KR101501851B1 - Apparatus and method for lane detection using hough transformation at optimized accumulator cells - Google Patents

Abstract

A method for detecting a traffic line using Hough transformation according to the present invention, comprises: an area of interest setting step of setting an area of interest in an image taken by a camera installed in a vehicle; a noise removal step of: removing noise by applying a Gaussian filter to the image, wherein the area of interest has been set in the image; an edge image generation step of generating an edge image by detecting an edge included in the image from which noise is removed; and a traffic line detection step of detecting a traffic line by applying Hough transformation in accumulator cells within a predetermined range in regard to the edge image. According to the present invention, even though a road environment is harsh, a traffic line can be detected with a high recognition rate. Also, by reducing unnecessary operations in a Hough transformation algorithm for traffic line detection, the speed of traffic line detection can be improved, and at the same time by preventing an unnecessary straight line component except for a traffic line from being detected, the accuracy of traffic line detection can be improved.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and a method for detecting a lane using a Huff transformation in an optimized cumulative cell,

The present invention relates to a lane detecting apparatus and method using Hough transform in an optimized accumulation cell.

Recently, various methods for detecting lanes have been studied, but they are showing limitations in detecting lanes with uniform high recognition rate in various road environments.

One such conventional technique is a lane departure warning apparatus using a laser. However, according to this method, since it is difficult to detect lanes due to the narrow range, there are many factors to be considered in the installation because of the large size of the apparatus. In the case that deformation occurs in the structure such as a mirror due to external factors, Errors occur when there are other objects or pedestrians, and it takes a long time and processing time to reflect by irradiating the laser, and it is not effective in situations such as shadows, light scattering, artificial objects on the road, There are several problems such as.

As another example of the prior art, Patent Document 1 discloses a lane-based high-speed detection method by extracting a region of interest. However, this method removes the shadow part through filtering and recognizes the lane. By removing the shadow part from the captured image through filtering, it reduces the error in the lane recognition, whereas in the lane recognizing part, The lane information is generated and the recognition rate is not high. That is, this method is a technique for reducing only the influence of shadows in lane recognition, and is unsuitable as a method for recognizing a lane boundary in a severe road environment, and fails to prepare a follow-up countermeasure if lane recognition fails.

Korean Patent Publication No. 10-2011-0001427

An object of the present invention is to provide an apparatus and method for detecting lanes using Huff transformation in an optimized cumulative cell capable of detecting lanes with high recognition rate in harsh road environments.

Another object of the present invention is to improve the accuracy of lane detection by increasing the lane detection speed by reducing the amount of unnecessary calculation in the Huff transformation algorithm for lane detection and by preventing an unnecessary straight line component other than the lane from being detected.

According to an aspect of the present invention, there is provided a method of detecting lanes using Hough transform, the method comprising: setting a ROI on an image captured by a camera mounted on a vehicle; An edge image generation step of generating an edge image by detecting an edge included in the noise-removed image, and a step of generating an edge image in a predetermined range of accumulator cells And a lane detection step of detecting a lane by applying Hough Transformation.

The method of detecting a lane using Hough transform according to the present invention may further comprise a step of converting the image of interest region into a monochrome image after the step of setting the region of interest.

In the lane detection method using Hough transform according to the present invention, the lane detecting step may include an angle calculating step of calculating the values of the Hough area within a range of a minimum threshold value and a maximum threshold value determined for the left lane and the right lane, And a distance calculating step of calculating a ρ value by applying a Hough equation to the θ average value.

In the lane detection method using Hough transform according to the present invention, the minimum threshold value of the left lane is 30 degrees, the maximum threshold value is 60 degrees, the minimum threshold value of the right lane is 120 degrees, and the maximum threshold value is 150 degrees .

In the lane detecting method using Hough transform according to the present invention, the lane detecting step is performed in parallel with respect to the left lane and the right lane.

In the lane detection method using Hough transform according to the present invention, the ROI is limited to a region including a lane.

The lane detecting method using Hough transform according to the present invention is characterized by further comprising a lane image displaying step of displaying a lane image detected by the lane detecting step to a user.

