KR20180093418A - Apparatus and method for detecting pedestrian - Google Patents

Abstract

An apparatus and a method for detecting a pedestrian are provided. A pedestrian detecting apparatus according to an embodiment of the present invention includes a first image obtained by photographing an object area through a first camera and a second image obtained by photographing the object area through a second camera, ; A classifying unit for classifying a time zone in which the target area is photographed as a day or a night; And a detector for assigning different weights to the first camera and the second camera according to the classified time zone, and detecting the pedestrian from the first image and the second image using the weight.

Description

TECHNICAL FIELD [0001] The present invention relates to a pedestrian detection apparatus,

Embodiments of the present invention relate to techniques for detecting pedestrians in an image.

Recently, the development and research of intelligent automobiles capable of autonomous driving have been increasing. Especially, automatic emergency braking (AEBS), which is one of Advanced Driving Assistance System (ADAS) System) has been growing.

AEBS is a technology to prevent collision in the front of the car. It aims at detecting pedestrians in front of the vehicle accurately and quickly in various environments. In order to accurately detect the pedestrian, it is necessary to extract and collect characteristics of the contrast between the pedestrian and the environment in the image. Generally, a CCD (Charge-Coupled Device) camera capable of viewing visible light in a bright day and an infrared (IR) camera capable of viewing infrared rays emitted from a pedestrian in a dark nighttime are used to detect a pedestrian do. However, there is a limitation in accurately detecting the pedestrian using only a single camera because the day / night environment changes. In other words, in the daytime environment, it is difficult to detect the pedestrian by thermal image because the temperature distribution of the surrounding environment and the pedestrian is not clearly discriminated by the sunlight, and it is difficult to detect the pedestrian by the visible image because the background is not clearly distinguished from the pedestrian in the nighttime environment have.

Korean Patent Registration No. 10-1455748 (Apr. 22, 2014)

Embodiments of the present invention are to provide a pedestrian detection technique considering external environmental factors by fusing images photographed from different types of cameras.

According to an exemplary embodiment of the present invention, there is provided an image processing apparatus comprising: a matching unit for matching a first image obtained by photographing an object area through a first camera and a second image obtained by photographing the object area through a second camera; A classifying unit for classifying a time zone in which the target area is photographed as a day or a night; And a detection unit that gives different weights to the first camera and the second camera, respectively, according to the classified time zone, and detects a pedestrian from the first image and the second image using the weight, Device is provided.

The first camera may be a CCD (Charge-Coupled Device) camera, and the second camera may be an infrared (IR) camera.

The classifying unit may classify the time zone in which the target region was photographed as day or night using the brightness value of the pixels included in the first image.

Wherein the classifying unit classifies the time zone as a day when the average brightness value of pixels positioned within a predetermined distance from the uppermost end of the first image exceeds a threshold value and classifies the time zone as night when the average brightness value is less than the threshold value can do.

Wherein the detection unit assigns a higher weight value to the first camera than the second camera when the classified time period is a week, and when the classified time zone is at night, assigns a higher weight value to the second camera Can be given.

Wherein the detection unit detects pedestrian candidates from the first image and the second image, extracts a pedestrian detection score for the detected pedestrian candidate, and calculates a damper-Schaefer theory based on the pedestrian detection score and the weight DST: Dempster-Shafer Theory) can be applied to detect the pedestrian.

Wherein the detection unit obtains a final pedestrian detection score by applying the weight according to the time zone classified into the pedestrian detection score when the DST is applied and detects the pedestrian by comparing the final pedestrian detection score and the set value have.

According to another exemplary embodiment of the present invention, a vehicle is provided, which is provided with the above-described pedestrian detecting device.

According to another exemplary embodiment of the present invention, in the matching unit, a first image obtained by photographing an object area through a first camera and a second image obtained by photographing the object area through a second camera are registered step; Classifying the time zone in which the object area is photographed into day or night, in the classifying unit; Assigning different weights to the first camera and the second camera, respectively, according to the classified time zone; And detecting the pedestrian in the first image and the second image using the weight, in the detecting unit.

