit section I _S may be a 50 cm (cm) and, correspondingly, the horizontal cross-section width W of the unit space 38 (dimension along the road reference line 30) And the depth D (dimension in the road crossing direction) can be a square of 50 cm. The height H of the subspace or the unit space can be matched with the height of the target space. In addition, the range in the height direction of the subspace or unit space is narrowed to the range including the step on the road surface and the road edge, thereby eliminating the influence of the point cloud by other structures and improving the road edge extraction accuracy. It can also be improved. In the program constituting the candidate segment search means 20, the width W, the depth D, and the height H are parameterized and can be changed by the user using the input device 8, for example.

Further, the fixed distance in step S20 for determining the range in the road crossing direction in which the unit space 38 is set is set to include a position where the road edge is presumed to exist, and is set in advance in consideration of, for example, the width of the road. It can also be appropriately set by the user from the input device 8.

The candidate segment search means 20 takes in a point cloud having coordinates in the set unit space 38 from the point cloud data stored in the storage device 6 (step S24). FIG. 5 is a schematic view illustrating the characteristics of the step on the road edge in the unit space, and FIG. 5A shows a plan view of the unit space 38 in which the edge 50 of the road edge exists, and FIG. b) shows a point cloud 52 in a vertical cross section along the line AA orthogonal to the edge 50 of FIG. 5 (a). The point cloud in the unit space including the step on the road edge has a height difference according to the step, while the point cloud in the unit space not including the step is basically located along the road surface and is flat with a small height difference. It is distributed in. Further, due to the point cloud distributed along the step side surface 54, the distribution of the point cloud projected on the horizontal plane of the unit space including the step includes a portion where the point cloud is densely linearly gathered along the edge 50. On the other hand, the distribution of the point cloud projected on the horizontal plane of the unit space not including the step becomes substantially uniform with a small difference in sparse and dense.

The candidate segment search means 20 extracts the presence / absence of a step in the unit space and the candidate segment based on the difference in the distribution of the point cloud according to the presence / absence of such a step. For example, the candidate segment search means 20 determines the height difference of the point cloud captured in step S24 (step S26). Then, when the distribution of the point cloud is flat without having a height difference according to the step (in the case of "Yes" in step S28), it is determined that there is no step in the currently set unit space 38. And return to step S20.

The candidate segment search means 20 sets the unit space 38 by sequentially shifting it toward the outside of the road within a certain distance from the road reference line 30 until the unit space 38 whose point cloud distribution is not flat is found.

When the distribution of the point cloud in the unit space 38 is not flat, that is, when there is a height difference according to the step on the road edge (when “No” in step S28), the candidate segment search means 20 uses the candidate segment search means 20 as a candidate segment. A line segment in which a group of points gathers in a horizontal plane is searched (step S30). When the candidate segment is extracted (in the case of "Yes" in step S32), the candidate segment is recorded in the edge list in association with the search unit interval (step S34), and the step in a certain search unit interval in FIG. The process of S8 or step S10 is terminated.

Further, when the unit space 38 which is not flat does not exist within a certain distance from the road reference line 30 (when “No” in step S20), and when there is a height difference in the point cloud, it is a line that is a candidate segment. If the minutes are not extracted (in the case of "No" in step S32), the process of step S8 or step S10 is also terminated.

If "No" is selected in step S32, the process may return to step S20 and the search for the candidate segment may be continued in the outer unit space 38. Further, when a candidate segment is extracted in a certain unit space 38 (in the case of “Yes” in step S32), the process returns to step S20, and the search for another candidate segment is continued in the outer unit space 38. You may.

The determination of the height difference of the point cloud in step S26 can be performed, for example, by using a plane detection process using a RANSAC (random sample consensus) algorithm. For example, if the horizontal plane estimated by RANSAC from the point cloud has an area corresponding to the horizontal cross section of the unit space, it can be determined that the point cloud of the unit space is flat. Further, if the height difference between the two estimated horizontal planes is less than the threshold value set according to the step on the road edge, it can be determined to be flat, and if it is equal to or more than the threshold value, it can be determined to be not flat. ..

The search for the candidate segment in step S30 can also be performed using, for example, RANSAC, and the line segment is estimated by RANSAC in the point cloud projected on the horizontal plane. The processing of steps S26 and S30 can be performed at high speed with a relatively small calculation load by using RANSAC.