The lane detection apparatus using Hough transform according to the present invention includes a ROI setting unit for setting a ROI on an image captured by a camera mounted on a vehicle, a noise eliminating unit for applying noise filtering to GI- An edge image generation unit for generating an edge image by detecting an edge included in the noise-removed image, a Hough transformation unit for performing a Hough transform in an accumulator cell within a predetermined range for the edge image, And a lane image display step of displaying the image of the lane detected in the lane detecting step.

The lane detecting apparatus using Hough transform according to the present invention may further comprise a monochrome image converting unit for converting the image of interest area set by the ROI setting unit into a monochrome image.

In the lane detecting apparatus using the Hough transform according to the present invention, the lane detecting section may include an angle calculating section for calculating the values of the Hough area within a range of a minimum threshold value and a maximum threshold value determined for the left lane and the right lane, And an angle averaging unit for averaging the values to calculate a? Average value, and a distance calculator for calculating a? Value by applying a Hough equation to the? Average value.

In the lane detecting apparatus using Hough transform according to the present invention, the minimum threshold of the left lane is 30 degrees, the maximum threshold is 60 degrees, the minimum threshold of the right lane is 120 degrees, and the maximum threshold is 150 degrees .

In the lane detecting apparatus using Hough transform according to the present invention, the lane detecting unit may perform the process of detecting the lane in parallel with the left lane and the right lane.

In the lane detecting apparatus using Hough transform according to the present invention, the ROI is limited to a region including a lane.

The lane detecting apparatus using Hough transform according to the present invention further includes a lane image display unit for displaying the image of the lane detected by the lane detecting unit to the user.

According to the present invention, there is provided an apparatus and method for detecting a lane using Huff transformation in an optimized accumulation cell capable of detecting a lane with a high recognition rate in a severe road environment.

Further, the unnecessary calculation amount in the Hough transform algorithm for lane detection is reduced to increase the lane detection speed, and at the same time, an unnecessary straight line component other than the lane is not detected, thereby improving the accuracy of lane detection.

1 is a diagram illustrating an example of a configuration of a device to which a lane detection method using Hough transform according to an embodiment of the present invention is applied.
2 is a diagram illustrating a lane detection method using Hough transform according to an embodiment of the present invention.
Fig. 3 is a diagram for explaining an example of a region-of-interest setting in this embodiment.
4 is a diagram for explaining a process of removing noise using a Gaussian filter in the present embodiment.
FIG. 5 is a diagram for explaining the relationship between the xy image space and the? -? Hough space.
6 is a diagram showing lane positions in the xy image coordinates in the present embodiment.
7 is a diagram for explaining a Hough transform algorithm for lane detection in this embodiment.
8 is a diagram for defining the minimum and maximum threshold values of the θ values for the left lane and the right lane set for the Huff transformation in the optimized accumulation cell in the present embodiment.
FIG. 9 is a diagram showing a cumulative cell of a general Huff area plane.
10 is a diagram showing an optimized accumulation cell in which a Hough transform algorithm is performed in the present embodiment.
11 is a diagram showing a left lane boundary detected in the intermediate process of the present embodiment.
12 is a diagram showing a right lane boundary detected in the intermediate process of the present embodiment.
13 is a diagram showing the result of applying the Hough transform algorithm of the present embodiment to a lane image passed through Gaussian filter.
FIGS. 14 and 15 are views showing examples of result screens finally provided to the user according to the present embodiment.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a diagram illustrating an example of a configuration of a device to which a lane detection method using Hough transform according to an embodiment of the present invention is applied. FIG. 2 illustrates a lane detection method using a Hough transform according to an embodiment of the present invention. FIG.

Referring to FIGS. 1 and 2, a lane detection method using Hough transform according to an embodiment of the present invention includes a step of setting a region of interest (S10), a monochrome image converting step (S20), a noise removing step (S30) A generating step 40, a lane detecting step S50, and a lane image displaying step S60.

In the ROI setting step S10, the ROI setting unit 10 sets a region of interest (ROI) on the image captured by the camera mounted on the vehicle. The area of interest is limited to the area including the lane in the photographed image. This process is used to acquire only the area where the lane is photographed in the entire image. Through this process, the calculation time for the entire algorithm application process is reduced, and the operation speed is improved. Also, since the upper part of the entire image other than the lower part where the lane exists is removed, it is also effective to remove the noise.

In the black-and-white image conversion step S20, the black-and-white image conversion unit 20 converts the image of interest into the black-and-white image in order to reduce the amount of calculation and to detect the lane efficiently.

