JP5152002B2 - Lane recognition device and lane recognition method - Google Patents

Description

  The present invention relates to a lane recognition device and method for recognizing a lane on a road.

  Conventionally, a method of recognizing a lane by detecting a roadway based on a captured image is known. The road fence is, for example, a thing such as Botts Dots or Cats Eye, and is discretely arranged on the road.

  The shape of the road fence changes depending on the lighting environment such as how the light hits during sunshine. Thus, for example, Patent Document 1 discloses an apparatus that improves detection accuracy regardless of the illumination environment by performing a morphological operation on a captured image.

JP 2007-141052 A

  However, according to the technique disclosed in Patent Document 1, since road noise is removed or noise is removed in order to remove noise based on the width of an object in an arbitrary horizontal line (scanning line). There is an inconvenience that cannot be completely removed. For this reason, there has been a problem in that it is not possible to detect a roadway stably.

  The present invention has been made in view of such circumstances, and an object thereof is to detect a roadway with high accuracy.

In order to solve such a problem, the present invention generates an edge image obtained by extracting a luminance edge in a captured image obtained by imaging a traveling road surface, and in this edge image, a candidate area which is an area extracted as an edge is generated. On the other hand, this region is expanded in all directions. At this time, the candidate area is expanded to such an extent that the candidate area corresponding to the road fence is not connected to the candidate area corresponding to the non-road fence. Then, the road candidate defining the lane on the road is detected based on the aspect ratio and the area with the expanded candidate area as a processing target, and thereby the lane on the road is determined based on the detected road fence. recognize.

  According to the present invention, the road fence and other elements can be clearly separated in terms of area and shape (aspect ratio) by expansion processing in all directions. Thereby, a roadway can be detected effectively.

1 is a configuration diagram schematically showing a lane recognition device 1 according to a first embodiment. Explanatory drawing showing vehicle C and imaging range The flowchart which shows the procedure of the lane recognition process concerning 1st Embodiment. Explanatory drawing which shows an example of an edge image typically Illustration of aspect ratio and area of white area Explanatory drawing which shows typically the captured image in the state which turned on the headlight 31 of the vehicle. The block diagram which shows schematically the lane recognition apparatus 1 concerning 2nd Embodiment. The flowchart which shows the procedure of the lane recognition process concerning 2nd Embodiment. The block diagram which shows schematically the lane recognition apparatus 1 concerning 3rd Embodiment.

(First embodiment)
FIG. 1 is a configuration diagram schematically showing a lane recognition device 1 according to a first embodiment of the present invention. The lane recognition device 1 according to the present embodiment is a device that recognizes a lane by detecting a roadway provided on a road using image processing. The lane recognition device 1 mainly includes a control unit 10, an image memory 20, and a camera 30, and the control unit 10 is connected to the camera 30 via the image memory 20. In the present specification, the term road ridge is used as a generic term for things that are discretely arranged on the road in order to demarcate the lane, such as botsdots, reflectors, and cat's eyes.

  The control unit 10 reads a captured image corresponding to one frame stored in an image memory 20 described later, and recognizes a lane by detecting a roadway based on the captured image. As the control unit 10, for example, a microcomputer mainly composed of a CPU, a ROM, a RAM, and an I / O interface can be used. In addition, an in-vehicle LAN is connected to the control unit 10, and the control unit 10 detects whether the headlight 31 (see FIG. 2) is in a lighting state or an extinguishing state through the in-vehicle LAN. Can do.

  The control unit 10 includes a candidate image generation unit 11, a roadway detection unit 12, a distortion correction unit 13, a lane recognition unit 14, and an imaging range change unit 15 when this is functionally captured. Yes. The candidate image generation unit 11 generates a captured image input to the image memory 20, that is, an edge image from which a luminance edge in the captured image captured by the camera 30 is extracted (first processing unit). In addition, the candidate image generation unit 11 expands this region in all directions with respect to a candidate region (a white region described later in the present embodiment) that is a region extracted as an edge in the generated edge image. By performing the processing, an expanded image is generated.

  The road tack detection unit 12 detects a road tack that defines a lane on the road based on the aspect ratio and area relating to the white region in the generated expanded image (road tack detection means). The distortion correction unit 13 corrects distortion aberration caused by the camera 30. The lane recognition unit 14 recognizes a lane on the road based on the detected roadway. The imaging range changing unit 15 adjusts the imaging range by the camera 30 according to the lighting state of the headlight 31 (adjustment unit).