The first camera may be a CCD camera, and the second camera may be an infrared camera.

The step of classifying the time period into day or night may classify the time zone in which the object area was photographed by day or night using the brightness value of the pixels included in the first image.

Classifying the time zone into day or night may include classifying the time zone as a day when the average brightness value of pixels located within a predetermined distance from the uppermost end of the first image exceeds a threshold value, , The time zone can be classified as night.

Wherein the weighting step assigns a higher weight value to the first camera than the second camera when the classified time period is a week, and when the classified time period is at night, Can be given a higher weight.

The step of detecting the pedestrian may include detecting pedestrian candidates from the first image and the second image, respectively, and extracting a pedestrian detection score for the detected pedestrian candidate; And detecting the pedestrian by applying the damper-Schaefer theory based on the pedestrian detection score and the weight value.

Wherein the step of detecting the pedestrian by applying the DST includes: obtaining a final pedestrian detection score by applying the weight according to the time zone classified into the pedestrian detection score when applying the DST, So that the pedestrian can be detected.

According to another exemplary embodiment of the present invention, there is provided an image processing apparatus, which is combined with hardware to generate a first image obtained by photographing an object area through a first camera and a second image obtained by photographing the object area through a second camera, Matching the images; Classifying the time zone in which the object area is photographed into day or night, in the classifying unit; Assigning different weights to the first camera and the second camera, respectively, according to the classified time zone; And a computer program stored in a computer-readable recording medium for causing the detecting unit to detect a pedestrian in the first image and the second image using the weight.

According to embodiments of the present invention, a pedestrian can be accurately detected by applying DST to an image photographed from cameras of different kinds. Particularly, according to the embodiments of the present invention, when the time zone in which the target area is photographed is classified into day or night, and different weights are assigned to each camera according to the time zone classified when the DST is applied, The pedestrian can be detected more accurately.

1 is a block diagram showing a detailed configuration of a pedestrian detecting apparatus according to an embodiment of the present invention;
2 is a view for explaining a process of matching a first image and a second image in a matching unit according to an embodiment of the present invention;
FIG. 3 is a view for explaining a process of classifying a time zone in which a target region is photographed by a classifying unit according to an embodiment of the present invention into a day or a night
4 is a view for explaining a DST according to an embodiment of the present invention;
5 is a diagram illustrating a process of detecting a pedestrian by applying DST in a detection unit according to an embodiment of the present invention.
6 is a diagram illustrating a process of detecting a pedestrian by applying DST in a detection unit according to an embodiment of the present invention.
7 is a diagram illustrating a process of detecting a pedestrian by applying DST in a detection unit according to an embodiment of the present invention.
8 is a graph showing the results of pedestrian detection according to an embodiment of the present invention.
9 is a view showing an example of a vehicle equipped with a pedestrian detecting device according to an embodiment of the present invention
10 is a block diagram illustrating and illustrating a computing environment including a computing device suitable for use in the exemplary embodiments.
11 is a flowchart for explaining a pedestrian detection method according to an embodiment of the present invention.

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. The following detailed description is provided to provide a comprehensive understanding of the methods, apparatus, and / or systems described herein. However, this is merely an example and the present invention is not limited thereto.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intention or custom of the user, the operator, and the like. Therefore, the definition should be based on the contents throughout this specification. The terms used in the detailed description are intended only to describe embodiments of the invention and should in no way be limiting. Unless specifically stated otherwise, the singular form of a term includes plural forms of meaning. In this description, the expressions "comprising" or "comprising" are intended to indicate certain features, numbers, steps, operations, elements, parts or combinations thereof, Should not be construed to preclude the presence or possibility of other features, numbers, steps, operations, elements, portions or combinations thereof.

1 is a block diagram showing a detailed configuration of a pedestrian detecting apparatus 100 according to an embodiment of the present invention. 1, a pedestrian detecting apparatus 100 according to an embodiment of the present invention includes an image obtaining unit 102, a matching unit 104, a classifying unit 106, and a detecting unit 108. As shown in FIG.

