KR101365557B1 - Apparatus and method for detecting traffic lane - Google Patents

Abstract

According to the present invention, a divided image having different resolutions is set for a captured image provided from a camera, and the lane for the corresponding frame is detected using the lane information extracted from the divided image, thereby minimizing the probability of lane detection error. The present invention relates to a lane detecting apparatus and method for providing safe operation of a vehicle.
The lane detection apparatus and method according to the present invention are installed in a vehicle and output a high resolution captured image, and resizing the high resolution captured image provided from the camera to at least one or more different resolutions, Image conversion means for generating a region-specific segmented image obtained by extracting only an image of a corresponding region from each image so as to correspond to a plurality of regions set according to distances between the image and the resizing image; And lane lane determination means for generating lanes for the corresponding frame by combining lane information extracted through the het filter, and detecting lanes through Hough transformation based on the generated lanes.

Description

Lane detection apparatus and its method {Apparatus and method for detecting traffic lane}

According to the present invention, a divided image having different resolutions is set for a captured image provided from a camera, and the lane for the corresponding frame is detected using the lane information extracted from the divided image, thereby minimizing the probability of lane detection error. The present invention relates to a lane detecting apparatus and method for providing safe operation of a vehicle.

Recently, with the development of information and communication technology, the lane departure warning system that provides a warning sound according to the lane departure of the vehicle or the driving, stop, rotation, acceleration, deceleration, etc. The automatic control system and the like in which the driving operation of the vehicle has been developed and the practical use is rapidly progressing.

Here, in a system providing convenience of a vehicle, such as the lane departure warning system or the vehicle automatic control system, a technique for detecting a lane from an image captured by a camera may be one of the most basic and core technologies.

However, in the above system, since a large amount of data processing must be performed in addition to the lane detection, the image provided from the camera is converted to a low resolution to detect the lane based on this.

However, the captured image provided by the camera installed in the vehicle contains elements that obstruct the lane detection such as shadows of road obstacles, other vehicles, random marks on the road, and the surrounding environment of the road, and thus, lane detection is relatively easy in the near area. However, there is a problem in the detection of lanes in the remote area. FIG. 1 is a diagram illustrating a situation in which another lane C is not detected as a noise and is recognized as a lane so that a lane is incorrectly detected.

Accordingly, the present invention was created in view of the above circumstances, and sets a divided image having different resolutions from a captured image provided from a camera, and uses the lane information for each region extracted from the divided image to lanes for the corresponding frame. It is a technical object of the present invention to provide a lane detection apparatus and method for minimizing the lane detection error probability while minimizing the data processing speed.

The lane detection apparatus according to the first aspect of the present invention for achieving the above object is a camera installed in the vehicle to output a high-resolution photographed image, and resizing the high-resolution photographed image provided from the camera to at least one or more different resolutions In addition, the image conversion means for generating a region-specific segmented image extracted only the image of the corresponding region from each image so as to correspond to a plurality of areas set according to the distance between the high-resolution photographed image and the resizing image, provided from the image conversion means And a lane determination means for generating lanes for the corresponding frames by combining lane information extracted through the top-hat filter from the divided images for each region, and detecting lanes through Hough transformation based on the generated lanes. Wherein, the image conversion means for performing a low-resolution first segmented image for the near area And output a second high resolution image of the high resolution to the distant region.

In addition, the lane detection method according to a second aspect of the present invention for achieving the above object is a first step of resizing a high-resolution photographed image provided from a camera installed in a vehicle to at least one or more different resolutions, and provided from the camera A second step of setting a plurality of divided areas according to distances based on preset reference distance information in the captured image; a third step of extracting corresponding divided areas from an image having a resolution corresponding to each divided area; A fourth step of extracting divided lane information from each divided region extracted in the step; a fifth step of generating lane information for the corresponding frame by connecting divided lane information of each divided region extracted in the fourth step; And a sixth step of deciding a lane by hough transforming the lane information generated in step 5, wherein the first step is performed by a camera. Resizing the high-resolution photographed image to low resolution, and the second step sets the near and far areas based on the reference distance information in the captured image, and the third step extracts the near area from the resized image and the camera. Characterized in that it is configured to extract the far region from the high-resolution photographed image provided from.

According to the present invention, a lane detection error is detected by extracting lane information from a low resolution image in a short range region and extracting lane information from a high resolution image in a far field and detecting a lane for a corresponding frame. By minimizing the probability, it is possible to provide a safer vehicle driving service.

