KR20100113371A - Method and apparatus for guarding pedestrian using far-infra-red stereo camera - Google Patents

Abstract

PURPOSE: A pedestrian protecting method using a FIR(far-infra-red) stereo camera is provided to prevent the collision with pedestrians by accurately recognizing background and pedestrian using a FIR stereo camera. CONSTITUTION: A pedestrian protecting method using an FIR(far-infra-red) stereo camera comprises an image acquisition part(120), a pedestrian recognition part(130), a collision risk determining part(140) and an anti-collision controller(150). The pedestrian recognition part takes image using FIR stereo camera which is calibrated. The pedestrian recognition part recognizes the pedestrians using the image. The anti-collision controller controls a brake device to prevent the collision with pedestrians.

Description

Method and device for pedestrian protection using far infrared stereo camera {Method and Apparatus for Guarding Pedestrian using Far-Infra-Red Stereo Camera}

The present invention relates to a pedestrian protection method and apparatus using a far infrared stereo camera. More specifically, the present invention relates to a pedestrian protection method and apparatus using a far-infrared stereo camera capable of accurately protecting a pedestrian by accurately distinguishing a background of surrounding buildings, objects, and the like, which should be excluded in order to accurately recognize a pedestrian.

Recently, a pedestrian (including a bicycle occupant) accident takes up a large part of a traffic accident, and such a pedestrian accident is an accident that must be prevented in that it leads to a big human accident such as death once an accident occurs.

Therefore, many studies on the protection of pedestrians (including bicycle occupants) are in progress. Nevertheless, the conventional pedestrian protection device, there is still a problem that can not accurately distinguish the background of the surrounding buildings, objects, and the like to be excluded in order to correctly recognize the pedestrian, and thus, the pedestrian protection is not properly There is a situation.

SUMMARY OF THE INVENTION An object of the present invention for solving the above problems is to provide a pedestrian protection device that can accurately protect a pedestrian by accurately distinguishing a background of surrounding buildings, objects, and the like, which should be excluded in order to correctly recognize a pedestrian. There is.

The present invention for achieving the above object, the image acquisition unit for obtaining an image through a calibrated Far-Infra-Red (FIR) stereo camera; A pedestrian recognition unit recognizing a pedestrian through the acquired image; A collision risk determination unit configured to determine a risk of collision with the recognized pedestrian based on the recognized distance information of the pedestrian and the speed information of the vehicle; And a collision avoidance control unit controlling a braking device so as to avoid a collision with the recognized pedestrian when it is determined that there is a risk of collision, which provides a pedestrian protection apparatus using a far-infrared stereo camera.

In addition, the present invention, the image acquisition step of acquiring an image through a calibrated Far-Infra-Red (FIR) stereo camera; A pedestrian recognition step of recognizing a pedestrian through the acquired image; A collision risk determining step of determining a collision risk with the recognized pedestrian based on distance information of the recognized pedestrian and vehicle speed information; And a collision avoidance control step of controlling a braking device to avoid a collision with the recognized pedestrian when it is determined that there is a risk of collision, and providing a pedestrian protection method using a far-infrared stereo camera.

As described above, according to the present invention, by using a far-infrared stereo camera to accurately distinguish the background and pedestrians to recognize the pedestrians, determine the risk of collision with the recognized pedestrians, and avoid collisions with pedestrians with a collision risk. By controlling the vehicle, there is an effect of providing a pedestrian protection device that can accurately protect pedestrians.

Hereinafter, some embodiments of the present invention will be described in detail through exemplary drawings. In adding reference numerals to the components of each drawing, it should be noted that the same reference numerals are assigned to the same components as much as possible even though they are shown in different drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

In addition, in describing the component of this invention, terms, such as 1st, 2nd, A, B, (a), (b), can be used. These terms are only for distinguishing the components from other components, and the nature, order or order of the components are not limited by the terms. If a component is described as being "connected", "coupled" or "connected" to another component, that component may be directly connected to or connected to that other component, but there may be another configuration between each component. It is to be understood that the elements may be "connected", "coupled" or "connected".

1 is a block diagram of a pedestrian protection device 100 using a far-infra-red (FIR) stereo camera 110 according to an embodiment of the present invention.