In the noise removing step S30, the noise removing unit 30 removes noise by applying Gaussian filtering to the black-and-white image in which the ROI is set. In the edge image generating step S40, the edge image generating unit 40 ) Detects an edge included in the noise-removed image and generates an edge image. The noise removing step S30 and the edge image generating step S40 are used to find out a boundary point of a lane and to remove an unnecessary part. Through this process, noise of other components other than the lane It is possible to increase the calculation time and accuracy.

In this process, a Gaussian filter is applied to minimize the influence of noise in the image. That is, as shown in the following Equation 1, the 2D distribution is used as a Gaussian smoothing function to remove noise in the image.

, X is the distance from the origin to the horizontal axis, y is the distance from the origin to the vertical axis direction, and? Is the standard deviation of the Gaussian distribution.

The Sobel mask is a 3x3 line, estimating the slope of the Y direction (row), using the pair of estimated X-direction (column) and other slope, which is expressed by the following equation (2).

The magnitude and direction of the tilt are defined by the following equations (3), (4) and (5).

The gradient magnitude of the image represents the edge very clearly and the direction of the edge must be determined and stored as shown in Equation 5 below.

, Gx in the equation (5) is a horizontal component, Gy is a vertical component, and? Is a rounding edge direction angle to one of four angles representing two diagonal lines in the vertical and horizontal directions.

A high threshold and a low threshold according to an edge slope value and a slope higher than a high threshold value, a strong edge pixel, and a slope lower than the low threshold value, Weak edge pixels are suppressed and edge pixels between the two thresholds are displayed.

4 (a) shows an input image having noise. When a general edge detection process is performed without using a Gaussian filter, the noise shown in FIG. 4 (c) The edge image of the lane including the lane.

However, if the input image of FIG. 4 (b) is subjected to an edge detection process using a Gaussian filter, an edge image of the lane without noise can be obtained as shown in FIG. 4 (d).

In the lane detecting step S50, the lane detecting unit 50 performs a process of detecting a lane by applying Hough Transformation in an accumulator cell within a predetermined range with respect to the edge image. The lane detecting step S50 is performed in parallel with respect to the left lane and the right lane.

For example, the lane detecting step S50 may include an angle calculating step S510, an angle average calculating step S520, and a distance calculating step S530.

In the angle calculation step S510, the angle calculation unit 510 calculates the θ values, which are angles of the Hough area within the range of the minimum threshold value and the maximum threshold value determined for the left lane and the right lane based on the traveling direction of the vehicle . Here, the minimum threshold of the left lane is 30 degrees, the maximum threshold is 60 degrees, the minimum threshold of the right lane is 120 degrees, and the maximum threshold is 150 degrees.

In the angle average calculation step S520, the angle average calculation unit 520 averages the calculated? Values to calculate the? Average value.

In the distance computing step S530, the distance computing unit 530 computes the ρ value by applying the Hough equation to the θ average value.

In the lane image display step S60, a process of displaying the lane image detected by the lane image display unit 60 in the lane detecting step S50 to the user is performed.

Hereinafter, the algorithm applied in the lane detection step S50, that is, the Huff transformation algorithm in the optimized accumulation cell, which is a main component of the present embodiment, will be described in further detail with reference to FIGS. 5 to 15.

Hough transform is a method that has a good effect for detecting it when it can be represented by parameters such as a line slope or an intersection point in an image.

The Hough transform algorithm can identify all the lines passing through the point in the process of iteratively calculating the values corresponding to all possible angles. The straight line passing through the point can be expressed by the following equation (6) Express.

The Hough transform generates parameters of θ and ρ and is represented by a sinusoidal curve in the Hough area and is transformed by the edge coordinates of the image coordinates (x, y).

The geometric analysis of the variables used in Figs. 5 and 6 for explaining the relation between the x-y image space and the p-theta Hough space is as follows. ρ is the length from the origin to the straight line, and θ is the angle between ρ and x axis. x and y mean the horizontal and vertical axes in the image. The horizontal line is θ = 0 degrees, and ρ is the same as the positive x intercept. The vertical line is θ = 90 degrees, ρ is the positive y intercept or θ = -90 degrees, and ρ is the negative y intercept.

