KR20090098167A - Method and system for detecting lane by using distance sensor - Google Patents

Abstract

A road line recognizing method and the system using a distance sensor and a simple road line recognition algorithm are provided to receive obstacle information difficult to obtain by a camera. A method for recognizing a road line comprises the following steps of: receiving distance data and video frames(S202); filtering the video frames using a steerable filter and extracting the filtering result(S204); setting a free space using the distance data(S206); applying a template matching algorithm to the filtering result and searching for line components existing within the free space(S210); setting a region of interest(ROI)(S212); and performing a line-fitting to the line components and extracting road lines(S214).

Description

Lane Recognition Method and System Using Distance Sensors {Method and System for Detecting Lane by Using Distance Sensor}

The present invention relates to a lane recognition method and a system using a distance sensor. More particularly, the present invention relates to a lane recognition method and a system for recognizing lanes by identifying obstacle information that is difficult to grasp only by image information through a camera.

Lane Keeping System (LKS) is a device that assists driving to the center of the lane by processing the image of the camera behind the camera and recognizing the white lane of the road.

The existing lane keeping system performs vehicle recognition using only a camera mounted at the rear of the vehicle.

1 is a diagram illustrating lane recognition when a conventional lane recognition system is used.

If another vehicle or the like exists in front of the driving vehicle as shown in FIG. 1A, when the lane is recognized only by the camera, the edge portion caused by the bus of the side lane as shown in FIG. 1B may be recognized as the lane, and thus, FIG. 1C. You may be mistaken that there is a lane like the red line of.

In this way, when only a camera is used for lane recognition, there is a possibility of recognizing a wrong lane due to ambient noise. In particular, when the lane is blocked by the front vehicle and the interrupting vehicle, it is difficult to perform the lane recognition. In addition, when a vehicle traveling forward is photographed on an image captured by a camera, an edge image generated by the vehicle may be incorrectly recognized as a lane.

In order to solve the above problems, an object of the present invention is to identify lanes by identifying obstacle information that is difficult to grasp only by image information through a camera.

In order to achieve the above object, the present invention provides a lane recognition system using a distance sensor, comprising: a camera for photographing the surrounding vehicle to generate an image frame; A distance sensor for generating data for detecting obstacles around the host vehicle; An image receiver which receives the image frame; A distance data receiver for receiving the distance data; A filter unit which extracts a filter result by using a steerable filter on the image frame; A prespace setting unit for setting a prespace using distance data from the distance sensor; A line component extracting unit which finds a line component existing in the prespace by applying a template matching algorithm to the filter result; A region of interest setting unit for setting a region of interest (ROI); And a lane extracting unit performing line fitting to the line component in the region of interest and extracting a lane.

In addition, the present invention to achieve the above object, the lane recognition method using a distance sensor, (a) receiving the image frame and the distance data photographed around the host vehicle; (b) extracting filter results from the image frame using a steerable filter; (c) setting a prespace using the distance data; (d) applying a template matching algorithm to the filter result to find a line component present in the prespace; (e) establishing a Region Of Interest (ROI); And (f) performing line fitting on the line components in the region of interest and extracting lanes.

According to the present invention, by using the distance sensor to obtain obstacle information that is difficult to know only by the camera information, the lane is recognized, and ROI (Region of Interest, ROI) from which the obstacle is removed using the obstacle information from the image input from the camera By setting it, it is effective to recognize lanes using simple lane recognition algorithms using distance sensor information such as noise (edge information generated by the front vehicle) that obstructs lane recognition due to obstacles, and whether lanes are covered. have.

In addition, since the distance sensor can be used already installed in the vehicle for the adaptive cruise control (ACC), lane recognition using the obstacle information on the road using only one camera without additional cost increases. It is possible.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. First of all, in adding reference numerals to the components of each drawing, it should be noted that the same reference numerals are used as much as possible even if displayed on different drawings. In addition, in describing the present invention, when it is determined that the detailed description of the related well-known configuration or function is obvious to those skilled in the art or may obscure the gist of the present invention, the detailed description thereof will be omitted.

2 is a flowchart illustrating a lane recognition method according to an embodiment of the present invention.

