JP6834330B2 - Track boundary detection method and track boundary detection device - Google Patents

Description

The present invention relates to a track boundary detection method and a track boundary detection device.

A roadside object detection device that detects a roadside object using a plurality of captured images is known (see Patent Document 1). In Patent Document 1, when the road surface structure near the vehicle is estimated from the parallax image obtained by using a stereo camera, the height of the road surface is read out along the scanning horizontal line on the image, and the height changes by about a threshold value, A step is detected.

Japanese Unexamined Patent Publication No. 2014-2608

However, a curb-like track boundary whose height changes sharply can be identified by a change in the height of the road surface in a narrow search width. However, it is difficult to detect a track boundary such as a slope whose height changes gently because the height change of the road surface in a narrow search width is small. In addition, it was difficult to accurately identify the end points of the track boundary when judging by the height change of the road surface over a wide search width.

The present invention has been made in view of the above problems, and an object of the present invention is to detect both a track boundary whose height changes sharply and a track boundary whose height changes gently. It is to provide a detection method and a track boundary detection apparatus.

The track boundary detection method and the track boundary detection device according to one aspect of the present invention set a first search area on the road surface around the vehicle, and for each position coordinate in the first search area, the road surface with the first search width. The first height change of is calculated, the second search area narrower than the first search area is set on the road surface, and the road surface with the second search width narrower than the first search width for each position coordinate in the second search area. The second height change of is calculated, and the position of the step is detected based on the second height change when the second search width is the narrowest.

According to the track boundary detection method and the track boundary detection device according to one aspect of the present invention, both a track boundary whose height changes sharply and a track boundary whose height changes gently can be detected.

FIG. 1 is a block diagram showing an overall configuration of the track boundary detection device 1 according to the embodiment. FIG. 2 is a perspective view showing an example of a linear step determination position (Pa) set on the road surface around the vehicle Vc. FIG. 3 is a bird's-eye view corresponding to FIG. FIG. 4 is a graph showing the height distribution of the road surface Fr at the step determination position Pa. FIG. 5 is for explaining a procedure for gradually narrowing down the step candidate positions (CP) based on the spatial frequency as an example of the height change of the road surface (Fr) and finally determining the step position (CP4). It is a figure of. FIG. 6 is a flowchart showing an example of a track boundary detection method using the track boundary detection device of FIG. FIG. 7 is a perspective view showing an example of a plurality of closed loop-shaped step determination positions (Pa1 to Pa4) set on the road surface around the vehicle Vc. FIG. 8 is a bird's-eye view corresponding to FIG. 7.

An embodiment will be described with reference to the drawings. In the description of the drawings, the same parts are designated by the same reference numerals and the description thereof will be omitted.

(First Embodiment)
The overall configuration of the track boundary detection device 1 according to the embodiment will be described with reference to FIG. The track boundary detection device 1 detects the height of the road surface (hereinafter referred to as "road surface") around the vehicle, and detects the position of the step on the road surface from the height change width of the road surface.

Specifically, the track boundary detection device 1 includes a distance measuring sensor 12, a microcomputer 13, and a database 19. The distance measuring sensor 12 detects the three-dimensional shape of the road surface around the vehicle. The microcomputer 13 executes a series of information processing for detecting the position of a step from the height data of the road surface by the distance measuring sensor 12. The database 19 stores various data used when executing a series of information processing for detecting the position of a step, and intermediate data generated during the execution of the series of information processing.

An example of the ranging sensor 12 is a laser range finder (LRF). The laser range finder detects the distance and orientation from the vehicle to the surface of the object by emitting a pulsed laser and detecting the reflected light from the object around the vehicle. The surface of an object includes roads, curbs, slopes, and embankment surfaces (road surfaces). Another example of the ranging sensor 12 is a stereo camera. A stereo camera can also record information in the depth direction of an object around the vehicle by simultaneously photographing an object around the vehicle from a plurality of different directions. By performing predetermined image processing on the stereo image obtained by the stereo camera, the distance and orientation to the surface of the object reflected in the stereo image can be acquired.

