Stop line recognition device

Abstract

To prevent false detection of a stop line when determining the stop line from a luminance difference of a road surface.SOLUTION: A camera unit 2 has a camera portion 11 and a preceding driving environment recognizer 13. The preceding driving environment recognizer 13 obtains a luminance difference of a road surface from an image acquired by the camera portion 11, detects section line(s) intersecting a stop line candidate when a stop line candidate is detected from the luminance difference of the road surface, examines whether or not a portion sandwiched between dark portions is detected in a preset search range of the section line, and determines that the stop line candidate is not a stop line when the portion sandwiched between dark portions is detected.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a stop line recognition device capable of preventing the erroneous detection of a stop line in advance.

Background Art

[0002] In recent years, various driving support devices for reducing the burden on drivers have been proposed for vehicles such as automobiles, and some have already been put into practical use. For example, there is also known a driving support device that recognizes a stop line drawn on the road ahead based on an image captured by an in-vehicle camera mounted on the host vehicle, automatically stops the host vehicle immediately before the stop line, or prompts the driver to decelerate.

[0003] As a technique for recognizing a stop line in this type of driving support device, for example, those disclosed in Patent Document 1 (Japanese Patent Laid-Open No. 6-233301) and Patent Document 2 (Japanese Patent Laid-Open No. 2015-179482) are known. In the techniques disclosed in these documents, first, when trying to detect a stop line (a white line in Patent Document 1 and a lane boundary line in Patent Document 2) based on an image captured by an in-vehicle camera, the luminance between the road surface and the stop line is detected for each pixel in the pixel column of the image. Then, a luminance value histogram is generated based on the detected luminance for each pixel, and when a bimodal distribution of the bright peak luminance value and the dark peak luminance value appears in the generated luminance value histogram, it is determined as a stop line.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the technologies disclosed in the aforementioned Patent Documents 1 and 2, even when shadows of three-dimensional objects such as two parallel utility poles are projected onto the road surface during the day, or when light rays from streetlights are projected onto the road surface at night, a bimodal distribution may appear in the luminance value histogram, potentially leading to misdetection as a stop line.

[0006] In other words, when the shadows of two utility poles are projected onto the road surface, the luminance value of each pixel in the image captured by the on-board camera will be lower in the section corresponding to the shadow of the utility poles than in the luminance value of the bare road surface. Therefore, the luminance value histogram will show a bimodal distribution with a dark peak luminance value in the shadowed area and a bright peak luminance value of the road surface in between the shadows. Also, when a single streak of light is projected onto the road surface at night, the luminance values ​​of the road surface on both sides of the light will be lower than the luminance value of the streak of light. Therefore, the luminance value histogram will show a bimodal distribution with a bright peak luminance value in the streak of light and a dark peak luminance value of the bare road surface on the adjacent side. As a result, in either of the above cases, if one attempts to determine the stop line from the bimodal distribution in the luminance value histogram, there is a possibility of misidentification as the stop line.

[0007] As a countermeasure, for example, Patent Document 3 (Japanese Patent Publication No. 2016-4287) discloses a technique that determines that a white line (stop line) has been falsely detected if the luminance (average luminance or representative luminance) of the candidate area for a white line (stop line) and the area of ​​the bare road surface are approximately the same. In other words, in this document, when an area between two shadows projected onto the road surface is detected as a candidate area for a white line, a road surface area parallel to this candidate area for a white line is set. Since both this candidate area for a white line and the road surface area are the bare road surface, the luminance of both areas will be approximately the same, and the difference in luminance between the two areas will be almost zero. Therefore, if the difference in luminance between the two areas is smaller than a preset threshold, the candidate area for a white line is determined to be a false detection.

[0008] However, in the technology disclosed in Patent Document 3, if the stop line is faded, the difference in brightness between the candidate area for the stop line and the road surface area in the faded portion becomes negligible. As a result, an undesirable situation arises where an area that is actually a stop line is excluded from the list of stop line candidates.