In the lane recognition method using a distance sensor according to an embodiment of the present invention, the method comprises: receiving an image frame and distance data photographing the periphery of a vehicle (S202), extracting a filter result from the image frame using a steerable filter; (S204), a binarization step (S206) of recognizing only a predetermined value or more pixels as a valid filter result, setting a prespace using the distance data (S208), and applying a template matching algorithm to the filter result A step (S210) of finding a line component present in the image, a step of setting a region of interest (ROI) (S212), and performing a line fitting on the line component and extracting a lane (S214). It provides a lane recognition method characterized in that.

In some cases, the binarization step S206 may be omitted.

The present invention provides a method for recognizing a lane using a distance sensor for lane recognition. With only one camera, it is difficult to know information about obstacles on the road. The proposed method uses information from the distance sensor to find out obstacle information for lane recognition. In general, the distance sensor can use the one already installed in the vehicle for the adaptive cruise control (ACC), so it is possible to recognize the lane using the obstacle information on the road using only one camera without any additional cost increase. .

In the lane recognition method according to the exemplary embodiment of the present invention, first, an operation (S202) of receiving an image frame photographing the surrounding vehicle is performed.

On the other hand, a region of interest (ROI) may be set after the step S202. In the present embodiment, the region of interest (ROI) may be set before the step S214 of finding a lane by line fitting.

For the following description, the region of interest will be briefly described herein, and more detailed description will be given in the description of the region of interest setting step S212.

The lanes shown in the video frame may be straight or curved depending on the shape of the road. The region of interest (ROI) required for lane recognition is set to include the shape of the lane while minimizing the range to be searched.

The preferred lane shape should be such that only lanes can be included and other objects can be excluded. The lane is divided into three stages so that the ROI becomes smaller as the lane looks smaller, considering that the lane looks smaller as it moves further away. The lane is divided into a left lane and a right lane. Therefore, six sub ROIs may be set in three stages.

3 is a diagram illustrating that a three-step ROI is set step by step in the region of interest setting step S212.

As shown in FIG. 3, there are six ROIs, three each on the left and right sides. That is, the first-level ROI may be divided into sub-ROI I and IV, the second-ROI may be sub-ROI II and V, and the third-ROI may be sub-ROI III and VI.

Sub-ROIs I and IV close to the vehicle may be fixed, and the remaining sub-ROIs may vary according to lane recognition results. That is, the second-level ROI may vary according to the first-level ROI, and the three-step ROI may vary according to the second-ROI.

After the step S202 is finished, the filter result is extracted using the steerable filter from the image frame (S204).

Since the lanes are inclined in the camera image frame and the slopes of the lanes are known, a steerable filter can be used to efficiently recognize the lanes.

The steerable filter may be defined using a 2D (2D) Gaussian function of [Equation 1]. If only the inner edge of the lane is detected, only the first derivative of the Gaussian function can be used. Equation 2 has a derivative with respect to the x-axis (θ = 0 °), and Equation 3 can obtain a filter having a derivative with respect to the y-axis (θ = 90 °).

Designing a filter for any angle gives Equation 4. That is, it is possible to design a filter that knows the response to θ.

4 is a diagram illustrating a stiffable filter result in a camera image in the filter result extraction step (S204).

For the image of FIG. 4A, the stiffable filter result for θ of 45 degrees is FIG. 4B, and the stiffable filter result for θ of −45 degrees is FIG. 4C.

Meanwhile, the lane slope used when the steerable filter is applied is an angle extracted using an edge distribution function (EDF) for an initial image frame, and a slope of a lane extracted from a previous image frame for a subsequent image frame. Can be used.

The steerable filter used to distinguish the lane marking from the road represents the response of each pixel according to the angle. Therefore, in order to generate a stiffable filter for distinguishing a lane marking from a road, an angle formed between the lane marking and the road must be given as an input. In the system of the present embodiment, the initial frame of the camera image uses a value displayed by using an edge distribution function (EDF), and the subsequent frame uses a fitting result of the lane detected in the previous frame. Since the EDF is calculated only in the initial frame, the amount of computation is reduced, and there is an advantage in that the stirrable filter can give an unknown initial value.

The EDF can be used to initialize the directional parameters of the steerable filter. EDF is a histogram in the edge pixel direction versus angle. Gradients are used by the EDF to calculate the directionality of each pixel. Equation 5 defines a gradient of pixels (x, y) in the camera image. D _x represents the change in intensity with respect to the x axis and D _y represents the change in intensity with respect to the y axis.

To calculate the gradient of Equation 5, Sobel Operator is used.