The microcomputer 13 is composed of, for example, a general-purpose microcomputer including a CPU, a memory, and an input / output unit, and by executing a computer program installed in advance, a plurality of information processing circuits (14) included in the track boundary detection device 1. ~ 18) is configured. Each part constituting the microcomputer 13 may be composed of one piece of hardware or may be composed of separate hardware. The microcomputer 13 may also be used as an electronic control unit (ECU) used for other controls related to the vehicle Vc such as automatic driving control. The microcomputer 13 repeatedly executes a series of information processing cycles for detecting a step on the road surface from the data acquired by the distance measuring sensor 12 at predetermined time intervals.

The plurality of information processing circuits configured by the microcomputer 13 include an arithmetic circuit 14, a step determination position circuit 15, a step candidate narrowing circuit 16, and a step detection circuit 18. The step candidate narrowing circuit 16 includes a height change circuit 17a and a search area circuit 17b.

The calculation circuit 14 constitutes the height detection sensor 11 together with the distance measuring sensor 12. The arithmetic circuit 14 applies coordinate conversion processing to the distance and orientation (polar coordinates) from the vehicle to the object obtained by the laser range finder to obtain a three-dimensional shape (cartesian coordinates) of the object around the vehicle. Get information. As a Cartesian coordinate system, for example, centering on the vehicle, if the x-axis is taken in the traveling direction, the y-axis is taken in the vehicle width direction, and the z-axis is taken in the vertical direction, the height detection sensor 11 is the road surface around the vehicle. The height (height data) of can be detected as the z coordinate.

The step determination position circuit 15 sets a linear step determination position for detecting the position of the step on the road surface on the coordinates of the height data. As an example, a linear step determination position extending in the vehicle width direction is set on the road surface. For example, as shown in FIGS. 2 and 3, the height data of the step determination position (Pa) extending in the direction orthogonal to the predetermined direction (Da) from the distance measuring sensor 12 by a predetermined distance in the predetermined direction (Da). Set to the road surface on the coordinates of. In the bird's-eye view shown in FIG. 3, it can be seen that the step determination position (Pa) is linear. 2 and 3 show an example in which the distance measuring sensor 12 is installed in the front part of the vehicle (Vc) and the traveling direction of the vehicle (Vc) is set to a predetermined direction (Da). Therefore, the step determination position (Pa) extending in the vehicle width direction is set in front of the vehicle (Vc). The step determination position (Pa) is set within the detection range of the distance measuring sensor 12. The predetermined direction (Da) is not limited to the traveling direction of the vehicle (Vc), and may be, for example, the vehicle width direction. The set step determination position (Pa) is stored in the database 19.

In the examples shown in FIGS. 2 and 3, a step (LD) in which the height of the road surface changes discontinuously or abruptly is formed on the shoulder, which is the end in the width direction of the roadway on which the vehicle (Vc) can travel. It is formed. On the outside of the roadway with the step (LD) as a boundary, a stepped portion (for example, a sidewalk or a shoulder) whose road surface is one step higher than that of the roadway is provided. As described above, in the examples shown in FIGS. 2 and 3, the road is composed of a roadway and a step portion (sidewalk or shoulder), and a step (LD) is formed at the boundary between the roadway and the step portion (sidewalk or shoulder). ing. The linear step determination position (Pa) extends in a direction crossing the roadway, the step (LD), and the step portion (sidewalk or shoulder).

The positional relationship of the step determination position (Pa) with respect to the distance measuring sensor 12 described above is only an example. Other examples will be described later with reference to FIGS. 7 and 8.

On the other hand, although not shown, there are cases where a slope or embankment whose height changes gently is formed on the shoulder, which is the end of the roadway in the vehicle width direction. The track boundary detection device according to the present embodiment detects slopes and embankments (gentle steps) formed at the boundary between the roadway (Rr) and the step portion (Rd) in the same manner as curbs (steep steps). can do.

The step determination position circuit 15 extracts the road surface height data at the step determination position (Pa) from the road surface height (height data) acquired by the calculation circuit 14. The vertical axis of FIG. 4 indicates the height of the road surface (Fr), and the horizontal axis indicates the step determination position (Pa) extending in the vehicle width direction. The extracted road surface height data is stored in the database 19.