[0009] The present invention aims to provide a stop line recognition device that can prevent erroneous detection of stop lines when determining stop lines from differences in brightness on the road surface. [Means for solving the problem]

[0010] The present invention relates to a stop line recognition device comprising: a driving environment information acquisition unit that acquires driving environment information in front of the vehicle; and a driving environment recognition unit that recognizes the driving environment in front of the vehicle based on the driving environment information acquired by the driving environment information acquisition unit, wherein the driving environment recognition unit comprises: a stop line candidate detection unit that detects stop line candidates from the brightness difference of the road surface based on the driving environment information acquired by the driving environment information acquisition unit; and a stop line determination unit that determines whether the stop line candidate detected by the stop line candidate detection unit is a stop line, wherein the driving environment recognition unit further determines whether the stop line candidate detected by the stop line candidate detection unit is a stop line The system includes a lane line detection unit that detects lane lines that intersect the road, a luminance distribution processing unit that determines the luminance distribution of the lane lines detected by the lane line detection unit, and a stop line false detection luminance pattern detection unit that, based on the luminance distribution of the road surface determined by the luminance distribution processing unit, checks whether a bright area sandwiched between dark areas has been detected in a preset search range of the lane lines detected by the lane line detection unit. The stop line determination unit determines that if a bright area sandwiched between dark areas has been detected in the search range by the stop line false detection luminance pattern detection unit, the stop line candidate detected by the stop line candidate detection unit is not a stop line. [Effects of the Invention]

[0011] According to the present invention, when a candidate stop line is detected from the brightness difference of the road surface, a lane marking that intersects this candidate stop line is detected, and it is checked whether a bright area sandwiched between dark areas is detected within a preset search range of the lane marking based on the brightness distribution of the road surface. If a bright area sandwiched between dark areas is detected, it is determined that the candidate stop line is not a stop line. Therefore, when determining a stop line from the brightness difference of the road surface, it is possible to prevent false detection of a stop line. [Brief explanation of the drawing]

[0012] [Figure 1] Schematic diagram of the driver assistance system [Figure 2] Flowchart showing the stop line candidate evaluation processing routine [Figure 3] A flowchart showing the processing routine for detecting gaps in dark areas. [Figure 4] An overhead view showing a situation where the stop line in front of the vehicle is faded. [Figure 5] This is an explanatory diagram showing the forward view in the state shown in Figure 4, as captured by the in-vehicle camera. [Figure 6] An overhead view showing two shadows projected onto the road surface in each lane. [Figure 7A] This is an explanatory diagram showing the forward view captured by the onboard camera in the state shown in Figure 6. [Figure 7B] This diagram shows a sequence of points sampled along the grid lines, based on the image in Figure 7A. [Figure 7C] An explanatory diagram showing a sequence of points sampled along the lane markings near the stop line. [Figure 8] An explanatory diagram showing a nighttime image of the area ahead captured by an in-vehicle camera. [Figure 9] An overhead view showing a beam projected from a vehicle in an adjacent lane onto the vehicle's own lane. [Figure 10] This is an explanatory diagram showing the forward view captured by the in-vehicle camera in the state shown in Figure 9. [Figure 11A] Histogram of luminance values ​​on a dividing line with a bimodal frequency distribution [Figure 11B] Histogram of luminance values ​​on a dividing line with a unimodal frequency distribution on the bright side. [Figure 11C] The luminance value histogram on the dividing line having a unimodal frequency distribution on the dark part side [Figure 12] Explanatory diagram showing the luminance pattern in the image search direction at the stop line candidate [Figure 13] Explanatory diagram showing the luminance pattern in the image search direction in the dividing line search area

Mode for Carrying Out the Invention

[0013] Hereinafter, an embodiment of the present invention will be described based on the drawings. The driving support device 1 shown in FIG. 1 is mounted on the host vehicle M (see FIG. 4). This driving support device 1 includes a camera unit 2 having a function as a stop line recognition device and a driving support control unit 3, and these two units 2 and 3 are connected so as to be capable of two-way communication through an in-vehicle communication line such as CAN (Controller Area Network) communication.

[0014] This driving support control unit 3 and the forward driving environment recognition unit 13 of the camera unit 2 described later are each composed of a microcontroller including a CPU, a RAM, a ROM, a rewritable nonvolatile memory (flash memory or EEPROM), and peripheral devices. Programs and fixed data necessary for the CPU to execute each process are stored in the ROM. The RAM is provided as a work area for the CPU, and various data in the CPU are temporarily stored. The CPU is also called a MPU (Microprocessor) or a processor. Instead of the CPU, a GPU (Graphics Processing Unit) or a GSP (Graph Streaming Processor) may be used. Alternatively, the CPU, the GPU, and the GSP may be selectively combined and used.