The image obtaining unit 102 obtains the first image and the second image with respect to the target area. Here, the target area may be an area to be photographed, for example, a front area of the vehicle, an area to be monitored on the road, or the like. The image acquiring unit 102 acquires a first image by photographing an object area through a first camera (not shown), and acquires a second image by photographing the object area through a second camera (not shown) have. In this case, the first camera may be a CCD (Charge-Coupled Device) camera, for example, as a camera suitable for daytime environment (or daylight time). In addition, the second camera may be an infrared (IR) camera, for example, as a camera suitable for a night environment (or night time zone). In case of CCD camera, it is effective to detect the pedestrian in the daytime environment because it can see the visible light. Infrared camera is effective to detect the pedestrian in the nighttime environment because infrared can be seen. However, the first camera and the second camera are suited only to cameras in daytime and nighttime environments, respectively, and the types of the first camera and the second camera are not limited thereto. Meanwhile, the first camera and the second camera may be fixed in close contact with each other, for example, in a slide bar type frame (not shown). In this case, the interval between the first camera and the second camera is minimized, so that substantially the same image can be obtained from the first camera and the second camera.

The matching unit 104 matches the first image acquired by the first camera and the second image acquired by the second camera. Even if the first camera and the second camera photograph the same area at the same time, a coordinate error between the same target of two images may be generated due to the difference in characteristics (or difference in resolution) of the camera lens. Accordingly, the matching unit 104 may perform an image matching operation for matching the subject in the first image and the subject in the second image by adjusting the coordinates of the two-dimensional first image and the second image.

2 is a diagram illustrating a process of matching a first image and a second image in the matching unit 104 according to an embodiment of the present invention. FIG. 2 (a) is a view showing a phenomenon in which an intra-target coordinate error occurs due to a difference in characteristics between the first camera and the second camera, and FIG. 2 (b) And the coordinates in the image are adjusted.

As an example, the matching unit 104 may model the relationship between the first image and the second image by homography, and adjust the coordinates in the first image or the second image through projective transformation. However, the method of matching the first image and the second image by the matching unit 104 is not limited thereto, and the matching unit 104 may be configured to perform various methods based on mutual information, cross-correlation, The first image and the second image can be matched using the image matching technique.

Returning back to Fig. 1, the classifying unit 106 classifies the time zone in which the target area was photographed as day or night. Here, the day may be a day time zone, for example, a time zone (for example, from 6:00 am to 7:00 pm) before sunrise and sunset. In addition, the nighttime may be a night time zone, for example, a time zone from sunrise until the sunrise (for example, from 7:00 pm to 6:00 am).

FIG. 3 is a view for explaining a process of classifying a time zone in which a target region is photographed by the classifying unit 106 according to an embodiment of the present invention, as a day or a night. 3 (a) and 3 (b) show a first image.

As an example, the classifying unit 106 may classify the time zone in which the target area was photographed as a day or a night using the first image. At this time, the classifying unit 106 may use the brightness value of the pixels included in the first image in the process of classifying the time zone in which the target area is photographed. More specifically, the classifying unit 106 calculates an average brightness value of pixels (for example, 50 pixels shown in FIG. 3) positioned within a predetermined distance from the uppermost end of the first image, When the threshold value is exceeded, the time zone in which the target area is photographed is classified into day, and when the average brightness value is less than the threshold value, the time zone in which the target area is photographed can be classified as night. In general, the top of the image is highly likely to include the sky in which the difference in the brightness value of the pixel appears clearly according to the day / night environment. Therefore, the classifying unit 106 classifies the average brightness of the pixels positioned within the predetermined distance from the top of the first image Value can be used to classify the time zone in which the target area was photographed as day or night. However, the method for classifying the time zone by the classifying unit 106 is not limited thereto. For example, the classifying unit 106 may classify the time zone into day or night using an average brightness value of all the pixels included in the first image.

1, the detection unit 108 assigns different weights to the first camera and the second camera (or the first image and the second image) according to the classified time zone, and uses the weight to generate the first image And a pedestrian is detected in the second image.