  The image memory 20 has a function of inputting a captured image captured by the camera 30 (image input means), and includes a frame memory configured by a RAM. The image memory 20 has a capacity capable of storing a captured image (digital video) for one frame output from the camera 30. The captured image captured by the camera 30 is stored in the image memory 20 for each frame.

  The camera 30 is an imaging unit mainly composed of an optical system and an image sensor (for example, a CCD or CMOS sensor), and creates a digital image having a predetermined luminance gradation (for example, a gray scale of 256 gradations). . The camera 30 outputs the digitized captured image for one frame to the image memory 20. As shown in FIG. 2, the camera 30 is attached, for example, in the vicinity of a room mirror of the vehicle C, and images a landscape including a road (traveling road surface) ahead of the vehicle C.

  FIG. 3 is a flowchart showing a procedure of lane recognition processing according to the present embodiment. The processing shown in this flowchart is called at a predetermined cycle and executed by the control unit 10.

  First, in step 1 (S1), the candidate image generation unit 11 reads a captured image corresponding to one frame from the image memory 20, and performs differential processing on the captured image as a processing target to generate a differential image. Specifically, the candidate image generation unit 11 creates a differential image by performing spatial differential processing on the captured image using a differential operator such as a sobel filter, for example. This differential image is an image indicating the edge strength related to each pixel constituting the captured image.

  In step 2 (S2), the candidate image generation unit 11 performs binarization processing on the generated differential image as a processing target. Specifically, the candidate image generation unit 11 generates an edge image by performing binarization using a predetermined threshold for the edge intensity for each pixel, that is, for each pixel. For example, the candidate image generation unit 11 converts a pixel having an edge strength equal to or greater than a threshold value to white, and converts a pixel having an edge strength smaller than the threshold value to black.

  FIG. 4 is an explanatory diagram schematically illustrating an example of an edge image. In the example shown in the figure, for convenience of explanation, a state in which white and black are reversed, that is, a portion corresponding to binarized black is drawn in white and corresponds to binarized white. The portion is drawn in black (the same applies to the drawings shown below). As shown in the figure, in the edge image, for example, a region (candidate region) extracted as an edge corresponding to the contours of the botsdots S1, the reflector S2, the white line S3, the road surface dirt S4, the preceding vehicle S5, etc. is white, Other areas are represented as black.

  In step 3 (S3), the candidate image generation unit 11 performs an expansion process using a morphological operation or the like, thereby creating an expanded image in which a white region (candidate region) included in the edge image is expanded in all directions. . Specifically, this expansion processing is performed by setting each pixel corresponding to white in the edge image as a processing target, a pixel group existing around the pixel (for example, eight pixels existing around the processing target pixel, or , The binary attribute of 8 pixels existing around the pixel to be processed and 16 pixels existing around the 8 pixels is set to white. The number of pixels to be expanded corresponds to the noise element including the road surface dirt S4 to the extent that the white area corresponding to the road fence is not connected to the white area corresponding to an element other than the road fence (non-road fence). It is desirable that the white areas are connected to each other, and the optimum value is set through experiments and simulations. For example, the botsdots S1 or the like is represented as a ring-shaped region in the edge image. By expanding this region in all directions, the ring-shaped inner peripheral portion is filled with white in the expanded image, and toward the outer peripheral side. Represented as an enlarged circular area.

  In step 4 (S4), the road fence detection unit 12 detects a group of continuous white areas using the expanded image as a processing target. Specifically, the road tack detection unit 12 scans the dilated image to detect each white area that is a lump. The road tack detector 12 calculates the area, aspect ratio, and coordinates of the representative point for each detected white region. For example, in the white region So shown in FIG. 5, the area is 11 pixels, and the aspect ratio is calculated by (yb−yt) / (xr−xl). Here, yt is the y coordinate of the uppermost point of the white region So, yb is the y coordinate of the lowermost point, xl is the x coordinate of the leftmost point, and xr is the x coordinate of the rightmost point. In addition, the representative point can be set to an arbitrary point (for example, the center of gravity) on the white area, but it is necessary to determine the same point for all the white areas So.