1, the pedestrian protection device 100 using the FIR stereo camera 110 according to an embodiment of the present invention, the image acquisition unit for obtaining an image through a calibrated FIR stereo camera 110 120; A pedestrian recognition unit 130 for recognizing a pedestrian (including a bicycle occupant) through an image acquired through the FIR stereo camera 110; A collision risk determination unit 140 that determines a risk of collision with the recognized pedestrian based on the recognized distance information of the pedestrian and the speed information of the vehicle; And a collision avoidance control unit 150 for controlling the braking device to avoid collision with the recognized pedestrian when it is determined that there is a collision risk.

The above-described FIR stereo camera 110 includes a first FIR camera (for example, a left FIR camera) and a second far infrared camera (for example, a right FIR camera), and an image obtained through the FIR stereo camera 110 may be obtained. And a first image and a second image acquired through the first far infrared camera and the second far infrared camera, respectively.

The characteristics of the FIR stereo camera 110 will be briefly described below.

Cameras using common CCD and CMOS devices detect and project light in the visible region, so images similar to those seen by the human eye can be obtained. On the other hand, the FIR stereo camera 110 projects light in an infrared band that is not seen by a person. Infrared refers to light of 750nm to 1mm of the wavelength of light. Among these infrared bands, NIR (Near Infra Red) light refers to a wavelength of 700nm to 1400nm, and NIR band of light is invisible to human eyes, but CCD It can also be sensed by CMOS devices, and filters can only detect light in the NIR band. In contrast, FIR light is also referred to as Long Wavelength Infra Red (LWIR), and infrared light represents a band of 8 to 15 μm of light wavelengths. In particular, the FIR band has the advantage of distinguishing the temperature because the wavelength changes with temperature. Human body temperature has a wavelength of 10 μm.

The above-mentioned calibration (Calibration) refers to the relationship between the image obtained by the FIR camera 110 and the real world. That is, the relationship between the coordinates of the image projected by the FIR camera 110 and the three-dimensional actual obtained. The relationship between the camera coordinates and the real world coordinates is shown through FIG. 4 and Equation 1.

In Equation 1, (xi, yi) represents image coordinates, and (Xi, Yi, Zi) represents coordinates of the real world.

Calibration of the FIR stereo camera 110 is performed by obtaining P matrix of Equation 1. To find the P matrix, we need to know the camera coordinates and the real world coordinate values for at least six points. More accurate P matrix values can be obtained using a large number of points. More accurate values are obtained using the Least Squares method for a large number of points.

Since the FIR stereo camera 110 acquires images simultaneously from two FIR cameras (a first FIR camera and a second FIR camera), it is possible to acquire distance information that cannot be acquired using a single FIR camera. . In the FIR stereo camera 110, three-dimensional depth information may be acquired using a relationship between two FIR cameras (a first FIR camera and a second FIR camera), and the three-dimensional depth information means distance information. As illustrated in FIG. 5, the FIR stereo camera 110 may be configured with two FIR cameras including a first FIR camera 510 and a second FIR camera 520. As shown in FIG. 5, the points X 'and X' 'of the real world differ in the points projected from the images acquired by the first FIR camera 510 and the second FIR camera 520, respectively. The difference between the points projected on the respective FIR cameras 510 and 520 is different depending on the distance, and the difference between the points projected on the respective FIR cameras 510 and 520 represents three-dimensional depth information (ie, distance information).

Speaking of each FIR camera (510, 520) _L, x _"R" x each point of the projected image coordinate "point X to the real world, two-coordinate difference is represented as in equation (2).

In Equation 2, Z 'is three-dimensional depth information (distance information), f is a focal length of the FIR cameras 510 and 52, and B is a distance between the FIR cameras 510 and 52. In order to obtain 3D distance information in Equation 2, it is converted as in Equation 3, and 3D depth information (distance information) of each pixel of the image may be obtained using the converted Equation 3.

The image acquisition unit 120 for acquiring an image through the FIR stereo camera 110 having such characteristics is a FIR stereo camera 110 calibrated using a calibration pattern having a difference in heat reflectivity or heat absorption. Acquire an image through.

The calibration pattern mentioned above may be, for example, a pattern manufactured using white aluminum and black paper having a difference in heat reflectivity or heat absorption. Such a calibration pattern is exemplarily illustrated in FIG. 6, and in an image obtained by using the calibration pattern, as shown in FIG. 7, the white aluminum part does not absorb heat and appears black and is black. It can be seen that the paper part absorbs heat and becomes relatively hot and appears white.

2 is a block diagram of the pedestrian recognition unit 130 included in the pedestrian protection device 100 using the FIR stereo camera 110 according to an embodiment of the present invention.