First, the detection unit 108 assigns a higher weight to the first camera than the second camera when the classified time period is a week, and gives a higher weight to the second camera than the first camera when the classified time period is at night . As an example, the detection unit 108 may assign weights 0.8 and 0.2 to the first camera and the second camera, respectively, if the classified time period is the week, and assign a weight value of 0.2 and 0.2 to the first camera and the second camera, And 0.8, respectively.

Next, the detection unit 108 detects a pedestrian candidate from the matched first and second images, extracts a pedestrian detection score for the detected pedestrian candidate, and calculates a pedestrian detection score based on the pedestrian detection score and the weight, (DST: Dempster-Shafer Theory) can be applied to detect pedestrians.

In one example, the detecting unit 108 detects a pedestrian candidate in the first image and the second image by comparing the subject included in the matched first image and the second image with a pre-stored pattern, The pedestrian detection score for the detected pedestrian candidate can be extracted. The pedestrian detection score is a value obtained by digitizing the probability that the detected pedestrian candidate corresponds to an actual pedestrian, and may have a value of 0 to 1, for example. In the above example, the pedestrian detection score may be increased as the degree of similarity between the detected pedestrian candidate and a previously stored pattern (i.e., a pedestrian pattern) is higher. However, the method for detecting the pedestrian candidate and extracting the pedestrian detection score is not limited to this, and the detecting unit 108 may be configured to detect the pedestrian candidate and the pedestrian candidate from the first image and the second image using various image analysis techniques, The pedestrian detection score for the pedestrian candidate can be extracted.

Thereafter, the detection unit 108 can detect the pedestrian by applying DST based on the pedestrian detection score and the weight. DST is a theory developed by Shafer by Dempster, where the degree of certainty is expressed in terms of intervals and can also be expressed numerically in terms of ambiguous measurements. First, a general DST will be described in detail with reference to FIG.

을 멱집합(Power Set)이라 하고 Θ로 기재한다(수학식 1). 상기 멱집합의 각 원소는 하나의 명제에 대한 대답으로 해석되며 이를 분별 프레임(FOD : Frame of Discernment)라 칭한다.4 is a diagram for explaining a DST according to an embodiment of the present invention. According to DST, a set of mutually exclusive hypotheses about the object of interest is established, such as probability theory first. A set of all subsets of the set

Is called a power set and is described as? (Equation 1). Each element of the power set is interpreted as an answer to a single proposition, which is referred to as a frame of discernment (FOD).

[Equation 1]

Here, one evidence may affect the degree of belief in a particular element of the power set, and the degree of belief may be expressed as a value between 0 and 1. In DST, a function mapped to [0, 1] is called a Basic Probability Assignment (BPA) and expressed as m. The BPA can be defined as: < EMI ID = 2.0 > Here, a subset of a power set with m greater than 0 is called a focal element.

&Quot; (2) "

Referring to FIG. 4, in the DST, the reliability measure for the element of interest is expressed as a section using two variables, which is called an evidential interval. The first variable Bel (Belief) is a numerical representation of the degree of belief in a hypothesis by the given evidence, and the second variable PI (Plausibility) is the degree of plausibility that the hypothesis is not negated. In DST, the above two variables can be used to express the degree of 'assurance', 'distrust', and 'uncertainty' of the hypothesis. Table 1 below shows the meanings according to the range of the above-mentioned evidence section.

Evidence section meaning [1, 1] A sure truth [0, 0] A definite lie [0, 1] Unknown 0 < PI < Bel < 1 A positive trend 0 <Bel <PI <1 A tendency to deny 0 < Bel < PI &lt; 1 Positive and negative exist simultaneously

When the BPA is assigned to the element of interest, the values of the two variables can be obtained using the combination rule according to DST. The combination rule according to the DST is defined as Equation (3) below.

&Quot; (3) &quot;

Hereinafter, a process of detecting the pedestrian using the DST by the detection unit 108 will be described.