  In step 5 (S5), the road fence detection unit 12 extracts, as the road fence area, a white area So whose area and aspect ratio are within a predetermined range, with each detected white area So as a processing target (filtering). processing). Specifically, the road fence detection unit 12 extracts a white area So whose area S and aspect ratio A satisfy the following formula as a road fence area.

(Formula 1)
SL ≦ S ≦ SH
AL ≦ A ≦ AH
Here, SL is a lower limit value with respect to the area S of the white region So to be extracted, and SH is an upper limit value thereof. AL is a lower limit value with respect to the aspect ratio A of the white area So to be extracted, and AH is an upper limit value thereof. These parameters SL, SH, AL, and AH are set to optimum values through experiments and simulations, taking into account the characteristics on the picked-up image of the road surface projected by the camera 30.

  Further, the area S of the white area So corresponding to the road fence located near the host vehicle tends to be large, and the area S of the white area So corresponding to the road fence located far away tends to be small. Therefore, when the y coordinate of the white area So is small, that is, when the white area So is positioned at the upper part of the image, the upper and lower limit values SL and SH regarding the area S may be adjusted to be small. When the y coordinate of the white area So is large, that is, when the white area So is located at the lower part of the image, the upper and lower limit values SL and SH regarding the area S may be adjusted to be large.

  In step 6 (S6), the distortion correction unit 13 geometrically corrects the distortion included in the representative point coordinates for the road area extracted in step 5 using a technique such as affine transformation. Thereby, the representative point coordinates in which the influence of distortion or the like caused by the optical system of the camera 30 is suppressed are newly obtained for the roadside area.

  In step 7 (S7), the lane recognition unit 14 recognizes the road shape, that is, the lane, based on the representative point coordinates regarding each roadway area. Specifically, the lane recognition unit 14 recognizes the lane by performing linear regression on the representative point coordinates regarding each roadway area using a least square method or the like.

  FIG. 6 is an explanatory diagram schematically showing a captured image in a state where the headlight 31 of the vehicle is turned on. Also, at night, the illuminance difference in the captured image increases due to the lighting of the headlight of the host vehicle. Although it is possible to correct a certain illuminance difference by shading correction or the like, there is a limit to the correction due to insufficient dynamic range of the camera. As shown in the figure, the range irradiated by the headlight 31 of the host vehicle with respect to the normal imaging range (outermost rectangular frame) corresponds to the frame S5 in the figure. For this reason, the road fence S6 and the surrounding road surface existing in the frame S5 cause a phenomenon that the subject is completely whitened, that is, so-called whiteout, and edge detection becomes difficult. On the other hand, the road fence S7 and the surrounding road surface existing outside the irradiation range (frame S5) of the headlight 31 are difficult to detect edges because of insufficient illuminance.

  Therefore, the imaging range changing unit 15 acquires the state of the headlight 31 through the in-vehicle LAN, and changes the imaging range of the camera 30 to the frame CA when in the lighting state. The frame CA is set so that the imaging range of the camera 30 substantially includes the irradiation range of the headlight 31 (frame S5). At that time, the imaging range changing unit 15 also changes the photometric area used for exposure correction and the like in the same manner as the imaging range.

  Thus, in the present embodiment, the candidate image generation unit 11 generates an edge image by performing differentiation processing and binarization processing on the captured image captured by the camera 30 as a processing target. In addition, the candidate image generation unit 11 performs an expansion process to expand this region in all directions with respect to the white region that is the candidate region in the generated edge image, and generates an expanded image. The road tack detection unit 12 detects the road tack based on the aspect ratio and the area related to the white region in the generated expanded image. The lane recognition unit 14 recognizes a lane on the road based on the detected roadway.

  According to such a configuration, the road fence and other elements can be clearly separated in terms of area and shape (aspect ratio) by expansion processing in all directions. As a result, it is possible to effectively detect a roadway while removing noise.

  In the present embodiment, the candidate image generation unit 11 connects the white areas corresponding to the noise elements to each other to the extent that the white area corresponding to the road fence does not connect to the white area corresponding to the non-road fence. Expansion processing is performed to such an extent that According to such a configuration, it is possible to connect noise regions such as road dirt and to change the area / aspect ratio thereof to those greatly different from that of the road surface. Thereby, it becomes possible to detect a roadway effectively.