1 is a diagram showing a lane misdetection example of a conventional lane detection apparatus.
2 is a diagram showing a schematic configuration of a lane detection apparatus according to a first embodiment of the present invention;
Figure 3 is a block diagram showing the functional separation of the internal configuration of the image conversion means 200 shown in FIG.
FIG. 4 is a diagram for describing an image splitting method of the image converting means 200 shown in FIG. 3.
Figure 5 is a block diagram showing the functional separation of the internal configuration of the lane determination means 300 shown in FIG.
FIG. 6 is a flowchart for explaining a lane detection method of the lane detection device shown in FIG. 2; FIG.
7 is a view for explaining a lane detection method according to a second embodiment of the present invention;

Hereinafter, embodiments according to the present invention will be described with reference to the drawings. However, the embodiments described below represent one preferred embodiment of the present invention, and are not intended to limit the scope of the present invention. The present invention can be variously modified and carried out without departing from the technical gist thereof.

2 is a view showing a schematic configuration of a lane detection apparatus according to a first embodiment of the present invention.

As shown in FIG. 2, the apparatus for detecting lanes according to the present invention includes at least one different resolution of a camera 100 installed in a vehicle and outputting a high resolution captured image, and a high resolution captured image provided from the camera 100. Image conversion means 200 for resizing and segmenting output by area, and detecting respective lanes from the divided images for each area provided from the image conversion means 200, and combining the lanes with respect to the corresponding frame. It is configured to include a lane determining means 300 to determine the.

The camera 100 photographs the front surface of the vehicle and provides a high resolution photographed image to the image converting means 200. For example, the camera 100 outputs a high definition (HD) captured image having a resolution of 1,280 × 720.

The image converting means 200 resizes at least one or more different resolutions to a high resolution photographed image provided from the camera 100, and sets a plurality of regions, for example, a short range region and a far region according to a distance, in the photographed image. A low resolution resized segmented image is extracted for the near region and a high resolution segmented image is extracted for the far region.

The lane determining means 300 extracts lane information from each divided image provided from the image converting means 200, and combines the extracted region divided lanes to determine the lane in the corresponding frame.

FIG. 3 is a block diagram showing functional separation of the internal configuration of the image conversion means 200 shown in FIG.

As shown in FIG. 3, the image converting means 200 includes a resizing processor 210, an area setting unit 220, and an image divider 230.

The resizing processor 210 resizes the high resolution captured image provided from the camera 100 to at least one image having different resolutions. For example, the resizing processor 210 resizes and outputs a high resolution photographed image in a low resolution, for example, QVGA (Quarter Video Graphics Array) method having a resolution of 320 × 240.

The area setting unit 220 sets a plurality of divided areas according to distances based on preset reference distance information in the captured image provided from the camera 100. That is, the region setting unit 220 first extracts the vanishing point Y from the captured image as shown in FIG. 4. Here, the vanishing point (Y) may be a horizon in the captured image. Subsequently, the area setting unit 220 sets the ROI to include the lane L in the road area below the vanishing point Y. In this case, the region of interest (ROI) may be a trapezoidal area including a lane (L). Thereafter, the area setting unit 220 sets the near area A and the far area B through camera modeling based on the reference distance information. For example, when the reference distance R information is set to "30m", an area within 30m from the vehicle may be set as a short range area and a subsequent area as a far area. Here, since the distance calculation method through the camera modeling is a general method calculated by the lens characteristics, the installation position and the installation angle of the camera, detailed description thereof will be omitted.

The image segmentation processor 230 divides images of different resolutions provided by the camera 110 and the resizing processor 210 so as to correspond to region setting information provided by the region setting unit 220. That is, the image segmentation processor 230 extracts the low resolution resizing image provided from the resizing processor 210 for the near region, and extracts the high resolution image provided from the camera 100 for the far region, respectively, to determine the lanes. Provided by means 300.

FIG. 5 is a block diagram illustrating a functional separation of the internal configuration of the lane determination means 300 shown in FIG. 2.

As shown in FIG. 5, the lane determining unit 300 includes a lane information extracting unit 310, a lane combining unit 320, a linear component detecting unit 330, and a lane expanding unit 340.