As shown in FIG. 4, a step (LD) is formed at the boundary between the roadway (Rr) and the step portion (Rd). The road surface (Fr) of the roadway (Rr) is provided with a cant whose height decreases from the central portion toward the shoulders at both ends. This is a general road design and structure that improves the drainage of the roadway and prevents water from accumulating on the roadway. In the step (LD), the road surface (Fr) rises sharply, and the road surface (Fr) of the step portion (Rd) forms a flat surface one step higher than the roadway (Rr). For example, a sidewalk is provided.

Next, with reference to FIG. 5, the step candidate position (CP) is gradually narrowed down from the road surface height data at the step determination position (Pa: measurement line), and finally the step position (CP4) is determined. The procedure will be explained. In this embodiment, the spatial frequency is used as an example of the height change of the road surface (Fr).

FIG. 5 shows the relationship between the height distribution of the road surface (Fr) at the step determination position (Pa) and the spatial frequency (f _{0 to} f _n). The spatial frequency (f _{0 to} f _n ) is a period of height fluctuation with respect to the distance in the vehicle width direction. The intensity of the spatial frequency becomes stronger in the part where there is a step and the height of the road surface changes abruptly. In other words, the intensity of the spatial frequency increases in the part where the change in height appears.

In FIG. 5, a plurality of spatial frequencies (f _k ) are used to detect the step position. k is an integer that varies in the range 0 to n (n is a natural number). Spatial frequency (f ₀ ) indicates the search for the highest frequency, and spatial frequency (f ₀ ) has the narrowest search width. As k increases, the spatial frequency (f _k ) becomes a search for a lower frequency, and the search width becomes wider. The spatial frequency (f _n ) indicates the search for the lowest frequency, and the spatial frequency (f _n ) has the widest search width.

In the high frequency search, the height difference of the road surface can be sensitively detected, and the position of the step (LD) can be accurately detected. However, it also reacts to changes in the height of the road surface due to noise and the like. On the other hand, in the low frequency search, a gentle road surface height difference can be detected, and it is also resistant to noise. However, since the step position is specified in units of a wide search width, the detection error is large.

Therefore, the step candidate narrowing circuit 16 first detects high-intensity position coordinates by searching for a low frequency, then detects a strong position as a step candidate position by searching for a high frequency around it, and gradually increases the height. The search is performed by frequency, and the step detection circuit 18 finally identifies the position of the step (LD) from the result of narrowing down the step candidate positions.

As shown in FIG. 5, the search area circuit 17b (first search area circuit) sets a search area (SR1: first search area) on the step determination position (Pa). The height change circuit 17a (first height change circuit) changes the height of the road surface (Fr) in the first search width (first height) for each position coordinate in the search region (SR1: first search region). Change) is calculated. The height change (first height change) of the road surface (Fr) in the first search width corresponds to the intensity p (y, f) of the _{spatial frequency (f n) shown in FIG.} The height change circuit 17a first calculates the intensity p (y, f) of the _{spatial frequency (f n) having the widest search width.}

The search area circuit 17b (second search area circuit) is a position coordinate in which the height change of the road surface (Fr) in the first search width, that is, the intensity p (y, f) of the _{spatial frequency (f n) is equal to or more than the threshold value.} (CP1) is detected as a step candidate position. The search area circuit 17b sets a new search area (SR2: second search area) with respect to the position coordinates (CP1). Specifically, the second search area (SR2) is set in a region including the position coordinates (CP1) and narrower than the first search region (SR1).

For example, the second search area (SR2) may be set to be included in the first search area (SR1). The width of the second search area (SR2) may be set to the first search width. Alternatively, the width of the second search region (SR2) may be narrowed to half the search width (first search width) of _{the previously searched spatial frequency (f n).}

In the first embodiment, when two or more step candidate positions (CP1) are simultaneously detected from the first search area (SR1), the second step candidate position (CP1) having the largest first height change is found. The search area (SR2) is set. Here, the case of narrowing down from one search area to one step candidate position is shown, but of course, the second search area SR2 is set in all the position coordinates (CP1) where the intensity p (y, f) is equal to or more than the threshold value. It doesn't matter. It is possible to detect steps on the road surface without omission.

The height change circuit 17a (second height change circuit) is the height of the road surface (Fr) in the second search width narrower than the first search width for each position coordinate in the search region (SR2: second search region). Calculate the change (second height change). The height change (second height change) of the road surface (Fr) in the second search width corresponds to the intensity p (y, f) of the _{spatial frequency (f 2) shown in FIG.} Here, n = 3.