Through this function, a point in the x-y plane of the image is displayed as a curve in the Hough area (ρ-θ plane). Therefore, in order to find a straight line through it, we find a point where the line is gathered in the Huff area, and that point is the accumulator cells.

Generally, all the θ values are calculated to recognize all the straight lines in the x-y plane of the image, so the boundaries of the lane and other straight lines are also detected, and the result is in a deteriorated state. Therefore, the present embodiment proposes an algorithm in which accumulator cells are optimized in the Huff area to solve this problem.

In this embodiment, the boundary of the lane is divided into left and right parts using Hough transform. The flowchart of the algorithm shown in FIG. 7 is shown below. The basic concept of this algorithm is to optimize the accumulator cells in the Huff area. First, the boundary of the lane is applied to the x-y plane, as shown in Fig.

6 shows two lanes consisting of a left lane boundary and a right lane boundary in the 2D plane of the road input image. ρ1 is the vertical distance between the origin and the left lane boundary, θ1 is the angle between the y-axis and the left lane boundary, and the vehicle's direction is the y-axis direction. ρ2 is the vertical distance between the origin and the right lane boundary, and θ2 is the angle between the y-axis and the right lane boundary. Here, the gradient vector maintains an angle of 90 degrees with the lane to be detected and output.

FIG. 7 shows a flowchart for the proposed Hough transform algorithm.

Referring to FIG. 7, when one frame of an image comes in as an input and an image is read in the ROI, an algorithm for detecting the left lane and the right lane in the ROI is executed in parallel. In the case of the left lane detection, if? 1 is larger than LLI in the left region,? 1 is calculated through Equation (6). Then, ρ1 and θ1 are stored in the (x, y) area and the corresponding detection lane is output. Also, the angle of the next frame (θnext) is the value changed by θ1. If θ1 is equal to HLI + Δθ, the next frame is applied to the algorithm, and if it is not the same, it is feedback to the part that calculates ρ1. In the same algorithm, the right lane is detected in the right area, and the detected lanes are simultaneously output in one frame.

Here, the minimum and maximum threshold values for the left lane and the right lane, which are set for the Hough transform in the optimized accumulation cell, are defined with reference to FIG.

8, assuming that the traveling direction of the vehicle is the y-axis direction, and the vehicle is located at the center of the roadway, the lane is located at 45 degrees on both sides of the traveling direction of the vehicle, The accumulator cell data is collected by 95% or more. Therefore, the LLI (Lower Left Interval) is set at 30 degrees, the HLI (High Left Interval) is set at 60 degrees, the LRI (lower right interval) is set at 120 degrees, and the HRI (high right interval) is set at 150 degrees.

Figure 9 is accumulator cells of a general Huff area plane. The (? Min,? Max) and (? Min,? Max) of the Huff space are the expected ranges of the parameter values. Usually, the range of these values is -90 degrees??? 90 degrees, -D?? Where D is the distance from the origin to the opposite vertex in x-y space.

10 is a diagram showing an optimized accumulation cell in which a Hough transform algorithm is performed in the present embodiment.

Referring to FIG. 10, for the left lane boundary detection, theta value between LLI and HLI is calculated. And straight lines extracted from the accumulator cells of that portion were selected.

Also, for the right lane boundary detection, straight lines extracted from the part between LRI and HRI were selected.

When LLI1 &thetas; ≤ HLI is extracted by applying DELTA &thetas; 1 in the accumulator cells in Fig. 11, n straight lines are recognized in the left lane.

12, when LRI ≤ &thetas; 2 ≤ HRI is extracted as DELTA &thetas; 2, m straight lines are recognized in the other right lane. In the case of the conventional Hough transform, it is difficult to recognize an accurate lane by recognizing a plurality of lines in the ROI. In this embodiment, only the required lane can be detected and output through the improved accumulator cells of the proposed algorithm .

FIG. 13 is a diagram of the proposed algorithm applied to a lane image passing through a Gaussian filter. As described with reference to FIGS. 11 and 12, it can be seen that a plurality of lanes are superimposed on each of the left lane and the right lane. However, since the lane boundary of the proposed algorithm does not require a plurality of lanes and only one lane is required for both lanes, the average value of? Is calculated as shown in the following Equations (7) and (8).

이다.Taking the result calculated through equations (7) and (8), that is, the average value of?,? Is expressed as the following equations (9) and (10), and the range of the axes of? 1 and?

to be.