The edge direction of each pixel was defined in [Equation 6] by using the values of D _x and D _y extracted in [Equation 5].

The EDF can be formed by accumulating pixels with respect to the edge direction.

FIG. 5 is a diagram illustrating a result using a Sobel operator with respect to the input image of FIG. 4A in the filter result extraction step (S204), and FIG. 6 illustrates an EDF for the input image of FIG. 4A in the filter result extraction step (S204). It is a graph shown.

In FIG. 6, the peak value on the left side and the peak value on the right side represent left and right lanes, respectively.

As shown in FIG. 6, the test result is divided into two regions based on 90 degrees, and the peak values of each region are set to the angle of the right lane and the angle of the left lane. The left region of FIG. 6 corresponds to the lane of the sub ROI I and the right region corresponds to the lane of the sub ROI IV.

The lanes of the sub ROI I and IV, which are the 1st region of interest, which is the area close to the car, can be approximated by a straight line, and the angle of the extracted peak value represents the lane direction of each sub ROI, so the angle corresponding to the extracted peak value is steered. Can be used as an initial parameter of the flexible filter.

After the step S204 is completed, the binarization step S206 may be performed in which only pixels having a predetermined value or more are recognized as valid filter results.

That is, since the binarization step S206 is performed, pixels having a predetermined value or less in the filter result may be recognized as an invalid result and excluded.

After the binarization step S206, a step of setting a prespace using the distance data obtained from the distance sensor may be performed (S208). Scanning laser radar may be used as the distance sensor.

Prespace means an area where obstacles and the like are removed from the filter result. Therefore, by using a distance sensor in the filter result, it is possible to disregard the areas with unnecessary obstacles and to extract only the areas within the free space so that the lanes without obstacles can be properly recognized. Can be achieved.

In order to set the prespace using the data input from the distance sensor in addition to the input image frame coming into the camera, the relationship between the coordinates of the distance sensor and the input image coordinate must be known. If the pixel coordinate of the input image is (x _i , y _i ) and the coordinate of the distance sensor is (Xw, Yw, Zw), the relationship between the two coordinates can be expressed as [Equation 7] and [Equation 8]. . Where H is the homography that relates the two coordinates.

Since the part of the vehicle in the image starts from the road surface, that is, the point where Yw becomes zero, it is necessary to know the coordinates of the image in which Yw is zero. So we need to know the coordinates in the image where Yw is zero. H of [Equation 7] for obtaining image coordinates where Yw becomes 0 is equal to [Equation 9].

는 카메라의 틸트각(Tilt Angle)이다.Where Hc is the height of the camera, θ is the yaw angle of the camera,

Is the tilt angle of the camera.

FIG. 7 is a diagram showing coordinates of data obtained from a distance sensor such as a scanning laser radar in a prespace setting step S208, and FIG. 8 is a diagram of data coordinates converted from an image sensor into image coordinates in a prespace setting step S208. It is a figure which shows one result.

Based on the coordinate transformation result, the prespace in the input image using the distance sensor can be set. Values input through the distance sensor form points as shown in FIG. If the data value from the distance sensor is converted into image coordinates, the coordinates in the image are not continuous as shown in FIG.

Therefore, the value input from the distance sensor is set to have a continuous value for one object through a data clustering process. When the coordinates of the distance data input from the distance sensor are converted from the polar coordinate system to the Cartesian coordinate system, and the change amount of the x-axis coordinate and the change amount of the y-axis coordinate are below a threshold value, for example, 50 cm or less. Classify into one cluster.

The free space is below the boundary line of the recognized data cluster and below the horizon line (or the top line of the region of interest of the third stage described later). As such, by limiting the space below the cluster boundary to the free space, obstacles forming the cluster can be eliminated, thereby achieving the object of the present invention.

FIG. 9 is a diagram illustrating a prespace extracted by using the input values of FIGS. 7 and 8 in the prespace setting step (S208). In the prespace, the pixel value is set to '1', and thus the white space is shown as shown in FIG. By setting the prespace in this way, it is possible to prevent the lane from being erroneously recognized by an obstacle such as a vehicle.

After the step S208 is completed, a template matching algorithm is applied to the filter result to find a line component existing in the prespace (S210). Here, the Hough transform may be used as the template matching algorithm.

Hough transform is first performed to find a lane from the filter image extracted from the steerable filter. Hough transform is used to find the linear components in the image. Hough transform extracts possible linear components at each pixel. Equation 10 is a conversion equation of x-y coordinates for the Hough transform.