  Further, in the present embodiment, the road tack detection unit 12 variably sets the area thresholds SL and SH for detecting the road tack according to the position of the white region in the expanded image. According to such a configuration, noise removal can be promoted by setting the threshold values SL and SH with respect to the area large in the own vehicle proximity region (on the lower side of the image) where road surface dirt is conspicuous. In addition, in a distant area where the road surface dirt is not conspicuous (on the upper side of the image), by setting a small threshold for the area, it is possible to promote the extraction of road fences that appear small.

  Moreover, in this embodiment, the imaging range change part 15 adjusts the imaging range by the camera 30 according to the lighting state of the headlight 31 of a vehicle. According to such a configuration, by limiting the imaging range to the light irradiation range of the host vehicle, it is possible to suppress the difference in illuminance in the captured image, so it is possible to improve the detection accuracy of the roadway.

  In addition, it has been clarified by experiments by the inventors that the difference in aspect ratio between the botsdots and cat's eyes as a roadway is almost the same. Therefore, even if the difference between the lower limit value AL and the upper limit value AH of the aspect ratio A is not set corresponding to the botsdots and the cat's eye, both can be detected as road traps using common parameters. Is possible.

(Second Embodiment)
FIG. 7 is a configuration diagram schematically showing a lane recognition device 1 according to the second embodiment of the present invention. The lane recognition device 1 according to the present embodiment is different from the first embodiment in that a mask image is generated. In the following, the description of the same configuration as that of the first embodiment will be omitted by citing the reference numerals, and the following description will be focused on the differences.

  As in the first embodiment, the lane recognition device 1 according to the present embodiment is mainly configured by a control unit 10 and an image memory 20. The control unit 10 is connected to the camera 30 via the image memory 20. Has been.

  The control unit 10 includes a mask image generation unit 16, a candidate image generation unit 11, a roadway detection unit 12, a distortion correction unit 13, and a lane recognition unit 14 when this is functionally captured. Yes. Here, the candidate image generation unit 11, the roadway detection unit 12, the distortion correction unit 13, and the lane recognition unit 14 are the same as those in the first embodiment. Moreover, in this embodiment, although the imaging range change part 15 is not provided, the structure which has this may be sufficient.

  The mask image generation unit (first processing means) 16 performs binarization processing on the captured image input to the image memory 20 as a processing target, and performs dilation processing in all directions for the image after binarization processing. Thus, a mask image is generated. In this embodiment, the candidate image generation unit 11 performs mask processing on the dilated image using a mask image.

  FIG. 8 is a flowchart showing a procedure of lane recognition processing according to the present embodiment. The processing shown in this flowchart is called at a predetermined cycle and executed by the control unit 10.

  First, in step 10 (S10), the mask image generation unit 16 reads a captured image corresponding to one frame from the image memory 20, and performs a binarization process on the captured image as a processing target to generate a binary image. . Specifically, the mask image generation unit 16 generates a binary image by performing binarization using a predetermined threshold for luminance for each pixel, that is, for each pixel. For example, the mask image generation unit 16 converts a pixel having a luminance equal to or higher than the threshold to white, and converts a pixel having a luminance smaller than the threshold to black. As a result, in the binary image, for example, road fences, white lines, sky, and preceding vehicles are displayed in white, and other areas including road surface dirt are displayed in black. Here, the threshold value is set as a value that can separate the road fence to be detected from the road surface dirt that is a noise element, and the binary image is displayed on the image. The binarization process is performed so that the pixel corresponding to the road fence has the same binary attribute (white in this embodiment) as the pixel corresponding to the road fence in the expanded image generated by the candidate image generation unit 11. .

  In step 11 (S11), the mask image generation unit 16 performs an expansion process using a morphological operation or the like on the binary image generated in step 10 as a processing target. Thereby, the mask image generation unit 16 creates a mask image in which the white region included in the binary image is expanded in all directions. The expansion processing in this step is basically the same as the expansion processing method shown in the first embodiment, but the size of the roadway in the mask image is a contracted image after the contraction processing in step 15 described later. The expansion rate is set to be larger than the size of the roadway.

  In step 12 (S12), the candidate image generation unit 11 reads a captured image corresponding to one frame from the image memory 20, and performs differential processing on the captured image as a processing target to generate a differential image. In step 13 (S13), the candidate image generation unit 11 generates an edge image by performing binarization processing on the generated differential image as a processing target. In step 14 (S14), the candidate image generation unit 11 performs an expansion process using a morphological operation or the like, thereby creating an expanded image in which the white region included in the edge image is expanded in all directions. Each process in steps 12 to 14 corresponds to each process in steps 1 to 3 in the first embodiment.