In the present embodiments, the power set element

? = {ped}, {background}, {ped, background}, unknown

Lt; / RTI &gt; Here, ped represents a pedestrian, background (or bg) represents a background, and unknown represents an unknown. That is, θ may be any one of a 'pedestrian area', a 'surrounding background area', a 'area where the pedestrian and surrounding background coexist' and an 'unknown area'. In this embodiment, (Or pedestrian zone).

As described above, the detection unit 108 detects pedestrian candidates from the first image obtained from the first camera and the second image obtained from the second camera, respectively, and extracts pedestrian detection scores for the detected pedestrian candidates have. Further, the detection unit 108 performs a normalization operation on the pedestrian detection score, and accordingly, the reliability of the pedestrian detection score can be assigned in the range of [0, 1]. Also, the detection unit 108 can assign reliability to each of the background detection score and the score for the unknown area. The reliability assigned in this way can be shown in Table 2 below.

Here, _mccd represents the first camera and m _ir represents the second camera. Also, ped_score represents the reliability of the pedestrian detection score, and bg_score represents the reliability of the background detection score. As an example, the pedestrian detection score may be higher as the degree of similarity between the detected candidate pedestrian and a previously stored pattern (i.e., pedestrian pattern) is higher, and the similarity between the detected background candidate and a previously stored pattern (i.e., background pattern) The background detection score can be increased. The background candidate may be detected through the same method as the method of detecting the pedestrian candidate or various image analysis techniques. In addition, the reliability of the pedestrian detection score and the reliability of the background detection score may be proportional to the pedestrian detection score and the background detection score, respectively. For example, ped_score corresponding to m _ccd indicates the reliability of the pedestrian detection points extracted from the first image, and to bg_score corresponding to m _ccd means confidence in the background detection points extracted from the first image. Also, ped and bg mean the reliability of the pedestrian candidate score or the background candidate score in the area where the pedestrian candidate and the background candidate exist at the same time, and unknown means the reliability of the unknown area (uncertain area). The user can arbitrarily set the reliability in consideration of the pedestrian detection score, the background detection score, and the like. The internal values in Table 2 above can be calculated by multiplying the respective reliabilities corresponding to the horizontal and vertical regions.

Thereafter, the detection unit 108 can obtain the final pedestrian detection score by applying the combination rule according to the DST. In the above Table 2, when the above-mentioned Equation 3 is applied,

Pedestrian detection score = {ped (1) + ped (2) + ped (3)} / {1- (unknown (1)

Lt; / RTI &gt;

This can be expressed by the following equation (4).

&Quot; (4) &quot;

That is, the final pedestrian detection score for the pedestrian candidate A is calculated by applying the DST combination rule to the reliability of the pedestrian detection score extracted from the first camera and the reliability of the pedestrian detection score extracted from the second camera with respect to the pedestrian candidate A .

At this time, the detection unit 108 may assign a weight according to the time zone to the pedestrian detection score (or the reliability according to the pedestrian detection score) when the combination rule according to the DST is applied. For example, if the classification zone is daytime, the detector 108 may be given a weight of 0.8 to obtain the final pedestrian detection score to assign a weight from 0.2 to ped_score corresponding to m _ir the ped_score corresponding to m _ccd have. In this case, the probability that the candidate of the pedestrian detected in the first image is finally determined as a pedestrian increases, and the probability that the pedestrian candidate detected only in the second image is finally determined as a background rather than a pedestrian increases. The weight may be a camera confidence, for example, a value between 0 and 1. In addition, the weight may vary depending on the classified day or night.

If the pedestrian candidate is detected from only one camera, the detection unit 108 multiplies the reliability of the pedestrian detection score for the pedestrian candidate by the weight (or camera reliability) of the camera detected by the pedestrian candidate, Can be obtained.

This can be expressed by the following equation (4).

&Quot; (5) &quot;

Here, Ped represents the final pedestrian detection score, m _score represents the reliability of the pedestrian detection score, and m _camera represents the camera weight (or camera confidence), respectively.