FIG. 6 is a schematic flow chart of the road edge determination process (step S12 in FIG. 2). The segment connection means 22 checks for the presence or absence of the connection relationship of the candidate segments extracted from the adjacent search unit intervals, and connects the candidate segments having the connection relationship to generate a candidate segment sequence (step S40).

It is presumed that the road edge segment row generation means 24 is not due to the road edge due to an isolated candidate segment that was not connected to another candidate segment in the segment connection process S40, for example, a vehicle on the road, or the like. A process of removing the candidate segment sequence (segment cleansing process) is performed (step S42). In step S42, candidate segment rows presumed to be due to the road edge are further connected to generate a road edge segment row.

The road edge segment row generation means 24 defines each road edge segment row as a group (step S44), and smoothes the shape of the segment row constituting each group (step S46). As a result, the planar shape of the road edge is determined. The road edge segment row generation means 24 further adds an elevation value to the road edge to determine the three-dimensional shape of the road edge (step S48). Specifically, the road edge segment row generation means 24 sets the lowest elevation value of the point cloud in the area separated from the position by a predetermined distance on the road reference line 30 side as the elevation value of the road edge at an arbitrary position. Set. Various types of dust are likely to exist near the road edge, and the altitude value of the point cloud obtained from them is a value off the road surface. Therefore, the altitude value at a position away from the road edge is set as the elevation value of the road surface below the road edge to ensure accuracy.

FIG. 7 is a schematic flow chart of the segment connection process (step S40 of FIG. 6). The segment connection means 22 forms a flag indicating whether or not each candidate segment obtained by the candidate segment search means 20 has a connection relationship with the candidate segment of the adjacent search unit interval, that is, a candidate segment sequence. Initialize the flag F indicating the presence / absence state and the value C indicating the status in the candidate segment sequence (step S60). Specifically, the flag F (*) of any candidate segment "*" is set to "OFF" indicating that there is no connection relationship, and the status C (*) is set so that the candidate segment sequence is not formed. The indicated value is “0”.

The segment connection means 22 sequentially connects candidate segments that are presumed to be lined up along the road edge. The connectivity check process is performed by selecting unprocessed candidate segments according to the order of the search unit intervals, and until there are no unprocessed candidate segments, that is, for all the candidate segments extracted by the search means 20. This is repeated until the connectivity check is completed (in the case of "No" in step S62).

The segment connection means 22 selects search unit intervals in order, and when a candidate segment corresponding to the search unit interval exists (in the case of “Yes” in step S62), the segment connection means 22 pays attention to the candidate segment which is an unprocessed segment. set to the candidate segment _{S B} (step S64). The segment connecting means 22, focused candidate segment adjacent to the currently selected search unit section corresponding to the S _B, if the candidate segment search unit intervals selected immediately before this current search unit section is present, the candidate segment and preceding candidate segments S _a.

Preceding If the candidate segments S _A is not present (in the case of "No" in step S66), since the candidate segment for the target candidate segments S _B are subject to connectivity checks it does not exist, without connectivity check process Returning to step S62, an unprocessed segment to be a new candidate segment of interest is searched for. At this time, the target candidate segments S _B and the preceding candidate segments S _A.

On the other hand, (if in step S66 "Yes") when the preceding candidate segments S _A is present, the preceding candidate segments S _A is already checked connectivity. Segment connection means 22, the preceding candidate indicates that the segment S flag _A F (S _A) has a connection relationship examines whether "ON" (step S68), according to the result processing Make a branch.

Figure 8 is a schematic diagram illustrating a process of (the case of "No" in step S68) If F _{(S A)} is "OFF". If F _{(S A)} is "OFF", the preceding candidate other candidates segments in the segment _{S A} is not yet connected, that is, prior candidate segments _{S A} across _N 1, _{N 2} of candidate segments connected It is an open end that has not been done. Therefore, there is a possibility that either of the end points _N 1, _{N 2} are connected to the focused candidate segments _{S B.} Therefore, the segment connecting means 22 calculates the endpoints distance for the combination of the end points of the four ways between the preceding candidate segments _{S A} across _N 1, _{N 2} and focused candidate segments _S across _N 3 _B in, _{N 4} (Distances shown by dotted lines in FIG. 8A), and their minimum values L are obtained (step S70).