(θ1, ρ1) to recognize the left lane boundary and the right lane boundary (θ2, ρ2).

Table 1 below shows the results of actual application of the present embodiment to lane detection, and FIGS. 14 and 15 show examples of result screens ultimately provided to the user according to the present embodiment.

From these results, it can be seen that when the present embodiment is applied to lane detection, an average hit rate of 95% or more is obtained.

As described above in detail, according to the present invention, there is provided an apparatus and method for detecting lanes using Huff transformation in an optimized cumulative cell capable of detecting lanes with high recognition rate in a severe road environment.

Further, the unnecessary calculation amount in the Hough transform algorithm for lane detection is reduced to increase the lane detection speed, and at the same time, an unnecessary straight line component other than the lane is not detected, thereby improving the accuracy of lane detection.

While the present invention has been described in connection with what is presently considered to be preferred embodiments, it is to be understood that the invention is not limited to the disclosed embodiments. In addition, it is a matter of course that various modifications and variations are possible without departing from the scope of the technical idea of the present invention by anyone having ordinary skill in the art.

10: region of interest setting section
20: B &
30: Noise canceling
40: edge image generating unit
50: lane detecting section
60: lane image display section
510:
520: Angle average calculating unit
530:
S10: Interest zone setting step
S20: B & W image conversion step
S30: Noise removal step
S40: Edge image generation step
S50: lane detection step
S60: Lane image display step
S510: Angle calculation step
S520: Angle average calculation step
S530: Distance calculation step

Claims (14)

A method for detecting a lane using Hough transform in a cumulative cell,
An interest area setting step of setting an area of interest in an image taken by a camera mounted on a vehicle;
A black and white image conversion step of converting the image in which the ROI is set into a monochrome image;
A noise removal step of removing noise by applying Gaussian filtering to the black-and-white image in which the ROI is set;
An edge image generation step of generating an edge image by detecting an edge included in the noise-removed image; And
And a lane detection step of detecting a lane by applying Hough Transformation in an accumulator cell within a predetermined range for the edge image,
The lane detection step
An angle calculating step of calculating the H values of the Huff area within a predetermined minimum threshold value and a maximum threshold value with respect to the left lane and the right lane;
An angle average calculation step of averaging the calculated? Values and calculating a? Average value; And
And a distance calculating step of calculating a? Value by applying a Hough equation to the? Average value.
delete delete The method according to claim 1,
Wherein the minimum threshold of the left lane is 30 degrees, the maximum threshold is 60 degrees, the minimum threshold of the right lane is 120 degrees, and the maximum threshold is 150 degrees. The method according to claim 1,
Wherein the lane detecting step is performed in parallel with respect to the left lane and the right lane. The method according to claim 1,
Wherein the area of interest is limited to a region including a lane. The method according to claim 1,
And a lane image display step of displaying a lane image detected by the lane detecting step to a user. A lane detecting device using Hough transform in a cumulative cell,
An interest area setting unit for setting an area of interest in a video image captured by a camera mounted on a vehicle;
A black-and-white image converting unit converting the image of the ROI set by the ROI setting unit into a monochrome image;
A noise eliminator for removing noise by applying Gaussian filtering to the black-and-white image in which the ROI is set;
An edge image generation unit for generating an edge image by detecting an edge included in the noise-removed image; And
And a lane detection unit for detecting a lane by applying Hough Transformation in an accumulator cell within a predetermined range for the edge image,
The lane-
An angle calculator for calculating θ values of the Hough area within a predetermined minimum threshold value and a maximum threshold value with respect to the left lane and the right lane;
An angle averaging unit for averaging the calculated? Values to calculate a? Average value; And
And a distance arithmetic unit for calculating a value of? By applying a Hough equation to the? Average value.
delete delete 9. The method of claim 8,
Wherein the left lane minimum threshold is 30 degrees, the maximum threshold is 60 degrees, the right lane minimum threshold is 120 degrees, and the maximum threshold is 150 degrees. 9. The method of claim 8,
Wherein the lane detecting unit performs the process of detecting the lane in parallel with respect to the left lane and the right lane. 9. The method of claim 8,
Wherein the area of interest is limited to a region including a lane. 9. The method of claim 8,
And a lane image display unit for displaying the image of the lane detected by the lane detecting unit to the user.

Images