Further, when the above-mentioned Equation 3 is applied in the same manner,

The final background detection score = {bg (1) + bg (2) + bg (3)} / {1- (unknown (1)

Lt; / RTI &gt;

Finally, the detection unit 108 can detect the pedestrian by comparing the final pedestrian detection score with the set value. Here, the set value may be the final background detection score, but is not limited thereto. For example, the set value may be a predetermined numerical value. If the set value is the final background detection score, the detection unit 108 finally determines the pedestrian candidate as a pedestrian when the final pedestrian detection score is larger than the set value, and if the final pedestrian detection score is smaller than the set value, As a background.

5 to 7 illustrate a process of detecting a pedestrian by applying DST in the detecting unit 108 according to an embodiment of the present invention. 5 to 7 (b) show the second image, and Figs. 5 to 7 (c) show the first image and the second image, respectively. Represents a fused image (i.e., matched image) of the image. 5 to 7 (a) and 7 (b) represent pedestrian detection scores for each of the pedestrian candidates, and the numerical values shown in Figs. 5 to 7 (c) ) Pedestrian candidate's final pedestrian detection score.

5, the classifying unit 106 classifies the time zone in which the target region was photographed based on the brightness values of the pixels included in the first image as a day, and the detecting unit 108 classifies the time And assigned a higher weight. Accordingly, it can be confirmed that the pedestrian candidate detected in the first image is finally detected as a pedestrian in the fused image.

6, the classifying unit 106 classifies the time zone in which the target region was photographed from the brightness values of the pixels included in the first image as a day, and the detecting unit 108 classifies the time And assigned a higher weight. Accordingly, the pedestrian candidate detected in the first image is finally detected as a pedestrian in the fusion image, and the pedestrian candidate detected in the second image is detected as the background (i.e., not detected as a pedestrian) in the fused image Can be confirmed.

7, the classifying unit 106 classifies the time zone in which the object region was photographed from the brightness values of the pixels included in the first image into the night, and the detecting unit 108 classifies the time And assigned a higher weight. Accordingly, the pedestrian candidate detected in the second image is finally detected as a pedestrian in the fused image, and the pedestrian candidate detected in the first image is detected as the background in the fused image (i.e., not detected as a pedestrian) Can be confirmed.

8 is a graph showing a result of detecting a pedestrian according to an embodiment of the present invention. 8 shows the pedestrian erroneous detection rate in the case of using only the CCD camera, the red dotted line in FIG. 8 shows the pedestrian erroneous detection rate in the case of using only the infrared camera, and the purple dotted line in FIG. And shows the pedestrian erroneous detection rate in the case of applying the DST to the fusion image according to the embodiment.

As shown in FIG. 8, when the single camera alone is used, the detection rate of the pedestrian is relatively high, whereas when the DST is applied to the convergence image according to the embodiment of the present invention, it is confirmed that the pedestrian misdetection rate is relatively low . In other words, according to the embodiments of the present invention, it is possible to accurately detect the pedestrian by applying the DST to an image photographed from cameras of different kinds. Particularly, according to the embodiments of the present invention, when the time zone in which the target area is photographed is classified into day or night, and different weights are assigned to each camera according to the time zone classified when the DST is applied, The pedestrian can be detected more accurately.

9 is an illustration of a vehicle 200 equipped with a pedestrian detection device 100 according to an embodiment of the present invention.

Referring to Fig. 9, the pedestrian detection device 100 may be installed on the roof of the vehicle 200, for example. In this case, the pedestrian detection device 100 can be used as an application for quickly and accurately detecting a pedestrian in front of the vehicle 200 and preventing a collision accident in front of the vehicle. At this time, the first camera and the second camera may be provided in the pedestrian detection apparatus 100, and may be fixed in close contact with each other, for example, in a slide bar type frame (not shown). In this case, the interval between the first camera and the second camera is minimized, so that substantially the same image can be obtained from the first camera and the second camera. The pedestrian detection device 100 may be used not only for the vehicle 200 but also for monitoring a pedestrian by providing a street light (not shown), a building (not shown), a building nearby, or a building internal structure . Furthermore, the pedestrian detecting device 100 may be disposed around military facilities and security facilities, and may be used for security, security, and the like.

