KR20210066430A - Apparatus for detecting edge of road image and method thereof - Google Patents

Abstract

The present invention relates to an apparatus for detecting a boundary line of a road image and a method thereof. An apparatus for detecting a boundary line of a road image in accordance with the present invention comprises: a preprocessing unit performing a preprocess of converting each frame unit image, extracted from the road image, into a grayscale image; a dividing unit dividing the preprocessed image into multiple areas with the same size; an extraction unit extracting a determination value for a central area of a window by using the window with a preset size and extracting determination values for all areas of the image while shifting the window in the image; and a boundary line output unit displaying and outputting the boundary line with respect to the image by using the extracted determination values for all areas. The present invention reduces a calculation amount to enhance the detection speed by minimizing the preprocessing step when the boundary line for extracting feature points of the road image on an embedded system is detected.

Description

Apparatus and method for detecting a boundary line of a road image {APPARATUS FOR DETECTING EDGE OF ROAD IMAGE AND METHOD THEREOF}

The present invention relates to an apparatus and method for detecting a boundary of a road image, and more particularly, to an apparatus and method for detecting a boundary of a road image for clearly detecting a boundary for extracting feature points of a road image on an embedded system.

The technology of extracting edges from an image is a technology used in various image processing technology fields, and is used to acquire boundary information itself of an image, as well as to identify feature points of objects in an image, or to mean an image. It is used for a variety of purposes, such as being used to divide the existing partial image into units or to analyze the properties of the image itself.

In particular, in a digital image, a boundary line is a part in which the brightness of an image changes rapidly, and means major changes in the image including a change in depth, a change in a surface direction, a change in a surface material, a change in lighting, and the like.

In the edge detection technique for detecting the edge of an image, there are various methods, from a method of extracting the edge by processing only the high frequency component by analyzing the frequency component of the entire image to a method of extracting the edge of the image through regional filtering. It exists and is widely used in the field of image processing and computer vision.

Through boundary detection, it is possible to significantly reduce the amount of data to be processed by emphasizing the prominent part and removing the less meaningful part, and it is particularly used to extract the features of a certain object.

Representative edge detection techniques include the Sobel edge detection method and the Canny edge detection method. First, the Sobel edge detection method is an algorithm developed by Irwin Sobel in 1968. The value of the surrounding pixels is calculated using the x, y Sobel filter, and if the value is close to 0, it is not a boundary line (edge) because there is no change, and if the value is large, it is recognized as a boundary line because the amount of change in a specific direction is large. In addition, a large mask has a thicker edge so that the boundary line becomes clear, but there is a problem in that the effect is low in the case of a dense or complex image with a change section of the contrast value.

In addition, the Canny boundary detection technique is an algorithm developed by Jone F. Canny in 1986. The optimization goal is to obtain a high signal to noise ratio (SNR), and Canny formulated a mathematical expression for this purpose, An 80% optimization was successfully achieved using the first derivative of the Gaussian approximation. However, since the Canny boundary line detection technique is more sensitive to weak boundaries and declares an unstable part as a boundary, there is a problem of obtaining a damaged boundary detection result and a slow speed.

Such conventional boundary detection techniques are configured in the form of a library including all algorithms, and several algorithms are complexly configured to produce the best results. However, since embedded systems generally use the minimum amount of computation for application optimization, large-capacity library-type image software is difficult to process in real time due to hardware limitations in embedded systems. There is a problem with the design. Therefore, it is necessary to develop a unit design algorithm suitable for embedded systems.

The technology that is the background of the present invention is disclosed in Korean Patent No. 10-155660 (announcement on October 01, 2015).

An object of the present invention is to provide an apparatus and method for detecting a boundary line of a road image for clearly detecting a boundary line for extracting feature points of a road image on an embedded system.

According to an embodiment of the present invention, there is provided an apparatus for detecting a boundary line from a road image, comprising: a pre-processing unit for pre-processing each frame unit image extracted from the road image into a gray scale image; a partitioning unit partitioning the pre-processed image into a plurality of regions having the same size; An extraction unit extracting judgment values for the central region of the window using a window of a preset size, and extracting judgment values for all regions of the image while shifting the window in the image ; and a boundary line output unit for displaying and outputting boundary lines for the image by using the judgment values for all the extracted regions.

In addition, the window is set to have a size of an n x n area, and may be set to a size of a square matrix in which n is an odd number.

In addition, the extractor may extract the determination value for the central region of the window using the following equation.

Here, P _avg is the average value of the pixel values of each area included in the window, Pi is the pixel value of the i-th area, n ² is the size of the window, W is the weight value according to the window, and P _c is the center area of the window. A pixel value, k is a fixed constant, and p' is a judgment value for the center region of the window.

In addition, the boundary line output unit may output the boundary line of the image in gray scale by using the determination value of each area extracted as 0 or 1, but display the area corresponding to the determination value 1 as the boundary line and output it.