As illustrated in FIG. 5, the lane information extracting unit 310 may include first to N th lane information extracting modules for extracting lane information for split images having different resolutions provided from the image converting means 200. Not shown). For example, the lane information extracting unit 310 is composed of first and second lane information extracting modules (not shown) so that the first lane information extracting module is configured for a near-field segmented image provided from the image converting means 200. The first lane information is extracted, and the second lane information extraction module 312 extracts the second divided lane information from the far divided image. Here, the lane information extracting unit 310 extracts lane information in the form of a line using a top hat filter. That is, in the case of lanes, in general, about 15cm in Korea and about 10cm in the United States, the lane information is extracted by detecting a line having a certain thickness through a top hat filter. In addition, since the lane is usually white or yellow in color than the road, the lane can be easily extracted by detecting a portion having a higher gray level.

The lane combiner 320 combines divided lane information for each region provided from the lane extractor 310 to generate one lane for a corresponding frame. That is, one lane is generated by sequentially connecting the first divided lane information extracted from the near division image and the second divided lane information extracted from the far divided region.

The linear component extractor 330 may be obtained by extracting a linear equation using a two-dimensional Hough transform based on the lane information provided from the lane coupler 320. The concept of Hough transform is to accumulate the data by creating a space with the parameters of a function as a coordinate system to find a function that approximates some data, and to find the desired parameter by obtaining the coordinates of the largest value. The detailed description is omitted because it is a formula.

In addition, the lane extension 340 determines a lane among a plurality of linear equations extracted through the Hough transform in the linear component extractor 330. That is, the lane extension 340 compares one or more straight lines by the extracted linear equation with the lane determined in the previous frame and determines that the difference in the slope is the smallest as the lane of the corresponding frame.

At this time, if the lane is not determined because the lane is not detected in the previous frame or the like, the lane is not determined. If the other lane is determined in the current frame, the distance from the lane is a certain distance. It can be done so that the straight line in the area of about 2.5m) is determined as the lane. In addition, when lane determination is not performed for both lanes in the previous frame, the HFT transformation may be performed to determine the line having the largest voting number as the lane. In addition, the lane information is not extracted in the current frame, but when the lane is determined in the previous frame, the lane information is mainly a dotted line lane on the general road. Can be done.

Next, a lane detection method of the lane detection apparatus having the above-described configuration will be described with reference to the flowchart shown in FIG.

First, a user sets reference distance information for setting a near area and a far area in a captured image. In this case, the reference distance information may be set at the time of manufacture of the lane detection apparatus, or the user may arbitrarily change and set the reference distance information set at the time of manufacture.

In the above state, the camera 100 installed in the vehicle provides a high-resolution photographed image of the front of the vehicle to the image converting means 200 (ST1).

The image converting means 200 resizes the high resolution photographed image provided from the camera 100 to at least one resolution different from each other (ST2). That is, the image converting means 200 resizes the high resolution captured image provided from the camera 100 to a low resolution image.

In addition, the image converting means 200 sets a plurality of divided regions according to distances based on predetermined reference distance information in the captured image provided from the camera 100 (ST3). That is, the image converting means 200 sets a trapezoid-shaped ROI in the road area based on a horizon, and selects a first division area corresponding to the near area and a second division area corresponding to the far area in the ROI. Set it.

Subsequently, the image converting means 200 extracts the divided region from the image having the resolution corresponding to each divided region (ST4). That is, the first divided image having a low resolution is extracted for the first divided region, which is the near region, and the second divided image having high resolution is extracted for the second divided region, which is the far region, and provided to the lane determining means 300.

Lane determination means 300 extracts the respective divided lane information from the plurality of divided images using the top-hat filter (ST5). That is, the first divided lane information of the near area and the second divided lane information of the far area are extracted from the first and second divided images.

The lane determination unit 300 connects the divided lane information extracted from each divided image to generate one lane information of the current frame (ST6).

Thereafter, the lane determination means 300 calculates a plurality of linear equations through Hough transform based on one lane information generated for the current frame, and determines one of the calculated linear equations as a lane (ST7). In this case, the lane determination means 300 may be set to a straight line component having the smallest slope difference based on the lane information determined in the previous frame.

In addition, the lane determination unit 300 determines the straight line information having the largest voting number in the Hough transform as the lane when there are no lanes determined in the previous frame.

In addition, when there is no lane determined in the previous frame, the lane determination unit 300 determines a straight line component having a predetermined distance from the other lane determined in the previous frame as the lane.

That is, according to the embodiment, a segmented image of a distant region is extracted from a high resolution photographed image provided from a camera, and a segmented image of a near region is extracted from a low resized image, and a lane is extracted from each of the extracted segmented images. The information is extracted, and the lane is detected using the extracted lane information.