Detecting the point where ρ and θ varying according to the position of each pixel can find the partial pixel having the largest linear component in the image.

FIG. 10 is a diagram illustrating a result of the Hough transform of the filter image in the linear component extraction step (S210).

FIG. 10A is a filter image, and FIG. 10B is a result of Hough transforming the filter image, and FIG. 10C is a diagram in which the Hough transform result is mapped (red line) to the image.

As described above, the lane display candidate group is generated using the Hough transform result. In particular, since the own lane area is generated based on the inner line of the lane display, the inner lane display portion is set to the lane display candidate group pixel. Therefore, the nearest pixel in the lane line region is found based on the straight line shown in the Hough transform result.

FIG. 11 is a diagram illustrating pixels closest to the own lane in the Hough transform result in the line component extraction step (S210).

After the line component extraction step S210 is finished, a region of interest (ROI) may be set (S212).

As described above, it may be set before the step S214 of finding a lane by line fitting.

The lanes shown in the video frame may be straight or curved depending on the shape of the road. The region of interest (ROI) required for lane recognition is set to include the shape of the lane while minimizing the range to be searched.

The preferred lane shape should be such that only lanes can be included and other objects can be excluded. The lane is divided into three stages so that the ROI becomes smaller as the lane looks smaller, considering that the lane looks smaller as it moves further away. The lane is divided into a left lane and a right lane. Therefore, six sub ROIs may be set in three stages.

As shown in FIG. 3, there are six ROIs, three each on the left and right sides. That is, the first-level ROI may be divided into sub-ROI I and IV, the second-ROI may be sub-ROI II and V, and the third-ROI may be sub-ROI III and VI.

The sub-ROIs I and IV close to the vehicle may be fixed, and the remaining sub-ROIs may vary depending on the following equation. That is, the second-level ROI may vary according to the first-level ROI, and the three-step ROI may vary according to the second-ROI.

For example, by dividing up and down 1/3 points from the bottom of the camera image to left and right, sub ROI I and IV are designated to designate the ROI stage 1, and the ROI of the subsequent stage is the lane within the ROI stage 1 After recognizing this, it may be set step by step using the recognized lane information.

In the ROI setting, the ROI of the second stage is set based on the lane information recognized in the first stage, and the ROI of the third stage is set using the lane information recognized in the second stage. By setting the ROI in stages, there is an advantage of eliminating false lane recognition probability due to ambient noise. In addition, the lane of the first stage, that is, the near lane, can be recognized as a straight line. Therefore, there is an advantage that it is easy to find lanes in the first stage and narrow the ROI in the second and third stages of the curve, which increases the probability of finding the lane. However, the noise caused by the front vehicle or the interrupting vehicle is difficult to remove by ROI using only the image.

In the input image from the camera, the lanes appear as edges with arbitrary angle components. The best way to find the lane is to extract the edge responsiveness for any angle. In particular, the obstructing shadows of the front and side lanes have different angles from the lanes, so if you know the lane angle in the input image, you can easily extract the edge components of the input image from the input image and exclude the components that make noise around you. have.

After the region of interest setting step S212, line fitting is performed on the line components within the set region of interest and step S214 is performed.

The lane marking candidate group pixels found as a result of the Hough transform may be found by performing line fitting. The line fitting may be made of two types: linear fitting and curve fitting.

Equation (11) represents a straight line fitting, and Equation 12 represents a curve fitting.

Since the first-stage ROI can be approximated by a straight line, line fitting can be performed with a straight fitting, and the point where the straight line obtained by using the line fitting result meets the boundary of the second-stage ROI and the first-stage ROI It is the x coordinate center of the ROI.

As described above, when the lane of the first-level ROI is extracted, the lane of the camera image frame may be extracted by performing the step-by-step process of extracting the lane into the second-phase ROI and the three-step ROI.

In addition, it may be performed in the second region of interest based on the result of the first region of interest.

Since the second-stage ROI can be approximated by a straight line or a curve, line fitting can be performed by linear or curve fitting. The line obtained using the line fitting result meets the boundary line between the two-stage ROI and the three-stage ROI. Step 3 becomes the center of the x coordinate of the ROI.

After the line fitting in the first stage of interest region is finished, the line fitting in the second stage of interest region is performed in the same manner as in the first stage of interest region. In this case, the second-stage ROI may be set after extracting a lane from the first-stage ROI.