Then, the connectivity check process as in end points that gives the minimum value L and S _A and S _B is likely to be connected (step S72). In the example of FIG. 8, is the connectivity check process assumptions and end point N ₃ and is connected to S _A endpoints N ₂ and S _B of the distance is a minimum of N ₂ and N ₃ are carried out. Further, in the connectivity check process S72, the preceding candidate segment S _A is a vector whose end point is the end _{point N 2} corresponding to the minimum distance, _{and the attention candidate segment S B} is a vector whose start point is the _{end point N 3} corresponding to the minimum distance. _{a, S B} each orientation is defined. Thus, for example, two candidate segments arranged along the road edges _S A, _S A in the direction toward the _{S B} from _{S A} for _{S B, S B} each orientation is defined.

In the connection check processing S72 will be described later, when the the S _A and S _B is determined to have the connection relation, the flag F of the target candidate segments S _B (S _B) is set to "ON", the connection relationship when it is determined that no is, F (S _B) is maintained at the initially set "OFF".

If flag F _{(S B)} is "ON" (step S74 in a case of "Yes"), the segment connecting means 22, the status C of the prior candidate segments _{S A (S A),} the candidate segment column head of the value "1" indicating that the segment is, on the other hand, the status C of interest candidate segments S _B (S _B), and the value "2" indicating the tail of the segment of the candidate segment string (step S76) ..

Figure 9 is a schematic diagram illustrating a process of (the case of "Yes" in step S68) If F _{(S A)} is "ON". If F (S _A) is "ON", the preceding candidate segments S _A is a part of the candidate segment column, specifically a tail of the candidate segment column. In other words, the preceding candidate segments S already on one end point N ₁ is connected to the other candidate segment of _A, only the other endpoint N ₂ is an open end. Therefore, it focused candidate segments _{S B} may be connected to the end point _{N 2.} Therefore, the segment connecting means 22 calculates the endpoints distance for the combination of the end points of two ways between the preceding candidate segments S endpoints of N _{2 A} and focused candidate segments S across N _3, N ₄ and _B (FIG. 9 (Distance indicated by the dotted line in (a)), and the minimum value L thereof is obtained (step S80).

Then, the connectivity check process as in end points that gives the minimum value L and S _A and S _B is likely to be connected (step S82). In the example of FIG. 9, by connectivity check process assumptions and end point N ₃ and is connected to S _A endpoints N ₂ and S _B of the distance is a minimum of N ₂ and N ₃ are carried out. Further, in the connectivity check process S82, as a vector of interest candidate segments S _B is that starting from the end point N ₃ corresponding to the minimum distance, the orientation of the S _B are defined. Incidentally, in the processing that precedes for prior candidate segments S _A, already orientation defined as vector start point N _1, and N ₂ as the end point.

In the connection check processing S82 will be described later, when the the S _A and S _B is determined to have the connection relation, the flag F of the target candidate segments S _B (S _B) is set to "ON", the connection relationship when it is determined that no is, F (S _B) is maintained at the initially set "OFF".

If flag F _{(S B)} is "ON" (step S84 in a case of "Yes"), the segment connecting means 22, the status C of the prior candidate segments _{S A (S A),} the interior of the candidate segment column segment, i.e. the value "3" indicating that the segment other than the head and tail, while the status C of interest candidate segments S _B (S _B), a value indicating the tail of the segment of the candidate segment string " 2 ”(step S86). After steps S76 and S86, the process returns to step S62, and an unprocessed segment, which is a new candidate segment of interest, is searched for.

In the case of F _{(S B)} is "OFF" in step S74, S84, the process returns to step S62, unprocessed segment is searched for a new attention candidate segments. In this case, the status _{C (S A), C (} S B) is not changed. Specifically, in the case of F _{(S B)} is "OFF" in step _{S74, C (S A),} C (S B) are both maintained an initial value "0" at step S84 F ( for S _B) is _{"OFF", C (S a} ) is _{"2", C (S B} ) is maintained respectively the initial value "0".

FIG. 10 is a schematic flow chart of the connectivity check processes S72 and S82. The minimum value L of the inter-terminal distance calculated in step S70 or S80 of FIG. 7 _{is compared with a predetermined threshold value (first distance threshold value) L 1} (step S100). Threshold L ₁ is defined as the upper limit of the distance between the terminals is not denied connection relationship between S _A and S _B. If L is the threshold value _{L 1} or less (if at step S100 _"Yes"), S A, segment direction _{S B} in FIG. 8, is defined by the vector described using the example of FIG. 9 (step S102 ).