Referring to FIG. 2, the pedestrian recognition unit 130 included in the pedestrian protection device 100 using the FIR stereo camera 110 receives an image acquired through the FIR stereo camera 110 and receives an image. A stereo matching unit 210 for obtaining 3D distance information of each pixel of the input image through stereo matching; A pedestrian candidate group extracting unit 220 for binarizing the input image and extracting a pedestrian candidate group by performing pixel distribution analysis on the binarized image by using the 3D distance information obtained above; And extracting feature data on the extracted pedestrian candidate group and determining whether the extracted pedestrian candidate group is a pedestrian based on the extracted feature data, such as a pedestrian determination unit 230 for recognizing the pedestrian.

The three-dimensional distance information obtained by the stereo matching unit 210 is 'distance information with the recognized pedestrian' used by the collision risk determination unit 140 to determine the collision risk.

The stereo matching unit 210 performs stereo matching to obtain 3D distance information from each pixel of the input image. Here, the stereo matching method includes a region based stereo matching method and a region based stereo matching method. There is a feature based stereo matching method.

Region Based Stereo Matching method finds the best matching point by comparing the regions of the image.Feature Based Stereo Matching method extracts the features of the image and compares the feature points. How to find the matching point.

In the present invention, the stereo matching unit 210 uses a region based stereo matching method as stereo matching to obtain 3D distance information from each pixel of the input image. The image acquired through the FIR stereo camera 110 detects heat in the infrared band. Therefore, since the high heat is represented by the image having a high value, the feature can be obtained already in the image acquisition process. Accordingly, the stereo matching unit 210 is a stereo matching for obtaining three-dimensional distance information from each pixel of the input image. The stereo matching unit 210 compares regions of the input image and finds a corresponding point. A Region Based Stereo Matching method for obtaining distance information is used.

At this time, the stereo matching unit 210 does not perform stereo matching on all regions, but compares the corresponding points by comparing regions which are equal to or greater than a value corresponding to a threshold temperature (for example, T = 85) among all regions of the input image. Perform a stereo match to find. Because stereo matching is required for all areas of the image, many operations are required, and since the low heat that does not exceed a certain value corresponding to the threshold temperature is excluded from the region of interest, the part that is above a certain value Is to perform stereo matching only.

The region based stereo matching method used in the stereo matching unit 210 is performed by Equation 4 below.

In Equation 4, L is an image obtained from a first FIR camera (eg, a left FIR camera) of the FIR stereo camera 110, and R is a second FIR camera (eg, a right FIR camera) of the FIR stereo camera 110. ), S denotes the similarity of each pixel in the Region Based Stereo Matching method. Referring to Equation 4, s to make the similarity S of each pixel to be the minimum means three-dimensional distance information.

In other words, the stereo matching unit 210, when stereo matching, calculates the similarity (S) of each pixel of the input image, and based on the calculated similarity (S) three-dimensional distance information (s) in each pixel Acquire it.

FIG. 8 is an image (3 (c) of FIG. 8C) obtained by performing stereo matching on an image (FIGS. 8A and 8B) input to the stereo matching unit 210. )) Is an exemplary view showing. FIG. 8A illustrates an image obtained from a first FIR camera (eg, a left FIR camera) of the FIR stereo camera 110, and FIG. 8B illustrates a second FIR camera of the FIR stereo camera 110. Example: Image obtained from the right FIR camera.

The above-described pedestrian candidate group extracting unit 220 is a module for extracting a pedestrian candidate that may be recognized as a pedestrian from the input image. Since the input image is obtained through the FIR stereo camera 110, the pedestrian may exist. The part appears in bright colors. Accordingly, the pedestrian candidate group extracting unit 220 may extract a portion that becomes a pedestrian candidate when the input image undergoes a binary process as shown in Equation 5 below.

The binarized image obtained by binarizing the image input by the pedestrian candidate group extracting unit 220 using Equation 5 may be confirmed through FIG. 9. FIG. 9A illustrates an image input to the pedestrian candidate group extracting unit 220, and FIG. 9B illustrates a binarized image obtained by binarizing FIG. 9A.