10 is a block diagram illustrating and illustrating a computing environment 10 including a computing device suitable for use in the exemplary embodiments. In the illustrated embodiment, each of the components may have different functions and capabilities than those described below, and may include additional components in addition to those not described below.

The illustrated computing environment 10 includes a computing device 12. In one embodiment, computing device 12 may be one or more components included in pedestrian detection device 100, or pedestrian detection device 100.

The computing device 12 includes at least one processor 14, a computer readable storage medium 16, The processor 14 may cause the computing device 12 to operate in accordance with the exemplary embodiment discussed above. For example, processor 14 may execute one or more programs stored on computer readable storage medium 16. The one or more programs may include one or more computer-executable instructions, which when executed by the processor 14 cause the computing device 12 to perform operations in accordance with the illustrative embodiment .

The computer-readable storage medium 16 is configured to store computer-executable instructions or program code, program data, and / or other suitable forms of information. The program 20 stored in the computer-readable storage medium 16 includes a set of instructions executable by the processor 14. In one embodiment, the computer-readable storage medium 16 may be any type of storage medium such as a memory (volatile memory such as random access memory, non-volatile memory, or any suitable combination thereof), one or more magnetic disk storage devices, Memory devices, or any other form of storage medium that can be accessed by the computing device 12 and store the desired information, or any suitable combination thereof.

Communication bus 18 interconnects various other components of computing device 12, including processor 14, computer readable storage medium 16.

The computing device 12 may also include one or more input / output interfaces 22 and one or more network communication interfaces 26 that provide an interface for one or more input / output devices 24. The input / output interface 22 and the network communication interface 26 are connected to the communication bus 18. The input / output device 24 may be connected to other components of the computing device 12 via the input / output interface 22. The exemplary input and output device 24 may be any type of device, such as a pointing device (such as a mouse or trackpad), a keyboard, a touch input device (such as a touch pad or touch screen), a voice or sound input device, An input device, and / or an output device such as a display device, a printer, a speaker, and / or a network card. The exemplary input and output device 24 may be included within the computing device 12 as a component of the computing device 12 and may be coupled to the computing device 102 as a separate device distinct from the computing device 12 It is possible.

11 is a flowchart illustrating a pedestrian detection method according to an embodiment of the present invention. The method shown in Fig. 11 can be performed, for example, by the pedestrian detecting apparatus 100 described above. In the illustrated flow chart, the method is described as being divided into a plurality of steps, but at least some of the steps may be performed in reverse order, combined with other steps, performed together, omitted, divided into detailed steps, One or more steps may be added and performed.

First, the image acquiring unit 102 acquires the first image and the second image (S102). The image acquiring unit 102 acquires the first image by photographing the target area through the first camera, and acquires the second image by photographing the target area through the second camera.

Next, the matching unit 104 matches the first image and the second image (S104). The matching unit 104 may perform an image matching operation of matching the subject in the first image and the subject in the second image by adjusting the coordinates of the two-dimensional first and second images.

Next, the classifying unit 106 classifies the time zone in which the target area is photographed as day or night (S106). As an example, the classifying unit 106 may classify the time zone in which the target region was photographed as day or night using the brightness value of pixels included in the first image.

Next, the detection unit 108 assigns different weights to the first camera and the second camera (or the first image and the second image) according to the classified time zone (S108). As an example, the detection unit 108 may assign weights 0.8 and 0.2 to the first camera and the second camera, respectively, if the classified time period is the week, and assign a weight value of 0.2 and 0.2 to the first camera and the second camera, And 0.8, respectively.

Finally, the detection unit 108 detects the pedestrian from the first image and the second image using the weight (S110). As described above, the detection unit 108 detects pedestrian candidates from the first image and the second image, extracts a pedestrian detection score for the detected pedestrian candidates, and calculates, based on the pedestrian detection score and the weight, - The pedestrian can be detected by applying the Schaefer theory (DST).

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, I will understand. Therefore, the scope of the present invention should not be limited to the above-described embodiments, but should be determined by equivalents to the appended claims, as well as the appended claims.