The segment connecting means 22 further _{compares L with the threshold L 2} (step S104). The threshold value L ₂ is set to a value for determining whether _{L 2} <L _{1 and L≈0.} Here, the L ≒ 0, refers to the degree that can be regarded as substantially S _A, is S _B are connected by end points of the minimum distance L. Specifically, the threshold value L ₂ is a relatively small value such as a point cloud measurement error or a road edge unevenness, and can be, for example, several centimeters.

If it is determined that L ≒ 0 by a L ≦ _{L 2} in step S104 (if at step S104 _"Yes"), calculates the S A, the angle of _{S B} theta (Step S106) , Compared with the first angle threshold value θ ₁ (step S108). θ is the angle between the segment directions defined in step S102. The edges of the road can bend at large angles, such as at intersections of roads, but otherwise they can be smooth over relatively long distances, that is, the edges can be continuous without significantly changing direction. Therefore, the threshold value θ ₁ is set to a relatively small value corresponding to the angle fluctuation in the smoothly continuous portion of the road edge, and the threshold value θ 1 is imposed on the condition that the _{threshold value is θ 1 or less.} The candidate segment sequence is preferably extracted.

11 L ≒ 0, is a schematic diagram illustrating a connection form of the candidate segments S _A, S _B if that is the end point is close. 11 (a) is the step S104, S108 in _{L ≒ 0 (L ≦ L 2} ) cutlet if it is determined theta ≦ theta ₁ and _S A, shows the positional relationship between the _{S B.} In this case, the segment connecting means 22, as shown in FIG. 11 (b), _S A, joining the _{S B} with a minimum distance L become endpoint _N 2, _{N 3} of the midpoint _{P m} (Step S110). Specifically, moving the leading candidate segments _S endpoint _{N 2} and focused candidate segments _S starting _{N 3 B} in _A together midpoint _{P m.}

When L ≦ L ₁ but L> L _2, that is, when L ≈ 0 (“No” in step S104), the segment connection means 22 checks another connection form. In this case, similarly to step _S106, and calculates the S A, the angle of _{S B} theta (step S120), the requirements of connect θ ≦ θ ₁ (step S122).

In the case of θ ≦ θ ₁ (in the case of “Yes” in step S122), the segment connecting means 22 has a form in which the segments intersect each other (step S124), and one segment and the extension line of the other segment intersect as the connection form. It is sequentially examined whether or not the mode (step S128) and the end point of one segment exist in the buffer area centered on the other segment (step S132).

Figure 12 is a schematic diagram illustrating a bonding form in the case of candidate segments S _A, is S _B intersect. 12 (a) is, when the preceding candidate segments _{S A} and focused candidate segments _{S B intersect,} S A, shows the positional relationship between the _{S B.} In this case, the segment connecting means 22, as shown in FIG. 12 _(b), S _A, joining the _S A, _{S B} at the intersection _{P c} of _{S B} (step S126). Specifically, both move to the intersection _{P c} the preceding candidate segments _S endpoint _{N 2} and focused candidate segments _S starting _{N 3 B} in _A.

Figure 13 is a schematic view illustrating a bonding form of case candidate segments S _A, which is the one and the other extension of S _B intersect. 13 (a) shows the case where the preceding candidate segments _{S A} and extension 60 of the target candidate segments _{S B intersect,} S A, shows the positional relationship between the _{S B.} In this case, the segment connecting means 22, as shown in FIG. 13 (b), _S A at the intersection _{P c} between the extended line 60 of the _{S A} and _{S B,} joining the _{S B} (step S130). Specifically, both move to the intersection _{P c} the preceding candidate segments _S endpoint _{N 2} and focused candidate segments _S starting _{N 3 B} in _A.

Figure 14 is the candidate segment S _A, corresponding endpoint to the minimum distance L of either one segment of the S _B, which is a schematic diagram illustrating a bonding form when present in the buffer area set around the other segments. FIG. 14 (a), _S A when starting _{N 3} of the target candidate segments _{S B} is present in the buffer area 64 of the prior candidate segments _{S A,} shows the positional relationship between the _{S B.} Buffer area 64 of the prior candidate segments _{S A} is a band-shaped area having a width 2L ₃ around the straight line 62 that overlaps the _{S A.} L ₃ is a predetermined distance threshold value, and is basically set so that L ₃ <L ₁ . In this case, the segment connecting means 22, as shown in FIG. 14 (b), _S A, joining the _{S B} with a minimum distance L become endpoint _N 2, _{N 3} of the midpoint _{P m} (Step S134). Specifically, moving the leading candidate segments _S endpoint _{N 2} and focused candidate segments _S starting _{N 3 B} in _A together midpoint _{P m.}

Featured buffer area of the candidate segments S _B are also set similarly, the end point N ₂ prior candidate segments S _A is bonded at the midpoint P _m in the same manner when present in the buffer area.