Here, the pedestrian candidate group extracting unit 220 may perform pixel distribution analysis on the number of vertical pixels and the number of horizontal pixels in order to extract the pedestrian part from the binarized image of only the bright part. As illustrated in FIG. 9, since a portion of the pedestrian has a large number of bright color pixels, a large number of pixels may be determined as a pedestrian candidate in the distribution of the number of pixels. In particular, since the pedestrians are vertically distributed, after performing the vertical pixel distribution analysis, the horizontal pixel distribution analysis may be performed to better extract the pedestrian candidate group. Pixels having a predetermined ratio (for example, 40%) of the highest value among the vertical distribution of bright color pixels may be extracted as pedestrian candidates. 10 is an image of a pedestrian candidate group obtained through vertical pixel distribution analysis of bright color pixels.

After performing the vertical pixel distribution analysis, in order to perform the horizontal pixel distribution analysis, a portion extracted as the pedestrian candidate group from the vertical pixel distribution analysis result may be used. Since the position of the pedestrian candidate group is determined through the vertical pixel distribution analysis, a more accurate pedestrian candidate group can be extracted through the horizontal pixel distribution analysis. However, unlike the vertical pixel distribution analysis, the horizontal pixel distribution analysis may be greatly influenced by the ambient noise. As shown in FIG. 10, the ambient noises (eg, surrounding buildings) are influenced because they are arranged horizontally unlike pedestrians (ie, humans), which makes it difficult to analyze the horizontal pixel distribution. In particular, in FIG. 11, which shows a result obtained by performing a horizontal pixel distribution analysis using the results of the vertical pixel distribution analysis shown in FIG. 10, due to the ambient noise (eg, surrounding buildings), it may be recognized as a pedestrian. It can be seen that a situation where a part is divided into several parts may occur.

As shown in FIG. 11, when the pedestrian candidate group is divided into various parts due to noise, there is a problem in that pedestrian recognition cannot be accurately performed. At this time, the pedestrian candidate group can be more accurately extracted using distance information.

In order to use the distance information, it is necessary to know the distance of the pedestrian image, and to know the distance of the pedestrian image, the distance average of the longest part of the horizontal pixel distribution analysis is used. Then, the distance information can be known. The height of the pedestrian can be calculated from the analyzed distance information, and the height of the pedestrian can be obtained through Equation 6 below.

In Equation 6, Y means the height of the pedestrian, Z means the distance to the pedestrian, f means the focal length, and y means the y coordinate in the image.

In addition, it is possible to calculate the point where the road surface meets the pedestrian by using the calibration information. Through this, the relationship between the y coordinate in the road surface (floor) and the image can be obtained. By using the relationship, the height of the pedestrian is calculated using the height of a common person (pedestrian) (for example, 1.8m), and using the closest part of the obtained pedestrian height value, the candidate group of pedestrians as shown in FIG. 12 can be extracted more accurately. Can be.

In other words, referring to FIG. 3, the pedestrian candidate group extracting unit 220 extracting a pedestrian candidate that may be recognized as a pedestrian from the input image, analyzes a distribution of vertical pixels of the binarized image and positions the pedestrian candidate group. A vertical pixel distribution analyzer 310 for identifying information; And a horizontal pixel distribution analyzer 320 for analyzing the distribution of the horizontal pixels of the binarized image to determine the height information of the pedestrian candidate group.

In addition, referring to FIG. 3, the pedestrian candidate group extracting unit 220 removes noise, which is height information irrelevant to the pedestrian candidate group, from the height information obtained using the obtained 3D distance information. It may include, through which it is possible to extract a more accurate pedestrian candidate group.

The pedestrian determination unit 230 included in the pedestrian recognition unit 130 of the pedestrian protection device 110 may determine whether the pedestrian is using an image of the pedestrian candidate group extracted by the pedestrian candidate group extraction unit 220. As an example for this purpose, feature data is extracted from an image corresponding to the extracted pedestrian candidate group through a Gabor Filter Bank, and extracted feature data through a support vector machine (SVM). The pedestrian is finally recognized by determining whether the pedestrian candidate group classified and extracted is a pedestrian.

The collision risk determination unit 140 of the pedestrian protection device 110 measures the risk of collision with the pedestrian based on the distance information with the pedestrian and the speed information of the vehicle, based on the pedestrian recognition result from the pedestrian recognition unit 130. If it is determined that there is a risk of collision, the collision avoidance control unit 150 determines a braking time point based on the recognized distance information with the pedestrian, the speed information of the vehicle, and the braking distance information of the vehicle, and determines the determined braking time. By controlling the braking device through the electronic control unit (ECU) at the time point, it is possible to avoid the collision with the pedestrian and to protect the pedestrian.