Also, the pre-processing unit may pre-process the gray scale-converted image by applying a Gaussian filter.

In addition, the method for detecting a boundary line of a road image according to an embodiment of the present invention includes the steps of pre-processing by converting each frame-by-frame image extracted from the road image into a gray scale image; partitioning the preprocessed image into a plurality of regions having the same size; extracting judgment values for the central region of the window using a window of a preset size, and extracting judgment values for all regions of the image while shifting the window within the image; and displaying and outputting a boundary line for the image by using the judgment values for all the extracted regions.

As described above, according to the present invention, when a boundary line for extracting feature points of a road image is detected on an embedded system, the amount of computation is reduced and the detection speed is increased by minimizing the pre-processing process.

In addition, according to the present invention, it is possible to design a unit boundary detection algorithm, so that the weight of the embedded system is possible, and it can be applied to images in the field of machine learning such as detection of a specific object and recognition of style, so that it can be used in various applications.

1 is a block diagram illustrating an apparatus for detecting a boundary line of a road image according to an embodiment of the present invention.
2 is a diagram for explaining a process of extracting a determination value of a window center region in the apparatus for detecting a boundary line of a road image according to an embodiment of the present invention.
3 is a flowchart illustrating an operation flow of a method for detecting a boundary line of a road image according to an embodiment of the present invention.
4 is a diagram for comparing the performance of a conventional boundary line detection method and a detection method according to an embodiment of the present invention.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings. In this process, the thickness of the lines or the size of the components shown in the drawings may be exaggerated for clarity and convenience of explanation.

In addition, the terms described below are terms defined in consideration of functions in the present invention, which may vary according to the intention or custom of the user or operator. Therefore, definitions of these terms should be made based on the content throughout this specification.

First, an apparatus for detecting a boundary line of a road image according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2 .

1 is a block diagram illustrating an apparatus for detecting a boundary line of a road image according to an embodiment of the present invention.

As shown in FIG. 1 , the apparatus 100 for detecting a boundary line of a road image according to an embodiment of the present invention includes a pre-processing unit 110 , a dividing unit 120 , an extracting unit 130 , and a boundary line output unit 140 . .

First, the pre-processing unit 110 pre-processes each frame-by-frame image extracted from the road image by converting it into a gray scale image.

In this case, the preprocessor 110 may preprocess the image converted to gray scale by further applying a Gaussian filter.

Here, the Gaussian filter is a filter for removing the background noise of the image whose contrast is adjusted by gray scale processing, and it is a pre-processing technique that makes the surroundings invisible and smooth in the form of masking so that only the median value stands out.

In detail, the preprocessor 110 extracts a road image obtained through an imaging device (not shown) into a single frame-by-frame image, converts each extracted frame-by-frame image into a gray-scale image, and converts the converted image into a grayscale image. Preprocessing is performed using a Gaussian filter.

In this case, the road image acquired through the imaging device may be acquired as a single image based on a snapshot. In this case, it is preferable to omit the process of extracting the single image and directly convert the image into a gray scale image.

In addition, although the image obtained is defined as a road image in the embodiment of the present invention, any image may be applied as long as it is an image requiring boundary line detection, and the present invention is not limited thereto.

In addition, the partition unit 120 partitions the pre-processed frame-by-frame image into a plurality of regions having the same size.

The extraction unit 130 extracts a judgment value for the central region of the window using a window of a preset size, but determines all regions of the image while shifting the window within the image in units of frames extract the value

In this case, the window is set to have a size of an n x n area, and it is preferable to set the window to have a size of a square matrix in which n is an odd number.

For example, the size of the window may be set to the size of 3 x 3, 5 x 5, 7 x 7, ... n x n area, and the characteristic of the detected boundary line is different according to the window size.

In order to extract the judgment value for the central region of the window, the extractor 130 first _{calculates an average value P avg} of pixel values of each region included in the window using Equation 1 below.

Here, P _avg is the average value of the pixel values of each region included in the window, Pi is the pixel value of the i-th region, and n ² is the size of the window.

Then, the extraction unit 130 uses the average value (P _avg ) calculated in Equation 1 and the pixel value (P _c ) of the central region of the window, as in Equation 2 below, the weight value W according to the window to calculate

Here, W denotes a weight value according to the window, P _c denotes a pixel value of the central region of the window, and k denotes a fixed constant. In this case, the fixed constant k is preferably set to a constant between 0 and 0.1.

In addition, the weight value W is changed to a binary value according to the values of neighboring pixels, and varies according to the sum of these values and the pixel value of the original image.

Finally, the extractor 130 uses the weight value W according to the window calculated in Equation 2 and the pixel value P _c of the center area of the window as shown in Equation 3 below for the center area of the window. A judgment value p' is extracted.