  In step 15 (S15), the candidate image generation unit 11 performs a contraction process using a morphological operation or the like with the expanded image as a processing target. Specifically, the candidate image generation unit 11 contracts each of the white areas expanded in Step 14 with the expanded areas inward, so that each white area is subjected to the expansion process. The size is adjusted to be approximately the same as the size of the image (creation of contraction image).

  In step 16 (S16), the mask image generation unit 16 performs mask processing on the contracted image generated in step 15 using the mask image generated in step 11. Specifically, for each of the white pixels included in the contracted image, the mask image generation unit 16 changes the pixel to black if the pixel on the mask image corresponding to the pixel is black. . On the other hand, for each black pixel included in the contracted image, even if the pixel on the mask image corresponding to the pixel is white, the pixel is not changed to white. That is, for each of the white pixels included in the contracted image, only the portion where the pixel on the mask image corresponding to the pixel is white is white, and the other pixels are black. In other words, when the pixel included in the contracted image is white and the pixel on the mask image corresponding to the pixel is white, the pixel is not changed, and the pixel included in the contracted image If the pixel on the mask image corresponding to the pixel is different in color or if both are black, the pixel is black. As a result, an image in which noise such as road surface dirt included in the contracted image is suppressed can be obtained.

  In each process from step 17 (S17) to step 20 (S20), each process of area detection, filtering, distortion correction, and road shape estimation is performed with the image obtained in step 16 as a processing target. These processes correspond to the processes from Step 4 to Step 7 in the first embodiment.

  As described above, in this embodiment, the mask image generation unit 16 performs binarization processing on the captured image input to the image memory 20 as a processing target, and performs dilation processing in all directions for the image after binarization processing. By doing so, a mask image is generated. Then, the mask image generation unit 16 performs mask processing on the expanded image using the mask image. According to such a configuration, dirt on the road surface can be removed, so that it is possible to improve the detection accuracy of the roadway.

  In the present embodiment, the candidate image generation unit 11 further performs a contraction process for contracting the size of the expanded candidate area on the expanded image after performing the expansion process. According to this configuration, it is possible to remove fine noise around the roadside and shape the shape. Thereby, the detection accuracy of a roadway can be improved. In addition, about this method, it is also possible to apply this to the method of 1st Embodiment independently.

(Third embodiment)
FIG. 9 is a configuration diagram schematically showing a lane recognition device 1 according to the third embodiment of the present invention. The lane recognition device 1 according to the present embodiment is different from the second embodiment in that the lane recognition method can be switched. In the following, the description of the same configuration as that of the second embodiment will be omitted with reference to the reference numerals, and the description will be made focusing on the differences.

  As in the first embodiment, the lane recognition device 1 according to the present embodiment is mainly configured by a control unit 10 and an image memory 20. The control unit 10 is connected to the camera 30 via the image memory 20. Has been.

  When the control unit 10 grasps this functionally, the mask image generation unit 16, the candidate image generation unit 11, the roadway detection unit 12, the distortion correction unit 13, the lane recognition unit 14, and the detection target switching Part 17 and white line detection part 18. Here, the mask image generation unit 16, the candidate image generation unit 11, the roadway detection unit 12, the distortion correction unit 13, and the lane recognition unit 14 are the same as those in the second embodiment.

  The detection target switching unit 17 has a white line detection mode and a roadside detection mode as lane recognition operation modes, and performs lane recognition while alternatively selecting both modes. Here, the white line detection mode is a mode for recognizing a lane by detecting a lane division line (line) that demarcates a lane on the road, such as a white line, by the white line detection unit 18 described later. On the other hand, as shown in the second embodiment, the road ridge detection mode is performed on the road by the mask image generation unit 16, the candidate image generation unit 11, the road lane detection unit 12, the distortion correction unit 13, and the lane recognition unit 14. In this mode, a lane is recognized by detecting a roadway.

  The white line detection unit (line detection unit) 18 reads a captured image corresponding to one frame from the image memory 20 from the image memory 20 and defines a lane on the road based on the captured image (for example, a white line). And thereby recognize the lane. A well-known technique can be adopted as a technique for white line recognition. For example, as described in JP-A-2002-175535, a plurality of white line candidate points are extracted from a luminance change of a captured image, and a plurality of white lines extracted are extracted. A line defining a lane on the road can be recognized based on the candidate points (refer to Japanese Patent Application Laid-Open No. 2002-175535 for details).