That is, the segment connecting means 22, or the distance between the starting point _{N 3} and the line 62 of the _{S B} is _{L 3} or less, when the distance between the straight line that overlaps the end _{N 2} and _{S B} of _{S A} is _{L 3} or less In addition, the midpoint joining (step S134) is performed assuming that the condition of step S132 is satisfied.

Above _S A, it determines that the _{S B} have a connection relationship either case joining them (S110, S126, S130, S134 ), the flag F _{(S B)} is set to "ON" (Step S112) , The connectivity check processes S72 and S82 are completed.

On the other hand, when L> L ₁ (when “No” in step S100), when θ> θ ₁ (when “No” in steps S108 and S122), and in steps S124, S128 and S132. If that does not satisfy the conditions of any connection mode (if at step S132 "no"), the flag F _{(S B)} ends the connection remains check processing S72, S82 of the initial setting "OFF".

When the segment connection process (step S40 in FIG. 6) including the connectivity check process described above by the segment connection means 22 is completed, the segment cleansing process (step S42 in FIG. 6) by the road edge segment row generation means 24 is performed. ..

FIG. 15 is a schematic flow chart of the segment cleansing process S42. Among the candidate segments, in the segment connection process S40, the candidate segment that is not connected to other candidate segments and remains as a single unit is deleted because there is a high possibility of noise (step S140). Specifically, the candidate segment whose status C is the initial value "0" is deleted. As a result, only the candidate segment column remains.

For the candidate segment sequence, a process of connecting those presumed to be due to the road edge to generate a road edge segment sequence is performed. The connection process of the candidate segment column T is performed by sequentially selecting the candidate segment column T according to the order of the search unit sections, as in the connection process of the candidate segment S by the segment connection means 22 described above, and the selected candidate of interest. The connection relationship is determined by the segment column and the preceding candidate segment column that is adjacent to the attention candidate segment column and is selected first. That is, the connection process is performed by selecting the unprocessed candidate segment sequence according to the order of the search unit intervals, and is repeated until there are no unprocessed candidate segment sequences (in the case of "No" in step S142).

Here, it represents a flag indicating whether the selected target candidate segment sequence (target segment column T _B) has a connection relationship with the preceding candidate segment sequence (preceding the segment string T _A) and F (T _B). The flag is initially set to "OFF" for any candidate segment string, and is set to "ON" when it is determined that the flag has a connection relationship.

Road edge segment sequence generator 24, (if in step S142 "Yes") if the candidate segment column there is an unprocessed, and sets the candidate segment column to notice segment column T _B (step S144).

When the preceding segment sequence T _A with respect to the target segment column T _B is not present (if at step S146 "No"), the T because the candidate segment string to be determined object connection relation for the target segment column T _B is absent _{For B} , the process of checking the connection relationship is omitted, and the process returns to step S142 to search for an unprocessed segment string which is a new sequence of interest. At this time, the target segment sequence T _B and the preceding segment sequence T _A.

On the other hand, when the preceding segment sequence _{T A} for the target segment column _{T B} is present (if at step S146 "Yes"), between the top candidate segments _{S H} of trailing candidate segments _{S T} and _{T B} of the _{T A} to set the connection vector _{S M} (step S148). FIG. 16 is a schematic view showing this state. As shown in FIG, S _M is defined as a vector directed to the start of S _H from the end of S _T.