In the second-level ROI, without using the steerable filter (S204), the binarization step (S206), the pre-space setting (S208), and the hough transform (S210), which were performed to select the lane-display candidate group pixels in the first-step ROI. Next, the lane display candidate group pixels are selected based on the fitting line and the own lane area using the fitting information acquired in the first region of interest. The method of setting the third-level ROI in the second-level ROI is the same as the method of setting the second-ROI in the first-ROI.

FIG. 12 is a diagram illustrating an example of setting a two-step ROI and a three-step ROI in the ROI setting step S212.

Lane recognition is obtained by performing a straight line fitting of the lane display candidate groups of the areas set in steps 1 to 3 together. You can also use Kalman filters for lane tracking. As the measurement matrix of the Kalman filter, left lane angle, left lane offset, right lane angle, and right lane offset were used. That is, tracking is performed by using slopes and offsets of the input left and right lanes.

The angle of the lane marking of the next frame uses the angle of the lane marking that appears in the fitting result using the straight fitting result. The found angle is used as the input angle of the steerable filter in the next frame. Therefore, since each piece of information uses the information of the previous frame, the input angle of the steerable filter was obtained by using EDF in the first frame, but the calculation amount can be reduced because the angle obtained in the previous frame is input in the subsequent frame.

FIG. 13 is a diagram illustrating a result of recognizing a lane component (red line) by fitting a straight line in the lane extraction step (S214).

14 is a diagram illustrating a lane recognition system according to an embodiment of the present invention.

Lane recognition system using a distance sensor according to an embodiment of the present invention, the camera 1402 for generating a video frame by photographing the surrounding vehicle; An image receiver 1403 which receives the image frame; A distance sensor 1404 for generating distance data for detecting obstacles around the host vehicle; A distance data receiver 1405 for receiving the distance data; A filter unit 1406 for extracting a filter result by using a steerable filter on the image frame; A prespace setting unit 1408 for setting a prespace using the distance data; A line component extractor 1410 which finds a line component existing in the prespace by applying a template matching algorithm to the filter result; A region of interest setting unit 1411 for setting a region of interest (ROI) and a lane extracting unit 1412 that performs line fitting on the line component in the region of interest and extracts a lane, and a filter valid only for pixels having a predetermined value or more. And a binarization unit 1414 that performs binarization to recognize the result.

In some cases, the binarization unit 1414 may be omitted.

2 to 13 and FIG. 14 together illustrate a lane recognition system according to an embodiment of the present invention.

The camera 1402 generates an image frame by photographing the surrounding vehicle, and the distance sensor 1404 generates data for detecting an obstacle around the vehicle.

The image receiver 1403 and the distance data receiver 1405 perform an image frame and distance data receiving step S202.

The filter unit 1406 performs the filter result extraction step (S204), the prespace setting unit 1408, the binarization unit 1414 performs the binarization step (S206), and performs the prespace setting step (S208). .

The line component extractor 1410 performs the line component extraction step S210, the ROI setting unit 1411 performs the ROI setting step S212, and the lane extractor 1412 performs the lane extracting step S214. The object of the present invention can be achieved by

Meanwhile, the line component extractor 1410 may perform a Hough transform using a template matching algorithm.

The region of interest may be divided into a plurality of stages, in which the region of interest of the first stage is fixed, and subsequent regions of interest may be set using information on lanes of the region of interest of the previous stage.

Meanwhile, the lane slope used when the filter unit 1406 applies the steerable filter is an angle extracted using an edge distribution function (EDF) in the case of an initial image frame, and the previous image in the case of a subsequent image frame. The slope of the lane extracted from the frame may be used.

The above description is merely illustrative of the technical idea of the present invention, and those skilled in the art to which the present invention pertains may make various modifications and changes without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention but to describe the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The protection scope of the present invention should be interpreted by the following claims, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of the present invention.

According to the present invention, by using the distance sensor to obtain obstacle information that is difficult to know only by the camera information, the lane is recognized, and ROI (Region of Interest, ROI) from which the obstacle is removed using the obstacle information from the image input from the camera By setting it, it is effective to recognize lanes using simple lane recognition algorithms using distance sensor information such as noise (edge information generated by the front vehicle) that obstructs lane recognition due to obstacles, and whether lanes are covered. have.