10: Computing environment
12: computing device
14: Processor
16: Computer readable storage medium
18: Communication bus
20: Program
22: I / O interface
24: input / output device
26: Network communication interface
100: Pedestrian detection device
102:
104:
106:
108:
200: vehicle

Claims (16)

A matching unit for matching a first image obtained by photographing a target area through a first camera and a second image obtained by photographing the target area through a second camera;
A classifying unit for classifying a time zone in which the target area is photographed as a day or a night; And
And a detection unit that gives different weights to the first camera and the second camera, respectively, in accordance with the classified time zone, and detects a pedestrian from the first image and the second image using the weight, .
The method according to claim 1,
The first camera is a CCD (Charge-Coupled Device) camera,
Wherein the second camera is an Infrared Ray (IR) camera.
The method according to claim 1,
Wherein the classifying unit classifies the time zone in which the target area was photographed by day or night using the brightness value of pixels included in the first image.
The method of claim 3,
Wherein the classifying unit classifies the time zone as a day when the average brightness value of pixels positioned within a predetermined distance from the uppermost end of the first image exceeds a threshold value and classifies the time zone as night when the average brightness value is less than the threshold value The pedestrian detection device.
The method according to claim 1,
Wherein the detection unit assigns a higher weight value to the first camera than the second camera when the classified time period is a week, and when the classified time zone is at night, assigns a higher weight value to the second camera To the pedestrian detection device.
The method of claim 5,
Wherein the detection unit detects pedestrian candidates from the first image and the second image, extracts a pedestrian detection score for the detected pedestrian candidate, and calculates a damper-Schaefer theory based on the pedestrian detection score and the weight DST: Dempster-Shafer Theory) to detect the pedestrian.
The method of claim 6,
Wherein the detection unit obtains a final pedestrian detection score by applying the weight according to the time zone classified into the pedestrian detection score when the DST is applied and detects the pedestrian by comparing the final pedestrian detection score and a set value, Pedestrian detection device.
A vehicle comprising a pedestrian detection device according to any one of claims 1 to 7.
Matching the first image obtained by photographing the object area through the first camera and the second image obtained by photographing the object area through the second camera in the registration unit;
Classifying the time zone in which the object area is photographed into day or night, in the classifying unit;
Assigning different weights to the first camera and the second camera, respectively, according to the classified time zone; And
And detecting the pedestrian in the first image and the second image using the weight in the detection unit.
The method of claim 9,
The first camera is a CCD camera,
Wherein the second camera is an infrared camera.
The method of claim 9,
Wherein classifying the time zone into day or night classifies the time zone in which the object zone was photographed by day or night using brightness values of pixels included in the first image.
The method of claim 11,
Classifying the time zone into day or night may include classifying the time zone as a day when the average brightness value of pixels located within a predetermined distance from the uppermost end of the first image exceeds a threshold value, , The time zone is classified as night.
The method of claim 9,
Wherein the weighting step assigns a higher weight value to the first camera than the second camera when the classified time period is a week, and when the classified time period is at night, To a higher weight.
14. The method of claim 13,
The step of detecting the pedestrian may include:
Detecting respective pedestrian candidates from the first image and the second image, and extracting a pedestrian detection score for the detected pedestrian candidate; And
And detecting the pedestrian by applying a damper-Schaefer theory based on the pedestrian detection score and the weight.
15. The method of claim 14,
Wherein the step of detecting the pedestrian by applying the DST includes: obtaining a final pedestrian detection score by applying the weight according to the time zone classified into the pedestrian detection score when applying the DST, To detect the pedestrian.
Combined with hardware
Matching the first image obtained by photographing the object area through the first camera and the second image obtained by photographing the object area through the second camera in the registration unit;
Classifying the time zone in which the object area is photographed into day or night, in the classifying unit;
Assigning different weights to the first camera and the second camera, respectively, according to the classified time zone; And
Detecting, at the detecting unit, the pedestrian in the first image and the second image using the weight value
The computer program being stored on a computer readable recording medium.

Images