Road edge segment sequence generator 24, (if in step S150 "Yes") when the length of the connection vector _{S M} is a predetermined threshold value (second distance threshold) _{L 4} or less, further, connection vector S and _M, the prior segment sequence _{T a} an angle theta _{(S T, S} M) of the trailing candidate segments _{S T,} and connecting the vector _{S M,} the angle between the top candidate segments _{S H} of the target segment column _{T B} theta _(S H, _{S M)} determines whether is a second threshold angle theta ₂ below a predetermined (step S152). For example, an edge (road edge segment row) consisting of a row of candidate segments having a straight line or a gentle curvature of a certain road is naturally interrupted and connected at a part such as a crossroad or a T-junction in which another road connects to the road. There can be as much space as the width of the road. In other words, it is not appropriate to connect in such a candidate segment of about road width occurring between the column size also connected vector S _M to interval. Therefore, the threshold value L ₄ can be set to, for example, less than the road width, and can be set to, for example, about 4 meters. Further, θ ₂ can be set to a value common to, for example, θ _1.

θ _(S T, S _M) and _{θ (S H, S M)} ( if at step S152 "Yes"), then the theta ₂ less, in step S108, S122 in FIG. 10 for connection of the candidate segment for the same reason as that it has a condition θ ≦ θ _1, T is connected via the S _{M a,} determines that T _B is suitable as an edge of the road, F a (T _B) to "oN" Set (step S154). Thus, the prior segment sequence T _A and target segments string T _B the connection vector S road edge segment rows connected by _M is generated.

17, FIG. 18 illustrates the cleansing process of the segment string to be in θ _(S T, S _M) or θ _(S H, S _M) is greater than theta ₂ (if at step S152 "No") It is a schematic diagram. In the process shown in FIG. 17, if the threshold value L ₅ than the length of the target segment column T _B, and that there is a high possibility of noise caused such a vehicle that has been parked in the roadside, the target segment sequence T _B remove and return to (step S162) step S142, re-select the target segment columns _{T B.} In Figure 17, FIG. (A) Segment column shown in _T A, connecting vector _{S M} which is set between _{T B} is, theta _(S T, _{S M)} and theta _(S H, _{S M)} is theta ₂ below not satisfy the condition (step S152) that is also because the length of _{T B} is _{L 5} or less, deletes the target segment column _{T B.} FIG. 17 (b) shows a state in which deletes the _{T B} at step S162, the process returns to step S142, the segment sequence _{T C} is set as a new target segment column. Then, prior new connection vector between the segment columns T _A and T _C S _{M 'is} set, the connection relationship is determined.

18, if the length of _{T B} is greater than _{L 5} shows the process of (if at step S160 "No"), respectively, from the beginning of the tail and notice segment column _{T B} of the prior segment sequence _{T A} updates the leading candidate segment of tail candidate segment and T _B of the T _a by removing the candidate segments one by one, resets the connection vector. In other words, to remove the top candidate segments _{S H} of trailing candidate segments _{S T} and _{T B} of _{T A} shown in FIG. 18 (a). FIG. 18 (b) shows the state after the deletion, the prior segment sequence _{T A} is _{T A} is removed the _{S T} ', and the target segment column _{T B} is _{T B} is deleted to _{S H',} and the S _{M 'is} set as a new connection vector between the T _a' top candidate segments S _{H 'and} T _B' tail candidate segments S _T of _'. Then, this state, examining the connections between the _{T A} 'and _{T B'} by the same process as step S150 and subsequent steps (step S172). As a result, (if in step S172 "Yes") as long recognized connection relation, _{T B} ', that is, set to a new target segment sequence _{T B} of the flag F _{(T B)} the "ON" (step S174 ). Thus, FIG. 18 (b) _{T A} shown in _{', S M', T B} ' are integrated road edge segment column.

On the other hand, if the connection relationship was observed kept (if at step S172 "No"), _{T B} ', i.e. the flag F _{(T B)} of the new target segment column _{T B} is "OFF" (Step S176).

Also, if it is greater than the threshold _{L 4} length of _{S M} (if at step S150 "No") is also kept of the flag F of the target segment column _{T B (T B)} is "OFF" (Step S178).

After steps S154, S174, S176, and S178, the process returns to step S142, and a new sequence of interest segments is searched for.