[0015] Furthermore, the camera unit 2 includes a camera unit 11 as a driving environment information acquisition unit, an image processing unit 12, and a forward driving environment recognition unit 13. The camera unit 11 is a stereo camera consisting of a main camera 11a and a sub-camera 11b. Both cameras 11a and 11b are installed at equal intervals with a predetermined baseline length from the center in the vehicle width direction, for example, above the rearview mirror and close to the windshield. Each of these cameras 11a and 11b is equipped with an image sensor such as a CCD or CMOS, and these two image sensors capture the driving environment in front of the vehicle, including the driving lane of the vehicle M. The camera unit 11 captures a reference image with the main camera 11a and a comparison image with the sub-camera 11b. Note that the symbol If in Figures 4, 6, and 9 simply indicates the field of view area that can be captured by the camera unit 11.

[0016] The pair of analog images captured by both cameras 11a and 11b are processed by the image processing unit 12 and then output to the forward driving environment recognition unit 13. The forward driving environment recognition unit 13 performs various image processing operations, such as those shown below, for each frame based on the reference image data and comparison image data from the image processing unit 12.

[0017] The forward driving environment recognition unit 13 first sequentially extracts small regions, for example, 4x4 pixels, from the reference image, compares the brightness or color pattern of each small region with a comparison image to find the corresponding region, and determines the distance distribution across the entire reference image. Furthermore, the forward driving environment recognition unit 13 examines the brightness difference between each pixel on the reference image and adjacent pixels, extracts those whose brightness differences exceed a threshold as edge points, and generates a distance image (a distribution image of edge points with distance information) by assigning distance information to the extracted pixels (edge ​​points).

[0018] The forward driving environment recognition unit 13 then performs a well-known grouping process on the generated distance image and compares it with a pre-stored three-dimensional frame (window) to recognize forward driving environment information such as lane markings separating the left and right sides of the lane in which the vehicle M is traveling, preceding vehicles, stop lines, intersections, traffic lights, and pedestrians. This forward driving environment information recognized by the forward driving environment recognition unit 13 is output to the driver assistance control unit 3.

[0019] Furthermore, a vehicle status sensor 21 is connected to the input side of the driver assistance control unit 3. This vehicle status sensor 21 is a general term for a group of sensors that detect various states related to the vehicle M. These vehicle status sensors 21 include a vehicle speed sensor that detects the vehicle speed (vehicle speed) of the vehicle M, a steering angle sensor that detects the steering angle of the vehicle M, an acceleration sensor that detects the longitudinal acceleration acting on the vehicle body, a yaw rate sensor that detects the yaw rate acting on the vehicle body, an accelerator opening sensor that detects the amount the accelerator pedal is pressed, and a signal from a brake switch that turns ON when the brake pedal is pressed.

[0020] Furthermore, a control actuator 31 is connected to the output side of the driver assistance control unit 3. This control actuator 31 is a general term for various actuators that support the driver's operation by controlling the driving state of the vehicle M in accordance with the control signals from the driver assistance control unit 3. Examples of these control actuators include an EPS actuator that drives the electric power steering (EPS), a power actuator that controls the driving force of the drive source (engine, electric motor, etc.), and a brake actuator that controls the braking force by adjusting the brake fluid pressure supplied to the brake unit.

[0021] The driver assistance control unit 3 performs some or all of the driving operations (steering, acceleration / deceleration, and braking) performed by the driver, based on the forward driving environment information output from the forward driving environment recognition unit 13 and the state of the vehicle M detected by the vehicle state sensor 21. While adaptive cruise control (ACC) and active lane keep bouncing (ALKB) control are known driver assistance controls that perform at least some of the driver's actions, their explanation is omitted as they are well-known.

[0022] Incidentally, the forward driving environment recognition unit 13, as shown in Figure 5, reads the brightness value of each pixel in the i-direction of a row of horizontal lines j in image T for each frame, and sequentially switches these values ​​from the lowest horizontal line j to the upper horizontal lines j to read the brightness value of the entire image T. At that time, when extracting a stop line from the brightness of the image, as shown in Figure 12, the brightness values ​​of adjacent pixels above and below the sequentially switched horizontal lines j are compared with a preset stop line determination threshold and binarization is performed. Then, the bright region sandwiched between the starting point Pi, where it switches from dark to bright, and the ending point Po, where it switches from bright to dark, is estimated to be the stop line, and this region is extended in the lane width direction to set the stop line candidate Ls'. Therefore, in image T, the stop line candidate Ls' is detected from the brightness pattern of the road surface (dark → bright → dark).