13 is a flowchart illustrating a pedestrian protection method using a far infrared stereo camera according to an embodiment of the present invention.

Referring to FIG. 13, a pedestrian protection method using a far-infrared stereo camera according to an embodiment of the present invention may include obtaining an image through a calibrated far-infra-red (FIR) stereo camera 110. Image acquisition step S1300; Pedestrian recognition step (S1302) for recognizing the pedestrian through the obtained image; A collision risk determination step (S1304) of determining a risk of collision with the recognized pedestrian based on distance information of the recognized pedestrian and vehicle speed information; And a collision avoidance control step (S1305) of controlling the braking device to avoid a collision with the recognized pedestrian when it is determined that there is a collision risk.

As described above, according to the present invention, the far-infrared stereo camera 110 is used to accurately distinguish the pedestrian from the background (a neighboring building and the surrounding object) and the pedestrian, and determine the risk of collision with the recognized pedestrian. By controlling the vehicle to avoid a collision with a pedestrian having a collision risk, there is an effect of providing a pedestrian protection method and a pedestrian protection device 100 capable of accurately protecting a pedestrian.

In the above description, all elements constituting the embodiments of the present invention are described as being combined or operating in combination, but the present invention is not necessarily limited to the embodiments. In other words, within the scope of the present invention, all of the components may be selectively operated in combination with one or more. In addition, although all of the components may be implemented in one independent hardware, each or all of the components may be selectively combined to perform some or all functions combined in one or a plurality of hardware. It may be implemented as a computer program having a. Codes and code segments constituting the computer program may be easily inferred by those skilled in the art. Such a computer program may be stored in a computer readable storage medium and read and executed by a computer, thereby implementing embodiments of the present invention. The storage medium of the computer program may include a magnetic recording medium, an optical recording medium, a carrier wave medium, and the like.

In addition, the terms "comprise", "comprise" or "having" described above mean that the corresponding component may be included, unless otherwise stated, and thus excludes other components. It should be construed that it may further include other components instead. All terms, including technical and scientific terms, have the same meanings as commonly understood by one of ordinary skill in the art unless otherwise defined. Terms commonly used, such as terms defined in a dictionary, should be interpreted to coincide with the contextual meaning of the related art, and shall not be construed in an ideal or excessively formal sense unless explicitly defined in the present invention.

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the scope of 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.

1 is a block diagram of a pedestrian protection device using a far-infrared stereo camera according to an embodiment of the present invention,

2 is a block diagram of a pedestrian recognition unit included in a pedestrian protection device using a far-infrared stereo camera according to an embodiment of the present invention;

3 is a block diagram of a pedestrian candidate group extraction unit in a pedestrian recognition unit included in a pedestrian protection device using a far-infrared stereo camera according to an embodiment of the present invention;

4 is a diagram illustrating a relationship between camera coordinates and actual coordinates;

5 is a view showing the configuration of a far-infrared stereo camera;

6 exemplarily shows a calibration pattern;

7 is a diagram illustrating an image acquired by a far-infrared stereo camera calibrated using a calibration pattern;

8 is a diagram illustrating an image of 3D distance information obtained through stereo matching on an input image by way of example;

9 is a diagram illustrating an image in which an input image is binarized;

10 is a diagram exemplarily illustrating a result of analyzing a vertical pixel distribution of a binarized image;

11 is a diagram exemplarily illustrating a result of analyzing horizontal pixel distribution of a binarized image;

12 is a diagram illustrating an analysis result of a horizontal pixel distribution obtained by removing noise with respect to a horizontal pixel distribution of a binarized image.

13 is a flowchart illustrating a pedestrian protection method using a far infrared stereo camera according to an embodiment of the present invention.