Here, p' denotes a judgment value for the central region of the window.

2 is a diagram for explaining a process of extracting a determination value of a window center region in the apparatus for detecting a boundary line of a road image according to an embodiment of the present invention.

A process of extracting the determination value of the window center region using Equations 1 to 3 will be described with reference to FIG. 2 .

_{In FIG. 2 , a judgment value of the central region P 4 of} the window W is extracted using a window W having a size of 3×3 (that is, n=3), but for convenience of understanding, The calculation process will be described by substituting arbitrary numbers for pixel values.

_{First, the extractor 130 calculates the average value P avg} of the pixel values of each region of the frame unit image included in the window W by using Equation 1 as follows.

Pavg = (-124(P ₀ )-127(P ₁ )-121(P ₂ )+124(P ₃ )+126(P ₄ )+127(P ₅ )+122(P ₆ )+124(P ₇ ) +121(P ₈ ))/9 = 41

Then, the extractor 130 substitutes the average value (Pavg = 41) calculated by Equation 1 into Equation 2 to calculate the weight value W according to the window W as follows. At this time, it is assumed that the fixed constant k is 0.07 and is calculated.

W = ((126 * 0.07)/41) ² = 0.04627 (k=0.07)

Then, the extractor 130 substitutes the weight value (W=0.04627) calculated by Equation 2 and the pixel value (P ₄ =126) of the central region of the window W into Equation 3 to obtain the value of the window W. The judgment value (p') for the central region is calculated as follows.

Pc x W =126 x 0.04627 = 5.83xxxxx

In this case, |Pc x W| That is, since |5.83xxxxx| is greater than 1, the determination value p' for the central region of the window W becomes 1.

In addition, in order to explain another case, a different arbitrary number is substituted for the pixel values of each area.

_{First, the extractor 130 calculates the average value P avg} of the pixel values of each region included in the window W as follows by using Equation 1 .

P _avg =(-117(P ₀ )-116(P ₁ )-120(P ₂ )-116(P ₃ )-117(P ₄ )-120(P ₅ )-122(P ₆ )-122(P ₇ )-120(P ₈ ))/9 = -118.777

Then, the extraction unit 130 substitutes the average value (Pavg = -118.777) calculated by Equation 1 into Equation 2 to calculate the weight value W according to the window W as follows. At this time, it is assumed that the fixed constant k is 0.07 and is calculated.

W = ((-117 * 0.07)/-118.7) ² = 0.0048 (k=0.07)

Then, the extraction unit 130 substitutes the weight value (W=0.0048) calculated by Equation 2 and the pixel value (P ₄ =−117) of the central region of the window W into Equation 3 to obtain the window W The judgment value (p') for the central region of is calculated as follows.

Pc x W = -117 x 0.0048 = -0.5616xxxxx

In this case, |Pc x W| That is, since |-0.5616xxxxx| is not greater than 1, the determination value p' for the central region of the window W becomes 0.

Accordingly, the extraction unit 130 extracts judgment values for all regions of the image while shifting the window W within the frame-by-frame image as shown in FIG. 2 . That is, by shifting the window in the up, down, left and right directions by one space in the image so that each area becomes the center area of the window W, each judgment value for all areas is calculated using Equations 1 to 3 do.

Finally, the boundary line detection unit 140 displays and outputs the boundary line for the frame-by-frame image by using the judgment values for all regions extracted by the extraction unit 130 .

In detail, the boundary line output unit 140 outputs the boundary line of the image in gray scale by using the determination value of each area extracted as 0 or 1 by the extraction unit 130, but the area corresponding to the determination value of 1 is the boundary line. It is preferable to display and output as .

Hereinafter, a method for detecting a boundary line of a road image according to an embodiment of the present invention will be described with reference to FIGS. 3 and 4 .

3 is a flowchart illustrating an operation flow of a method for detecting a boundary line of a road image according to an embodiment of the present invention. A detailed operation of the present invention will be described with reference to this.

According to an embodiment of the present invention, first, the pre-processing unit 110 converts each frame-by-frame image extracted from the road image into a gray-scale image and pre-processes ( S310 ).

Next, the preprocessor 110 pre-processes the image converted to gray scale in step S310 by further applying a Gaussian filter (S320).

Next, the partition unit 120 partitions the frame-by-frame image preprocessed in steps S310 and S320 into a plurality of regions having the same size (S330).

Next, the extraction unit 130 extracts a determination value for the central region of the window using a window of a preset size (S340).

In this case, the window is set to have a size of an n x n area, and it is preferable to set the window to have a size of a square matrix in which n is an odd number.

Also, in step S340, a judgment value for the center region of the window is extracted using Equations 1 to 3.

Then, the extraction unit 130 extracts judgment values for all regions of the image in units of frames while shifting the window within the image in units of frames ( S350 ).