[0023] As shown in Figure 5, even if the stop line Ls is faded in the middle due to friction with tires as vehicles pass, the candidate stop line Ls' is set continuously in the lane width direction, but the brightness of the faded portion of the candidate stop line Ls' is almost the same as the brightness of the road surface. However, if the driver assistance control unit 3 excludes the candidate stop line Ls' from the target stop line due to the faded stop line, the driver assistance control described above will be impaired.

[0024] On the other hand, as shown in Figure 6, when two utility poles P1 and P2 erected on the shoulder or sidewalk of a driving lane are illuminated by light from a light source S such as the sun, and their two shadows P1s and P2s are projected parallel to the road surface of the driving lane, the area between the two shadows P1s and P2s becomes relatively brighter. Therefore, as shown in Figure 7A, when the forward driving environment recognition unit 13 attempts to extract a stop line from the brightness of the image T for each frame, similar to Figure 5 described above, the brightness pattern of the road surface becomes "dark → bright → dark" in the shadows P1s and P2s, and a stop line candidate Ls' is set. However, if the forward driving environment recognition unit 13 determines the stop line candidate Ls' set in the area between the two shadows P1s and P2s as a stop line, it will cause problems with the driving assistance control in the driving assistance control unit 3.

[0025] Furthermore, at night, the road surface may be illuminated by lighting, causing the illuminated areas to become "bright," and a "dark → bright → dark" brightness pattern may be formed on the road surface. For example, as shown in Figure 8, the lighting from streetlights Li installed on the shoulder of the road or sidewalk of the driving lane may illuminate the road surface, and this lighting may form a "dark → bright → dark" brightness pattern on the road surface. In such cases as well, the forward driving environment recognition unit 13 sets a candidate stop line Ls' on the road surface based on the "dark → bright → dark" brightness pattern.

[0026] Furthermore, as shown in Figure 9, during nighttime driving, a beam Be may be projected from the headlights (high beam, low beam, auxiliary lights, etc.) of a vehicle (adjacent vehicle) N traveling in a lane adjacent to the lane of the vehicle M, towards the direction of the vehicle M's lane. If this adjacent vehicle N slows down or stops before the stop line Ls, as shown in Figure 10, a "dark → bright → dark" brightness pattern is formed on the road surface in the image T of one frame due to the illumination of the beam Be. As a result, the forward driving environment recognition unit 13 sets the stop line candidate Ls' in the area illuminated by this beam Be.

[0027] Therefore, the forward driving environment recognition unit 13 is equipped with a function to evaluate whether or not the stop line candidate Ls' set on the image T is a stop line. Specifically, this function of evaluating stop line candidates provided in the forward driving environment recognition unit 13 is executed according to the stop line candidate evaluation processing routine shown in Figure 2.

[0028] In this routine, first, in step S1, forward driving environment information acquired by the camera unit 11 and processed according to a predetermined image by the image processing unit 12 is read. Next, the process proceeds to step S2, where a candidate stop line Ls' is detected based on this forward driving environment information. The processing in step S2 corresponds to the stop line candidate detection unit of the present invention. Note that the detection of this candidate stop line Ls' has been described previously, so the explanation is omitted here. Alternatively, this candidate stop line Ls' may be detected according to a different program.

[0029] Next, the process proceeds to step S3 to check whether a candidate stop line Ls' has been detected. If a candidate stop line Ls' has been detected, the process proceeds to step S4. If a candidate stop line Ls' has not been detected, the routine is exited. In the following, the left and right lane markings Ll and Lr may be collectively referred to as lane marking line L.

[0030] Proceeding to step S4, the system detects the left and right lane markings Ll and Lr that demarcate the driving lane near the designated stop line candidate Ls'. Various methods for detecting these lane markings L are known, for example, as disclosed in Japanese Patent Application Publication No. 2022-60118, already filed by the present applicant. The method for detecting lane markings L disclosed in the said publication will be briefly explained below. Note that the processing in step S4 corresponds to the lane marking detection unit of the present invention.