In this way, the track boundary detection device determines the step candidate position (CP) by repeating the setting of the search area and the detection of the step candidate position while gradually narrowing the search width (frequency) and the search area (SR). Gradually narrow down and finally determine the position of the step (CP4).

In the present embodiment, the spatial frequency is used as the height change of the road surface (Fr), but the spatial frequency can be obtained by using, for example, the wavelet transform. Create a wavelet with a search width by superimposing waves of the frequency you want to calculate. The higher the frequency, the narrower the search width of the wavelet. By correlating the wavelet with the height change of the road surface, the intensity p (y, f) _{of the spatial frequency (f k) can be obtained.} The height change of the entire road surface can be obtained by obtaining the intensity of each spatial frequency for each position coordinate of the search area. For example, the wavelet transform method disclosed in JP-A-2009-204462 can be used.

In this embodiment, the spatial frequency is taken up as an example of the height change of the road surface (Fr). However, the present invention is not limited to this, and statistical analysis within the search width including, for example, a histogram of the height of the road surface (Fr), the mean value or the standard deviation of the height of the road surface (Fr), is used instead of the spatial frequency. Alternatively, these may be used in combination with the spatial frequency.

An example of the track boundary detection method using the track boundary detection device of FIG. 1 will be described with reference to FIG. The track boundary detection device repeatedly executes a series of information processing cycles for detecting a step on the road surface at predetermined time intervals.

First, in step 01, the laser range finder measures the distance and orientation from the vehicle to the object by detecting the distance and orientation from the vehicle to the surface of the object. Then, the arithmetic circuit 14 acquires the three-dimensional shape of the object around the vehicle by performing coordinate conversion processing from the distance and the direction from the vehicle to the object obtained by the laser range finder.

Proceeding to step S03, k = n is set as the initial value of k of the spatial frequency (f _k). n is a predetermined natural number. The initial value (f _n ) of the spatial frequency is set to the frequency of the target calculated from the shape of the assumed runway boundary target. In the present embodiment, for example, a spatial frequency having twice the width of the target as the wavelength and the height of the target as the intensity (amplitude) is used as the _{initial value (f n).} In the case of a curb (steep step), the wavelength is twice the width of the curb, and in the case of a slope (gentle step), the wavelength is twice the width from the lowest point to the highest point of the slope. .. When a plurality of types of track boundary targets are assumed, the initial value (f _n ) of the spatial frequency may be determined according to the widest target.

Proceeding to step S05, the step determination position circuit 15 sets the step determination position (Pa) with respect to the three-dimensional shape of the object. The search area circuit 17b sets the first search area SR1 on the step determination position (Pa). Then, the height change circuit 17a changes the height of the road surface (Fr) in the first search width, that is, the intensity p (y, f) of the _{spatial frequency (f n) for each position coordinate in the first search region SR1.} Is calculated.

Proceeding to step S07, the search area circuit 17b (second search area circuit) detects the position coordinates (CP1) at which the intensity p (y, f) _{of the spatial frequency (f n) is equal to or higher than the threshold value as the step candidate position.}

The process proceeds to step S09, and it is determined whether or not the step candidate position has been detected. When the step candidate position (CP1) is detected (YES in S09), the process proceeds to step S11. Then, since the spatial frequency (f _n ) is not the smallest frequency (f ₀ ) (NO in S11), the process proceeds to step S17 in order to narrow down the step candidate positions.

In step S17, the search area circuit 17b sets a new search area (SR2: second search area) with respect to the position coordinates (CP1). The process proceeds to step S19, and k-1 is set as a new k. In other words, to set the high spatial frequency by one stage than the spatial frequency _{(f n) (f n-} 1). Along with this, the search width also becomes narrower. After that, returning to step S05, the search area circuit 17b sets the second search area SR2, which is narrower than the first search area SR1. Then, the height change of the road surface (Fr) in the second search width narrower than the first search width, that is, the intensity p (y, f) of the _{spatial frequency (f n-1) is calculated.}

If the cycle consisting of steps S05, S07, S09, S11, S17, and S19 described above can be repeated until k = 0 (YES in S11), the process proceeds to step S15, and the step detection circuit 18 finally detects the step. The step candidate position (CP4) is output as the step (LD) position. In the example of FIG. 5, n = 3 is set, but an integer other than this may be used.