FIG. 19 is a schematic view showing an example of the segment cleansing process S42. FIG. 19 (a) shows an example of the result of the segment connection process S40, while the candidate segment strings T ₁ through T ₄ are generated, the candidate segments S ₁ is remains alone. FIG. 19 (b) by the processing in step S140 in FIG. 15, in a state of candidate segments _{S 1} from the state itself shown in FIG. 19 (a) is deleted. FIG. 19 (c) shows the result of performing the processing of steps S142 to S178 on the candidate segment sequence group of FIG. 19 (b). For example, a candidate segment columns _T 1, _{T 2} is by the condition is satisfied in step S150, S152, it is connected via a connection vector _{S M1.} On the other hand, _{regarding the connectivity of the candidate segment columns T 2} and T ₃ , for example, the angle with the connection vector set between them _{exceeds the threshold value θ 2} (when “No” in step S152), and T ₃ Since the length is _{equal to or less than the threshold value L 5} (in the case of “Yes” in step S160), T ₃ which is a sequence of interest segments is deleted (step S162). Then, T ₄ is set as a new segment of interest and _{the connectivity of T 2} and T ₄ is examined. For example, these candidate segment sequences T ₂ and T ₄ satisfy the conditions of steps S150 and S152. , Connected via the _{connection vector SM2.} As a result, a road edge segment sequence in which _{T 1} , _SM 1, T ₂ , SM ₂ and T _{4 are connected is generated.}

FIG. 20 is a schematic diagram illustrating the smoothing process S46. FIG. 20A shows a state in which the road edge segment row 70 is cut at a certain distance (radius of a circle indicated by a dotted line), and the “x” mark in the figure is the cutting point. FIG. 20B shows a state in which the cutting points are connected in order to generate a new road edge segment row 72, which is the result of smoothing the road edge segment row 70. ..

2 Road edge extraction device, 4 Arithmetic processing device, 6 Storage device, 8 Input device, 10 Output device, 20 Candidate segment search means, 22 Segment connection means, 24 Road edge segment sequence generation means.

Claims (3)

It is a road edge extraction device that extracts road edges that form steps in the target space based on the three-dimensional coordinate data of the point cloud extracted from the surface of the feature.
A means for dividing a road into a plurality of search unit sections along its planar alignment and searching for a candidate segment of the road edge in a subspace corresponding to each search unit section in the target space. The point group projected on the horizontal plane is sandwiched between the vertical planes provided at both ends of the search unit interval based on the fact that the density of the point group is increased and the point group has a height difference equal to or higher than a preset step threshold value. Candidate segment search means for obtaining a line segment to be the candidate segment corresponding to the search unit interval from the distribution of the point group in the subspace.
It has a segment connection means for sequentially connecting the candidate segments presumed to be lined up along the road edge and generating a candidate segment sequence.
The segment connection means selects the candidate segment according to the order of the search unit intervals, and determines whether or not the selected candidate segment of interest has a connection relationship with the preceding candidate segment selected immediately before.
When there is no candidate segment already connected to the preceding candidate segment, the minimum distance among the four distances between both ends of the preceding candidate segment and each of both ends of the attention candidate segment is a predetermined first. Both of them when the preceding candidate segment is equal to or less than the distance threshold and the leading candidate segment is a vector whose end point corresponds to the minimum distance and the attention candidate segment is a vector whose start point is the end point corresponding to the minimum distance. If the angle formed by the vector is equal to or less than a predetermined first angle threshold value, it is considered that the candidate segment sequence has the connection relationship between the leading candidate segment and the end point at which the attention candidate segment is the minimum distance. To judge,
A road edge extraction device characterized by. In the road edge extraction device according to claim 1,
In the segment connecting means, when the leading candidate segment is already a part of the candidate segment sequence, the minimum distance among the two distances between the end point of the leading candidate segment and both ends of the attention candidate segment is If it is equal to or less than the first distance threshold value and the angle between the preceding candidate segment and the attention candidate segment is equal to or less than the first angle threshold value, the preceding candidate segment and the attention candidate segment become the minimum distance. Judging that the connection relationship is established between the end points and the candidate segments are included in the same candidate segment sequence.
A road edge extraction device characterized by. In the road edge extraction device according to claim 1 or 2.
The candidate segment row is selected according to the order of the search unit section, and the selected attention candidate segment row is connected to the preceding candidate segment row adjacent to the attention candidate segment row and previously selected, and the road edge segment row is connected. Further has a road edge segment sequence generating means to generate
In the road edge segment row generation means, the length of the connection vector set from the tail candidate segment of the leading candidate segment row to the leading candidate segment of the attention candidate segment row is equal to or less than a predetermined second distance threshold. If the angle formed by the connection vector and each of the tail candidate segment and the head candidate segment is equal to or less than a predetermined second angle threshold, the leading candidate segment sequence and the attention candidate segment sequence are connected by the connection vector. A road edge extraction device, characterized in that the road edge segment row is formed.

Images