[0031] In step S4, first, a search is performed along the horizontal line j of image T, starting from the center of the i-direction of image T and moving outwards. Then, edge points where the brightness of pixels inside the vehicle M's lane is relatively lower than the brightness of pixels outside, and where the derivative of the brightness, which indicates the amount of change, is greater than or equal to the set threshold on the positive side, are extracted as the starting points of the lane lines. Also, edge points where the brightness of pixels inside is relatively higher than the brightness of pixels outside, and where the derivative of the brightness, which indicates the amount of change, is less than or equal to the set threshold on the negative side, are extracted as the ending points of the lane lines.

[0032] Next, the starting and ending points of the boundary lines are plotted on the image as candidate boundary line points. Based on the sequence of points sampled from the bottom to the top of the image while switching the horizontal line j, an approximation line representing the edge (boundary) of the boundary line L is calculated. Then, the boundary line L for the current frame is generated using this approximation line. In this case, it may also be possible to determine whether the boundary line L is a solid line or a dashed line. If this sequence of candidate boundary line points is not set, the boundary line will not be detected.

[0033] Next, the process proceeds to step S5, where it is checked whether a lane marking L (Ll or Lr) is detected on at least one side. If no lane marking L is detected on either the left or right side, the routine is exited without determining whether the candidate stop line Ls' is a false detection. In this case, the driver may be notified that it was not possible to determine whether the candidate stop line Ls' is an actual stop line Ls.

[0034] Furthermore, if at least one of the left and right dividing lines Ll, ​​Lr is detected, the process proceeds to step S6. In step S6, the gap in the dark area of ​​dividing line L is detected. This gap in the dark area of ​​dividing line L refers to the region where the brightness due to dividing line L is considered a bright area, and it is sandwiched between shadows P1s and P2s on both sides.

[0035] The gap in the dark area is detected according to the dark area gap detection processing subroutine shown in Figure 3. If the left and right dividing lines Ll and Lr are detected, the gap in the dark area may be detected using each dividing line Ll and Lr. Alternatively, it may be detected based on one of the dividing lines L. However, even if the left and right dividing lines Ll and Lr are detected, as shown in image T of Figure 7A, for example, the left dividing line Ll may be crossed by shadows P1s and P2s, while the tip of shadow P1s may slightly overlap the right dividing line Lr. In such cases, detecting the gap in the dark area individually using both dividing lines Ll and Lr can prevent false detection of the stop line with higher accuracy.

[0036] In this subroutine, first, in step S11, a luminance value histogram is created based on the luminance of the lane markings L near the candidate stop line, with the luminance value on the horizontal axis and the count value (frequency) on the vertical axis. Note that the processing in step S11 corresponds to the luminance distribution processing unit of the present invention.

[0037] This luminance value histogram is first created by setting a search range (boundary line search range) Sf around the boundary line L (Ll in the figures) detected near the candidate stop line Ls', as shown in Figures 7B and 7C. This boundary line search range Sf is set to an area of ​​a specified length extending vertically, approximately centered on the width of the candidate stop line Ls'. The upper limit of this length is approximately 3.5 / 2 [m] vertically, approximately centered on the width of the candidate stop line Ls'.

[0038] Then, within the vertical region of this boundary line search range Sf, the brightness of each pixel Pl is detected along approximately the center of the boundary line L. If boundary lines L are detected on both the left and right sides, the boundary line search range Sf is set individually for the left and right boundary lines Ll and Lr, the brightness of the pixels Pl on each boundary line Ll and Lr is detected, and the brightness is counted for each bin of assigned brightness values ​​(see Figures 11A to 11C).

[0039] Next, the process proceeds to step S12, where a brightness threshold (brightness threshold) is set based on the frequency distribution of the brightness value histogram. As shown in Figures 11A to 11C, this brightness threshold is set by first searching the brightness value histogram from both the bright and dark sides, and starting from the first bin that exceeds a predetermined minimum count. If there are more than two consecutive bins that do not meet the minimum count, the bins from the starting point to the last bin that exceeds the minimum count are identified as a peak.

[0040] Next, a brightness threshold is set according to the number of identified peaks. For example, as shown in Figures 7A and 7B, when two shadows P1s and P2s cross the left dividing line Ll, the brightness of the dividing line L(Ll) in the area of ​​shadows P1s and P2s is low (dark area), and the brightness of the left dividing line Ll sandwiched between shadows P1s and P2s is high (bright area). Therefore, two different peaks (bimodality) are detected on the bright and dark sides, as shown in Figure 11A.