On the other hand, if the step candidate position is not detected in step S09 (NO in S09) before the cycle is repeatedly executed until k = 0, the process proceeds to step S13, and the step detection circuit 18 determines the step determination position (NO). The judgment result that there is no step (LD) on Pa) is output.

The threshold value used in step S07 is set according to the assumed shape of the track boundary target, particularly the height of the track boundary target. Specifically, it is set high when a high track boundary target is assumed, and set low when a low track boundary target is assumed. Alternatively, a plurality of threshold values may be prepared according to the spatial frequency. For example, the higher the frequency, the higher the threshold. As a result, it is possible to prevent erroneous judgment due to changes in noise and the like.

According to the first embodiment, the following effects can be obtained.

By starting the search from a step (low frequency) where the height changes slowly and continuing the search by gradually narrowing down the search area, it is possible to detect a low frequency, that is, a step whose height changes gently, and it is high. A steep step can be detected without being affected by the frequency, that is, the noise whose height changes suddenly.

A second search area is set for the position coordinates whose height change is equal to or greater than the threshold value. Since the position coordinates whose first height change is equal to or greater than the threshold value are extracted as step candidates and the search is continued, it is possible to suppress erroneous detection of steps having low intensity and accurately narrow down the step candidates.

The second search area is set for the position coordinate having the largest first height change among the two or more position coordinates whose height change is equal to or more than the threshold value. As a result, the position coordinates having the highest intensity can be extracted as a step candidate, so that erroneous detection due to the shape of an object constituting the step can be further suppressed, and the position can be narrowed down accurately.

The width of the second search area is set to the first search width. That is, the width of the search region (CP) is _{defined as} the search width of the spatial frequency (f k) used in the previous cycle. As a result, the search area can be narrowed down accurately, and while suppressing the error associated with the calculation of the width of the search area, the calculation processing is small and the processing can be performed at high speed.

The height change circuit 17a calculates the first height change as the first space frequency with respect to the first search width, and the second height change as the second space frequency with respect to the second search width. That is, the height change circuit 17a calculates the height change of the road surface as the cycle of the height change with respect to the distance in the vehicle width direction. As a result, the step on the road surface can be calculated as the frequency intensity, so that even when the step has a complicated shape, the step position can be accurately narrowed down by comparing the intensities.

(Second Embodiment)
The area where the vehicle (Vc) can travel is divided by the step, and the vehicle (Vc) cannot travel beyond the step. Therefore, there is a request to reliably determine the area where there is no step (travelable area). .. When the vehicle (Vc) is on the runway, the road surface of the runway is flat, so that the intensity of the spatial frequency does not increase on the flat surface. Therefore, when a plurality of position coordinates (step candidate positions) whose frequency intensity is higher than the threshold value are found, it is highly possible that the step candidate position closest to the vehicle (Vc) is the road boundary (step). Therefore, it can be determined that at least the area from the vehicle (Vc) to the step candidate position closest to the vehicle (Vc) is a travelable area.

Therefore, in the second embodiment, when two or more step candidate positions (CP1) are simultaneously detected from the first search area (SR1), the search area circuit 17b is among the two or more step candidate positions (CP1). The second search area (SR2) is set for the position coordinate (CP1) closest to the vehicle (Vc).

Specifically, when the first height change is equal to or more than the threshold value in a plurality of position coordinates in the first search area (SR1), the search area circuit 17b is the position coordinates where the first height change is equal to or more than the threshold value. The second search area (SR2) is set for the position coordinate (CP1) closest to the vehicle (Vc) in (CP1). That is, among the plurality of step candidate positions whose height change is equal to or greater than the threshold value, only the step candidate position on the side closer to the vehicle (Vc) is set in the search area in the next cycle. Other points are the same as those of the first embodiment, and the description thereof will be omitted. Alternatively, it may be carried out in combination with the first embodiment. That is, the second search area (SR2) may be set at the same time for the position coordinate having the largest change in the first height and the position coordinate (CP1) closest to the vehicle (Vc).

Since the position coordinates (CP) closest to the vehicle (Vc) can be extracted as the step candidate position, it is possible to reliably determine the area (travelable area) around the vehicle (Vc) where there is no step. The other points of the second embodiment are the same as those of the first embodiment, and the description thereof will be omitted.