<Description of Symbols for Main Parts of Drawings>

100: pedestrian protection device

110: Far Infrared (FIR) Stereo Camera

120: image acquisition unit

130: pedestrian recognition unit

140: collision risk determination unit

150: collision avoidance control

210: stereo matching unit

220: pedestrian candidate group extraction unit

230: pedestrian determination unit

310: vertical pixel distribution analyzer

320: horizontal pixel distribution analyzer

330: noise removing unit

Claims (13)

An image acquisition unit that acquires an image through a calibrated Far-Infra-Red (FIR) stereo camera; A pedestrian recognition unit recognizing a pedestrian through the acquired image; A collision risk determination unit configured to determine a risk of collision with the recognized pedestrian based on the recognized distance information of the pedestrian and the speed information of the vehicle; And If it is determined that there is a risk of collision, a collision avoidance control unit controlling a braking device to avoid a collision with the recognized pedestrian. Pedestrian protection device using a far-infrared stereo camera, characterized in that it comprises a. The method of claim 1, The far infrared stereo camera may include a first far infrared camera and a second far infrared camera, and the obtained image may include a first image and a second image acquired through the first far infrared camera and the second far infrared camera, respectively. Pedestrian protection device using a far-infrared stereo camera. The method of claim 1, The image acquisition unit, A pedestrian protection apparatus using a far-infrared stereo camera, characterized in that the image is acquired through the calibrated far-infrared stereo camera using a calibration pattern having a difference in heat reflectivity or heat absorption. The method of claim 3, wherein The calibration pattern is, The pedestrian protection device using a far-infrared stereo camera, characterized in that the pattern is made of white aluminum and black paper having a difference in the heat reflectivity or the heat absorption. The method of claim 1, The pedestrian recognition unit, A stereo matching unit which receives the obtained image and acquires 3D distance information of each pixel of the input image through stereo matching with respect to the input image; A pedestrian candidate group extraction unit for binarizing the input image and extracting a pedestrian candidate group by performing pixel distribution analysis on the binarized image using the obtained 3D distance information; And A pedestrian determination unit that extracts feature data for the extracted pedestrian candidate group and determines whether the pedestrian candidate group is a pedestrian based on the extracted feature data, thereby recognizing the pedestrian. Including, The acquired 3D distance information is a pedestrian protection device using a far-infrared stereo camera, characterized in that the distance information with the recognized pedestrian. The method of claim 5, The stereo matching unit, The pedestrian protection apparatus using a far-infrared stereo camera, characterized in that to obtain the three-dimensional distance information in each pixel through the stereo matching to find the corresponding point by comparing the input image area. The method of claim 6, The stereo matching unit, The stereoscopic pedestrian protection device using a far-infrared stereo camera, characterized in that to perform the stereo matching to find the corresponding point by comparing the area that is equal to or greater than a value corresponding to a threshold temperature of all the input image. The method of claim 5, The stereo matching unit, In the stereo matching, the similarity of each pixel of the input image is calculated, the pedestrian protection device using a far-infrared stereo camera, characterized in that to obtain the three-dimensional distance information on each pixel based on the calculated similarity. . The method of claim 5, The pedestrian candidate group extraction unit, A vertical pixel distribution analyzer configured to analyze the distribution of the vertical pixels of the binarized image and determine position information of the pedestrian candidate group; And Horizontal pixel distribution analysis unit for analyzing the distribution of the horizontal pixels of the binarized image, to determine the height information of the pedestrian candidate group Pedestrian protection device using a far-infrared stereo camera, characterized in that it comprises a. The method of claim 9, The pedestrian candidate group extraction unit, And a noise removing unit configured to remove noise, which is height information irrelevant to the pedestrian candidate group, from the determined height information by using the obtained three-dimensional distance information. The method of claim 5, The pedestrian determination unit, Through the Gabor Filter Bank, the feature data is extracted from the image corresponding to the extracted pedestrian candidate group, A pedestrian protection apparatus using a far-infrared stereo camera, wherein the pedestrian is recognized by classifying the extracted feature data and determining whether the extracted pedestrian candidate group is a pedestrian through a support vector machine (SVM). . The method of claim 1, The collision avoidance control unit, When it is determined that there is a risk of collision, the braking time is determined based on the recognized distance information with respect to the pedestrian, the speed information of the vehicle, and the braking distance information of the vehicle, and the electronic control unit may be used at the determined braking time. ECU: A pedestrian protection device using a far-infrared stereo camera, characterized in that for controlling the braking device through an Electronic Cotnrol Unit. An image acquisition step of acquiring an image through a calibrated far-infra-red (FIR) stereo camera; A pedestrian recognition step of recognizing a pedestrian through the acquired image; A collision risk determination step of determining a collision risk with the recognized pedestrian based on the recognized distance information of the pedestrian and the speed information of the vehicle; And A collision avoidance control step of controlling a braking device to avoid a collision with the recognized pedestrian when it is determined that there is a risk of collision Pedestrian protection method using a far-infrared stereo camera, characterized in that it comprises a.

Images