[0041] In contrast, if the area near the stop line Ls detected in image T of Figure 7A is set as the candidate stop line Ls', and as shown in Figure 7C, the shadows P1s and P2s are not projected onto the boundary line L (Ll in the figure) within the boundary line search range Sf corresponding to this candidate stop line Ls', the luminance histogram will be as shown in Figure 11B. That is, this luminance histogram will have a unimodal peak on the bright side. Similarly, even if the tip of the shadow P1s slightly overlaps the right boundary line Lr shown in image T of Figure 7A, the luminance histogram will have the shape shown in Figure 11B.

[0042] When the forward driving environment recognition unit 13 detects two peaks (bimodality) in the frequency distribution of the luminance histogram, it detects the endpoints of the last bins of both peaks (dark endpoint, bright endpoint), as shown in Figure 11A. Next, it sets the luminance value of the midpoint between these endpoints as the brightness threshold. On the other hand, when the frequency distribution of the luminance histogram has one peak (unimodality), the forward driving environment recognition unit 13 compares the number of bins remaining without reaching the minimum count on the bright side and the dark side, with the detected peak in between, and sets the endpoint of the peak with the larger number of remaining bins as the brightness threshold. Therefore, in Figure 11B, a brightness threshold with relatively low luminance is set. Also, in Figure 11C, a brightness threshold with relatively high luminance is set. Note that the above setting of the brightness threshold is just an example and is not limited to this.

[0043] Subsequently, the program proceeds to step S13, where it classifies the pixels in the boundary line search area Sf based on the brightness threshold set in step S12. This brightness classification of each pixel is performed, for example, by searching the image T shown in Figure 7A from bottom to top while switching the horizontal line j, and binarizing the pixels in the boundary line search area Sf using the brightness threshold.

[0044] As a result, for example, in image T in Figure 7A, the parts corresponding to shadows P1s and P2s become the dark areas shown in Figure 13, and the brightness and darkness of pixels within the boundary line search range Sf in image T become a brightness pattern of "bright → dark → bright → dark → bright" from the bottom to the top of image T, with bright areas detected in the gaps between the dark areas. In addition, a candidate stop line Ls' is set near the stop line Ls in Figure 7A, and in the boundary line search range Sf set intersecting this candidate stop line Ls', the brightness threshold is set relatively low, so the brightness pattern of the boundary line L becomes almost entirely "bright".

[0045] Next, proceeding to step S14, the system checks whether the brightness pattern is in a "gap between dark areas" based on the brightness and darkness of the pixels in the section line search area Sf defined in step S13, and then proceeds to step S7 in Figure 2.

[0046] In step S14, if the luminance pattern has "dark area → bright area → dark area" and the width of the bright area of ​​this luminance pattern (the vertical width of image T) falls within a preset range, this bright area is set as the "gap between dark areas". Also, if the luminance pattern is not "dark → bright → dark", or if the width of the bright area falls outside the preset range, that is, if the luminance pattern of the dividing line L is almost entirely "bright area" or almost entirely "dark area", the gap between dark areas is determined not to be detected. In this case, the upper limit of the width of the bright area is, for example, about 3.5 [m]. Note that the processing in steps S12 to S14 corresponds to the stop line misdetection luminance pattern detection unit of the present invention.

[0047] Subsequently, proceeding to step S7 in Figure 2, the system checks whether a gap in the dark area was detected in step S6. If a gap in the dark area is detected, it is determined that the stop line candidate Ls' is a false detection, and the system proceeds to step S8, where the gap detection flag F is set (F←1), and the routine is exited. On the other hand, if it is determined that no gap in the dark area was detected, it is determined that the stop line candidate Ls' is a stop line, and the system branches to step S9, where the gap detection flag F is cleared (F←0), and the routine is exited. Note that the processing in steps S7 to S9 corresponds to the stop line determination unit of the present invention.

[0048] The value of the dark area gap detection flag F is read by the forward driving environment recognition unit 13 when it recognizes a candidate stop line Ls'. If F=1, the recognized candidate stop line Ls' is not an actual stop line and is recognized as a false detection. As a result, the forward driving environment recognition unit 13 does not output information about the stop line Ls to the driver assistance control unit 3, and therefore the driver assistance control unit 3 continues driving without performing driver assistance controls such as automatic stopping control to stop the vehicle M at the stop line or notification control to prompt the driver to decelerate.