In addition, since the distance sensor can be used already installed in the vehicle for the adaptive cruise control (ACC), lane recognition using the obstacle information on the road using only one camera without additional cost increases. It is possible.

1 is a diagram illustrating lane recognition when a conventional lane recognition system is used.

2 is a flowchart illustrating a lane recognition method according to an embodiment of the present invention.

3 is a diagram illustrating that a three-step ROI is set step by step in the region of interest setting step S212.

4 is a diagram illustrating a stiffable filter result in a camera image in the filter result extraction step (S204).

FIG. 5 is a diagram illustrating a result using a Sobel operator on the input image of FIG. 4A in the filter result extraction step (S204).

FIG. 6 is a graph illustrating an EDF for the input image of FIG. 4A in the filter result extraction step S204.

FIG. 7 is a diagram illustrating coordinates of data obtained from a distance sensor such as a scanning laser radar in a prespace setting step (S208).

FIG. 8 is a diagram illustrating a result of converting data coordinates from a distance sensor into image coordinates in the prespace setting step (S208).

FIG. 9 is a diagram illustrating a prespace extracted by using the input values of FIGS. 7 and 8 in the prespace setting step (S208).

FIG. 10 is a diagram illustrating a result of the Hough transform of the filter image in the linear component extraction step (S210).

FIG. 11 is a diagram illustrating pixels closest to the own lane in the Hough transform result in the line component extraction step (S210).

FIG. 12 is a diagram illustrating an example of setting a two-step ROI and a three-step ROI in the ROI setting step S212.

FIG. 13 is a diagram illustrating a result of recognizing a lane component (red line) by fitting a straight line in the lane extraction step (S214).

14 is a diagram illustrating a lane recognition system according to an embodiment of the present invention.

<Description of Symbols for Main Parts of Drawings>

1402: camera 1403: video receiver

1404: distance sensor 1405: distance data receiver

1406: filter unit 1408: free space setting unit

1410: line component extractor 1411: ROI setting unit

1412: lane extractor 1414: binarization

Claims (10)

In the lane recognition method using a distance sensor, (a) receiving an image frame and distance data photographing the surrounding vehicle; (b) extracting filter results from the image frame using a steerable filter; (c) setting a prespace using the distance data; (d) applying a template matching algorithm to the filter result to find a line component present in the prespace; (e) establishing a Region Of Interest (ROI); And (f) performing line fitting on the line components in the region of interest and extracting lanes; Lane recognition method comprising a. The method of claim 1, wherein in step (b), Lane gradient used when applying the steerable filter, The first image frame is an angle extracted using an edge distribution function (EDF), and the subsequent image frame is a slope of the lane extracted from the previous image frame. The method of claim 1, After step (b), And further comprising a binarization step of recognizing only a pixel having a predetermined value or more as a valid filter result. The method of claim 1, wherein the region of interest is: The lane recognition method is divided into a plurality of stages, wherein the region of interest in the first stage is fixed and the subsequent region of interest is set using information on the lane of the region of interest in the previous stage. The method of claim 1, wherein the template matching algorithm, Lane recognition method characterized in that the Hough transform. In lane detection system using a distance sensor, A camera for photographing the surrounding vehicle to generate an image frame; A distance sensor for generating distance data for detecting obstacles around the host vehicle; An image receiver which receives the image frame; A distance data receiver for receiving the distance data; A filter unit which extracts a filter result by using a steerable filter on the image frame; A prespace setting unit configured to set a prespace using the distance data; A line component extracting unit which finds a line component existing in the prespace by applying a template matching algorithm to the filter result; A region of interest setting unit for setting a region of interest (ROI); And A lane extracting unit performing line fitting to the line component in the region of interest and extracting a lane; Lane recognition system comprising a. The method of claim 6, Lane gradient used when the filter unit applies the steerable filter, In the case of an initial image frame, the angle is extracted using an edge distribution function (EDF), and in the case of a subsequent image frame, the lane is an inclination of the lane extracted from a previous image frame. The method of claim 6, The lane recognition system, And a binarization unit configured to perform binarization for recognizing only pixels having a predetermined value or more as valid filter results. The method of claim 6, wherein the region of interest is: The lane recognition system of claim 1, wherein the region of interest of the first stage is fixed and the subsequent region of interest is set using information on the lane of the region of interest of the previous stage. The method of claim 1, wherein the template matching algorithm, Lane recognition system, characterized in that the Hough transform.

Images