  In the lane recognition device 1 having such a configuration, the detection target switching unit (switching unit) 17 switches the operation mode as follows. Specifically, the lane recognition device 1 switches to the other operation mode when the detection failure is formed for a predetermined time (for example, 3 seconds) or more in a state where any one of the operation modes is selected. When the lane recognition unit 14 performs linear regression on the representative point coordinates of the white area, the error between the regression line and the representative point coordinates exceeds a predetermined threshold. And so on. On the other hand, the detection failure determination in the white line detection mode is based on the condition that the degree of matching between the regression line and the edge point with respect to the white line is below a predetermined threshold.

  Thus, in this embodiment, since the existing white line detection technology and roadside detection can coexist, a lane can be recognized even in an environment where whitelines and roadsides are mixed.

DESCRIPTION OF SYMBOLS 1 Lane recognition apparatus 10 Control unit 11 Candidate image generation part 12 Roadway detection part 13 Distortion correction part 14 Lane recognition part 15 Imaging range change part 16 Mask image generation part 17 Detection target switching part 18 White line detection part 20 Image memory 30 Camera 31 Headlight

Claims (7)

Imaging means for imaging the road surface;
First processing means for generating an edge image obtained by extracting a luminance edge in a captured image captured by the imaging means;
In the edge image generated by the first processing means, candidate image generation for generating an expanded image by performing expansion processing for expanding the region in all directions with respect to a candidate region that is an area extracted as an edge Means,
In the expanded image generated by the candidate image generation means, road ridge detection means for detecting a road lane that defines a lane on the road based on the aspect ratio and area relating to the candidate area;
Based on the detected road stud by the road stud detection means have a recognition means for recognizing the lane on the road,
The lane recognition device according to claim 1, wherein the candidate image generation unit performs an expansion process so that a candidate area corresponding to a roadside is not connected to a candidate area corresponding to a non-roadside . The candidate image generating means, to the extent that the candidate regions each other corresponding to the noise components including the soil on the road surface are connected to each other, have been lane recognition device according to claim 1, which comprises carrying out the expansion process.   The lane recognition device according to claim 1 or 2, wherein the roadside detection means variably sets an area threshold for detecting a roadside according to the position of the candidate area in the expanded image. . The candidate image generation means generates a contracted image by contracting the size of the expanded candidate area with respect to the expanded image after performing the expansion process,
A binarization process is performed using a captured image captured by the imaging unit as a processing target and a value capable of separating a road fence and a noise element as a threshold value, and the image after the binarization process is processed in all directions. A second processing means for generating a mask image by further performing expansion processing;
The second processing unit uses the mask image with respect to the contracted image generated by the candidate image generating unit , and a pixel included in the contracted image is white and is positioned with respect to the pixel. If the pixel on the mask image corresponding to is white, the pixel is not changed, and the color of the pixel on the mask image corresponding to the pixel and the pixel included in the contracted image is different from both. The lane recognition device according to any one of claims 1 to 3, wherein in the case of black, mask processing is performed to make the pixel black . The lane recognition device according to any one of claims 1 to 4 , further comprising an adjusting unit that adjusts an imaging range of the imaging unit in accordance with a lighting state of a vehicle illumination. Line detecting means for recognizing a lane on the road by detecting a line defining a lane on the road based on a captured image captured by the imaging means;
The operation mode further includes a mode for recognizing a lane based on the detection of the line, a mode for recognizing a lane based on the detection of the roadside, and a switching unit that selectively switches the operation mode. The lane recognition device according to any one of claims 1 to 5 , wherein A first step of generating an edge image obtained by extracting a luminance edge in the image with the captured image captured by the imaging means as a processing target;
A second step of expanding the region in all directions with respect to a candidate region that is a region extracted as an edge in the generated edge image;
A third step of detecting road ridges defining lanes on the road based on the aspect ratio and area, with the expanded candidate area as a processing target;
Based on the detected road studs, it has a fourth step recognizes the lane on the road,
The second step is a lane recognition method , wherein the candidate area is expanded to such an extent that the candidate area corresponding to the road fence is not connected to the candidate area corresponding to the non-road fence .

Images