(Third Embodiment)
In FIGS. 2 and 3, a linear step determination position (Pa) extending in the vehicle width direction is illustrated, but a loop-shaped loop surrounding the vehicle is used by using a 360-degree radar capable of scanning all directions in the horizontal plane. The step determination position (Pa) may be set. 7 and 8 show an example in which a radar laser 12 capable of rotating 360 degrees is mounted on a vehicle (Vc) and a plurality of circular step determination positions (Pa1 to Pa4) are set. As described above, the linear step determination position (Pa) is not limited to a straight line, but may be a curved line, and the curved line may form a closed loop. Further, the linear step determination position (Pa) may be not limited to a single number but may be a plurality of positions. The other points of the third embodiment are the same as those of the first embodiment, and the description thereof will be omitted.

Each function shown in each of the above embodiments may be implemented by one or more processing circuits. The processing circuit includes a programmed processing device such as a processing device including an electric circuit. Processing devices also include devices such as application specific integrated circuits (ASICs) and conventional circuit components arranged to perform the functions described in the embodiments.

1 Track boundary detection device 11 Height detection sensor 12 Distance measurement sensor (radar laser, stereo camera)
13 Microcomputer 14 Calculation circuit 15 Step determination position circuit 16 Step candidate narrowing circuit 17a Height change circuit 17b Search area circuit 18 Step detection circuit CP1 to CP4 Step candidate position (position coordinates)
Da Predetermined direction F Road surface LD Step Pa, Pa1 to Pa4 Step determination position SR1 First search area SR2 Second search area Vc Vehicle

Claims (7)

A height detection sensor that detects the height of the road surface around the vehicle, and a step that sets a step determination position on the road surface that is separated from the height detection sensor by a predetermined distance in a predetermined direction and extends in a direction orthogonal to the predetermined direction. It is a track boundary detection method for detecting the position of a step on the road surface by using a determination position circuit.
A first search area is set on the step determination position, and
For each position coordinate in the first search area, the first height change of the road surface in the first search width is calculated.
Based on the first height change, a second search area narrower than the first search area is set on the step determination position.
For each position coordinate in the second search region, the second height change of the road surface in the second search width narrower than the first search width is calculated .
The cycle from the setting of the first search region to the calculation of the second height change is repeated until the second search width becomes the narrowest width.
Lane boundary detection method, wherein the second search width based on the previous SL second height change when the narrowest, to detect the position of the previous SL step. The track boundary detection method according to claim 1, wherein the second search region is set for the position coordinates whose first height change is equal to or greater than a threshold value. The first aspect of claim 1, wherein the second search region is set for the position coordinate closest to the vehicle among the two or more position coordinates whose first height change is equal to or more than a threshold value. Track boundary detection method. A claim characterized in that the second search region is set for the position coordinate having the largest first height change among the two or more position coordinates whose first height change is equal to or more than a threshold value. Item 4. The runway boundary detection method according to Item 1 or 3. The track boundary detection method according to any one of claims 1 to 4, wherein the width of the second search region is set to the first search width. Claims 1 to 5 are characterized in that the first height change is calculated as the first space frequency with respect to the first search width, and the second height change is calculated as the second space frequency with respect to the second search width. The track boundary detection method according to any one of the above. A height detection sensor mounted on the vehicle that detects the height of the road surface around the vehicle, and a step determination position that is separated from the height detection sensor by a predetermined distance in a predetermined direction and extends in a direction orthogonal to the predetermined direction. Is a track boundary detection device including a step determination position circuit for setting the above on the road surface.
A first search area circuit that sets the first search area on the step determination position,
For each position coordinate in the first search region, a first height change circuit that calculates the first height change of the road surface in the first search width, and
A second search area circuit that sets a second search area narrower than the first search area on the step determination position based on the first height change, and
A second height change circuit that calculates a second height change of the road surface in a second search width narrower than the first search width for each position coordinate in the second search region .
Among the second height variation which calculates repeatedly the second height changing circuit, based on the previous SL second height change when the second search width is the narrowest, the position of the step in front SL on the road surface A track boundary detection device comprising: a step position detection circuit for detecting.

Images