[0049] Furthermore, if the dark area gap detection flag F is cleared (F=0), the forward driving environment recognition unit 13 recognizes the recognized stop line candidate Ls' as the stop line Ls and outputs information about the stop line Ls to the driver assistance control unit 3. As a result, the driver assistance control unit 3 operates the control actuator 31 and performs driver assistance control as usual, such as decelerating and automatically stopping the vehicle M before the stop line Ls.

[0050] As explained above, in this embodiment, first, the forward driving environment recognition unit 13 searches for a brightness pattern on the road surface based on the forward driving environment information. If a "dark → light → dark" brightness pattern is detected on the road surface, the bright region between the starting point Pi where the brightness pattern switches from dark to light and the ending point Po where it switches from light to dark is extended in the lane width direction to set a candidate stop line Ls'. Next, the left and right lane markings Ll and Lr that demarcate the driving lane are detected near this candidate stop line Ls' and intersect with this candidate stop line Ls'.

[0051] Furthermore, if a dark gap is detected between two shadows P1s and P2s covering the boundary line Ll and Lr, the candidate stop line Ls' is determined not to be a stop line, thus preventing false detection of stop lines. In addition, since the candidate stop line Ls' is detected based on a "dark → light → dark" luminance pattern, as before, the stop line Ls can be correctly detected even if part of it is faded.

[0052] Furthermore, the present invention is not limited to the embodiments described above. For example, the brightness threshold set based on the luminance histogram may be improved by adding or multiplying different coefficients for daytime and nighttime. [Explanation of Symbols]

[0053] 1…Driving assistance system, 2...Camera unit, 3…Driving support control unit, 11…Camera Department, 11a... Main camera, 11b... Sub-camera, 12…Image processing unit, 13…Forward driving environment recognition unit, 21... Vehicle status sensor, 31... Control actuator, F... Narrowing detection flag, If…Angle of view area, Ls'... Candidate stop line, Li... streetlights, Ll...left lane line, Lr...right lane line, Ls…stop line, M... Own vehicle, N...Adjacent vehicle, P1s, P2s... Shadow, P1, P2... utility poles, Pi…starting point, Pl...pixel, Po... End point, S...Light source, Sf... Search range for lane markings, T...Image

Claims

1. A driving environment information acquisition unit that acquires driving environment information in front of the vehicle, A driving environment recognition unit recognizes the driving environment in front of the vehicle based on the driving environment information acquired by the driving environment information acquisition unit. It has, The aforementioned driving environment recognition unit is A stop line candidate detection unit detects a stop line candidate from the brightness difference of the road surface based on the driving environment information acquired by the driving environment information acquisition unit, A stop line determination unit determines whether the stop line candidate detected by the stop line candidate detection unit is a stop line or not. In a stop line recognition device equipped with, The aforementioned driving environment recognition unit further, A section line detection unit for detecting section lines that intersect the stop line candidate detected by the stop line candidate detection unit, A luminance distribution processing unit that determines the luminance distribution of the boundary lines detected by the boundary line detection unit, Based on the luminance distribution of the road surface obtained by the luminance distribution processing unit, the stop line misdetection luminance pattern detection unit checks whether a bright area sandwiched between dark areas has been detected within a preset search range of the lane line detected by the lane line detection unit. Equipped with, The stop line determination unit determines that if the stop line false detection brightness pattern detection unit detects a bright area sandwiched between the dark areas within the search range, the stop line candidate detection unit has detected the stop line candidate and that it is not a stop line. A stop line recognition device characterized by the following features.

2. The stop line misdetection brightness pattern detection unit checks whether the bright area sandwiched between the dark areas has been detected based on at least one of the left and right dividing lines that intersect the stop line candidate set by the stop line candidate detection unit. The stop line recognition device according to claim 1, characterized in that it is a stop line recognition device.

3. The stop line misdetection luminance pattern detection unit determines the frequency distribution of luminance values ​​based on the luminance distribution of the road surface obtained by the luminance distribution processing unit, calculates a brightness threshold based on the frequency distribution of luminance values, and distinguishes between the bright areas and the dark areas based on the brightness threshold. The stop line recognition device according to claim 1, characterized in that it is a stop line recognition device.

4. The stop line misdetection brightness pattern detection unit sets the midpoint between the two peaks as the brightness threshold when the frequency distribution of the brightness value is bimodal. The stop line recognition device according to claim 3.

5. The aforementioned light-dark threshold is set to different values ​​during the day and at night. The stop line recognition device according to claim 3 or 4.

Images