Meanwhile, in the above embodiment, the photographed image is divided into two regions, a near region and a far region. However, as shown in FIG. 7, the photographed image is divided into three regions, a near region, a far region, and a middle region. To detect a lane. In FIG. 7, (a) shows a partition area setting screen, and (b) shows lane information extracted from each partition area.

For example, the lane detecting apparatus converts a high resolution captured image into a low resolution and a medium resolution resizing image in a state where reference distance information for setting a near area, a far area, and a medium distance area in a captured image is set in advance, and then, A first segmented image is extracted from a low resolution resizing image, and a second segmented image is extracted from a medium resolution resizing image of a VGA (video graphics array) method having a resolution of 640 × 480 for a medium distance region. For example, the third segmented image may be extracted from a high resolution photographed image provided from the camera.

Accordingly, the present invention detects a lane more accurately than a method of detecting a lane using a low resolution shot image while minimizing a decrease in processing speed compared to a method of detecting a lane using a high resolution shot image for accurate lane detection. You can do it.

100: camera, 200: image conversion means,
300: lane determination means,
210: resizing processing unit, 220: area setting unit,
230: image segmentation unit, 310: lane information extraction unit,
320: lane coupling unit, 330: linear component detection unit,
340: Lane Confirmation.

Claims (13)

A camera installed in the vehicle and outputting a high resolution photographed image;
Resizing the high resolution captured image provided from the camera to at least one different resolution, and extracting only the image of the corresponding region from each image so as to correspond to a plurality of regions set according to the distance between the high resolution captured image and the resizing image Image conversion means for generating a separate image for each;
By combining the lane information extracted through the top-hat filter from the divided image for each region provided from the image conversion means to generate a lane for the frame, and detecting the lane through the Hough transform based on the generated lane Including lane-determining means,
And the image converting means is configured to output a low resolution first segmented image for a short range region and a second segment image of a high resolution for a far region. delete The method of claim 1,
And the first divided image of the image converting means is an image resized in a QVGA method, and the second divided image is an HD image provided from a camera. The method of claim 1,
The image converting means is configured to output a low resolution first segmented image for a short range region, a high resolution second segmented image for a far region, and a third segment image of medium resolution for a medium range region. Lane detection device characterized in that. The method of claim 1,
And the lane determination means determines the straight lane with the smallest inclination difference as the lane of the current frame based on the lane information determined in the previous frame. The method of claim 1,
And the lane determining means determines a straight line having a distance from the lane as a lane when the other frame lane in the current frame is not determined. The method of claim 1,
And the lane determining means determines a straight line having the largest voting number in the Hough transform as the lane when the lane is not determined in the previous frame and the lane is not determined for both lanes in the current frame. Resizing the high resolution captured image provided from the camera installed in the vehicle to at least one or more different resolutions;
A second step of setting a plurality of divided regions according to distances based on preset reference distance information in the captured image provided from the camera;
A third step of extracting a corresponding divided region from an image having a resolution corresponding to each divided region,
A fourth step of extracting divided lane information from each divided area extracted in the third step,
A fifth step of generating lane information on the corresponding frame by connecting the divided lane information of each divided region extracted in the fourth step;
And a sixth step of deciding a lane by Hough transforming the lane information generated in the fifth step.
The first step is to resize the high resolution captured image provided from the camera to a low resolution,
In the second step, the near and far areas are set based on the reference distance information in the captured image.
The third step is a lane detection method characterized in that it is configured to extract the near-area from the resizing image and the far-field from the high-resolution photographed image provided from the camera. delete 9. The method of claim 8,
The first step is to convert the high-resolution photographed image provided from the camera to a low resolution first resizing image and a medium resolution second resizing image,
In the second step, a short range region, a far range region, and a medium range region are set based on the reference distance information in the captured image.
In the third step, the lane detection method is configured to extract a short range region from a first resizing image, a middle distance region from a second resizing image, and extract a far region from a high resolution photographed image provided from the camera. . 9. The method of claim 8,
In the sixth step, the lane detection method of determining a straight lane having the smallest inclination difference as the lane of the current frame based on the lane information determined in the previous frame. 9. The method of claim 8,
The sixth step is a lane detection method, characterized in that if the previous frame lane is not determined and the other lane is determined in the current frame, a straight line having a predetermined distance from the lane is determined as the lane. 9. The method of claim 8,
In the sixth step, when the lane is not determined in the previous frame and the lane is determined in both lanes in the current frame, the lane having the largest voting number in the Hough transform is determined as the lane.

Images