Finally, the boundary line output unit 140 displays and outputs the boundary line for the image in units of frames by using the determination values for all regions extracted in step S350 ( S360 ).

In detail, the boundary line of the frame unit image is output in gray scale using the determination value of each area extracted as 0 or 1, and the area corresponding to the determination value 1 is displayed as the boundary line and output.

4 is a diagram for comparing the performance of a conventional boundary line detection method and a detection method according to an embodiment of the present invention.

4 (a) is an example image obtained by extracting a road image as a frame-by-frame image, (b) is an example of detecting a boundary line on the (a) image using the Canny Edge Detection Method, (c) is an example of detecting a boundary line for the image (a) using the boundary line detection technique according to an embodiment of the present invention.

As can be seen by comparing the (b) image and the (c) image of FIG. 4, it can be seen that the boundary line between the (b) image and the (c) image is clearer and more similar to the original image (a).

In addition, in terms of the boundary line detection speed, the Canny boundary line detection method took about 55 ms, and the boundary line detection method according to the embodiment of the present invention took about 20 ms, indicating that the detection speed was about 2 times faster than that of the Canny boundary line detection method. can

As described above, the apparatus and method for detecting a boundary line of a road image according to an embodiment of the present invention minimizes a preprocessing process when detecting a boundary line for extracting feature points of a road image on an embedded system, thereby reducing the amount of calculation and the detection speed has the effect of speeding up

In addition, according to an embodiment of the present invention, it is possible to design a unit boundary detection algorithm, so that the weight of the embedded system is possible, and it can be used in various application fields because it can be applied to images in the field of machine learning such as detection of specific objects and stylistic recognition. have.

Although the present invention has been described with reference to the embodiment shown in the drawings, which is only exemplary, those of ordinary skill in the art will understand that various modifications and equivalent other embodiments are possible therefrom. will be. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the following claims.

100: boundary line detection device 110: pre-processing unit
120: compartment 130: extraction unit
140: border line output unit

Claims (10)

In an apparatus for detecting a boundary line from a road image,
a pre-processing unit for pre-processing each of the frame-by-frame images extracted from the road image into gray scale images;
a partitioning unit partitioning the pre-processed image into a plurality of regions having the same size;
An extraction unit extracting judgment values for the central region of the window using a window of a preset size, and extracting judgment values for all regions of the image while shifting the window in the image ; and
and a boundary line output unit for displaying and outputting a boundary line for the image by using the determination values for all the extracted regions. According to claim 1,
The window is
A boundary line detection device configured to have a size of an nxn region, and set to a size of a square matrix in which n is an odd number. According to claim 1,
The extraction unit,
A boundary line detection apparatus for extracting a judgment value for the central region of the window using the following equation:

Here, P _avg is the average value of the pixel values of each area included in the window, Pi is the pixel value of the i-th area, n ² is the size of the window, W is the weight value according to the window, and P _c is the center area of the window. A pixel value, k is a fixed constant, and p' is a judgment value for the center region of the window. 4. The method of claim 3,
The boundary line output unit,
A boundary line detection apparatus for outputting the boundary line of the image in gray scale using the determination value of each area extracted as 0 or 1, but displaying the area corresponding to the determination value 1 as the boundary line and outputting the boundary line. According to claim 1,
The preprocessor is
A boundary line detection apparatus for pre-processing by applying a Gaussian filter to the grayscale-converted image. In the boundary line detection method performed by the boundary line detection apparatus of a road image,
converting and pre-processing each frame-by-frame image extracted from the road image into a gray scale image;
partitioning the preprocessed image into a plurality of regions having the same size;
extracting judgment values for the central region of the window using a window of a preset size, and extracting judgment values for all regions of the image while shifting the window within the image; and
and displaying and outputting a boundary line for the image by using the judgment values for all the extracted regions. 7. The method of claim 6,
The window is
A boundary line detection method that is set to have a size of an nxn region, but is set to a square matrix size where n is an odd number. 7. The method of claim 6,
The step of extracting the judgment value is
A boundary line detection method for extracting a judgment value for the central region of the window using the following equation:

Here, P _avg is the average value of the pixel values of each area included in the window, Pi is the pixel value of the i-th area, n ² is the size of the window, W is the weight value according to the window, and P _c is the center area of the window. A pixel value, k is a fixed constant, and p' is a judgment value for the center region of the window. 9. The method of claim 8,
The step of displaying and outputting the boundary line comprises:
A boundary line detection method of outputting the boundary line of the image in gray scale using the determination value of each area extracted as 0 or 1, but displaying the area corresponding to the determination value 1 as the boundary line and outputting the boundary line. 7. The method of claim 6,
The pre-processing step is
A boundary line detection method of pre-processing by applying a Gaussian filter to the grayscale-converted image.

Images