WO2016104831A1 - 저사양 영상 기기에서의 실시간 영상 인식 방법 - Google Patents

저사양 영상 기기에서의 실시간 영상 인식 방법 Download PDF

Info

Publication number
WO2016104831A1
WO2016104831A1 PCT/KR2014/012832 KR2014012832W WO2016104831A1 WO 2016104831 A1 WO2016104831 A1 WO 2016104831A1 KR 2014012832 W KR2014012832 W KR 2014012832W WO 2016104831 A1 WO2016104831 A1 WO 2016104831A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
information
tracking
interest
recognition method
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/KR2014/012832
Other languages
English (en)
French (fr)
Inventor
윤주홍
황영배
최병호
김정호
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Korea Electronics Technology Institute
Original Assignee
Korea Electronics Technology Institute
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Korea Electronics Technology Institute filed Critical Korea Electronics Technology Institute
Publication of WO2016104831A1 publication Critical patent/WO2016104831A1/ko
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/215—Motion-based segmentation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/223—Analysis of motion using block-matching
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T9/00—Image coding
    • G06T9/20—Contour coding, e.g. using detection of edges

Definitions

  • the present invention relates to image recognition technology, and more specifically to video equipment.
  • Real-time (every frame) image recognition is required to perform functions such as motion recognition-based human interaction, privacy protection, and vehicle collision prevention.
  • the present invention has been made to solve the above problems, and the object of the present invention is to provide a low-embedded low-speed operation by detecting a detector and a tracker.
  • Another object of the present invention is to provide a real-time image recognition method that can adjust the operation plan of the detector and the tracker by using the camera information and the image information.
  • Another purpose of the present invention is the unique information in the video (skin, boundary).
  • the image recognition method includes the steps of: detecting an object of interest at a first period in an image; and tracking the object of interest at a second period in the image; Including, The first cycle, The second cycle Longer than
  • the image recognition method by referring to at least one of the characteristic information of the image and the movement information of the camera photographing the image, the first period and the second stock price
  • the setting step may further include.
  • the image recognition method may further include a step of setting an object of interest through interpolation in a frame that does not detect and track among the frames constituting the image.
  • At least one of skin information and edge information may be used to detect an object of interest including at least one of a face and a vehicle license plate.
  • the object of interest can be detected only in a region where the skin density and the edge density are higher than the reference.
  • the detecting step and the tracking step may include a camera for capturing the image.
  • the tracking area can be set based on the object motion size for each area.
  • an image recognition method includes: a detector for detecting an object of interest at a first period in an image; a tracker for tracking an upper center of gravity object at a second period in the image; and the first period. It includes; setting unit for setting longer than the second period.
  • the number of operations of the detector and the tracker is set to be equal, thereby enabling real-time image recognition even in a low specification embedded environment.
  • FIG. 1 is a view provided in the description of an imaging apparatus according to an embodiment of the present invention.
  • FIG. 2 is a diagram showing an example of an image recognition operation plan setting result
  • 3 is a diagram showing a face detection result using skin information
  • FIG. 4 is a diagram showing a result of setting a search area using skin information
  • 5 is a diagram showing a result of minimizing a search area using edge information.
  • 6 is a view showing a license plate detection result using edge information
  • FIG. 7 shows the result of setting a license plate search area using geometry information of an image scene.
  • FIG. 8 is a diagram illustrating a result of setting a search area for vehicle and pedestrian detection using geometrical structure information of an image scene
  • FIG. 9 is a diagram illustrating a similarity map obtained as a result of a sliding window search and the result
  • FIG. 10 is a diagram illustrating movement information of an object and movement information by a scene. Best Mode for Carrying Out the Invention
  • the real-time image recognition method through the protection of personal information in the black box for the vehicle (mosaic treatment of face and vehicle license plate), human interaction system, criminal vehicle recognition, pedestrian / front vehicle recognition To prevent car stratification
  • the real-time image recognition method according to the embodiment of the present invention is concerned with all the frames constituting the image by operating the detector with a relatively large amount of calculations and minimizing it through tracking and interpolation with a small amount of calculations. This is how object recognition is performed.
  • the imaging device according to an embodiment of the present invention includes an operation plan setting unit 110 and a detector. 120 and tracker 130.
  • the detector 120 captures images of the camera using an object detection algorithm.
  • the object of interest is detected from the image, and the tracker 130 tracks the object of interest detected by the detector 120 using an object tracking algorithm.
  • the operation plan setting unit 110 sets operation cycles of the detector 120 and the tracker 130.
  • the operation plan setting unit 110 may refer to the feature information (skin information, edge information, etc.) and camera information (position information, speed information, etc.) of the image. .
  • the operation plan setting unit 110 operates the detector 120 at the minimum number of times, and auxiliaryly uses the tracker 130 with a relatively small amount of calculation to speed and All detection rates are achieved.
  • the object position of interest is interpolated and filled with movement information that can be calculated from the tracking results.
  • FIG. 2 is a diagram showing an example of a video recognition operation plan setting result.
  • the detector 120 operates once every 30 frames, the tracker 130 operates once every three frames, replacing the role of the detector 10, and frames in which the tracker 130 is not operating are moved based on a motion information-based link. You can see that the object of interest is set to interpolate and populate.
  • the image feature information skin information, edge information column
  • camera information position information, velocity information column
  • the operating periods of the detector 120 and the tracker 130 can be set according to the situation.
  • the face detection function of the detector 120 is stopped. If the speed of the highway or camera (vehicle) is fast, the image blur is generated due to the high speed, and the face information is severely distorted.
  • the unique characteristic information of each object is used in the image to improve the speed of the detector 120.
  • the color information skin color (skin color) that is common to the face is used.
  • Information is used, and license plates, vehicles, and pedestrians use edge information that exists around objects.
  • the detector 120 does not search the entire region of the image, but searches only the region having the skin information and the boundary line information that the object of interest may exist, thereby greatly reducing the calculation amount and reducing the false detection rate.
  • Fig. 6 shows the result of license plate detection using edge information. It was.
  • False detection rates can be greatly reduced, while skin / edge density for each detection result can be additionally considered to further reduce false detection rates.
  • the skin / edge density is obtained by dividing the number of skin color / edge pixels by the width of the detection object in the detection object.
  • the detection result with a threshold value of less than or equal to the threshold value is treated as having low reliability and is excluded.
  • the detector 120 and the tracker 130 set up an effective search area by using the geometrical structure information of the scene that the camera faces. That is, as shown in FIGS.
  • the area in which the object is located in the image can be roughly determined.This allows for unnecessary calculations that are consumed by the search by considering only the object size that fits the area, rather than searching for objects of varying sizes.
  • the tracker 130 uses the detected result to perform a search using the external model of the object in a predetermined area as shown in FIG. 9. As shown in FIG. Sliding window) The position of the pixel with the highest similarity in the similarity map obtained after the search is traced.
  • the feature information of the object can be used together to improve the tracking performance.
  • the skin density information obtained at the time of face detection is used, and the edge density information such as license plates are used for tracking. Divides into the background of the visual field to prevent tracking failures.
  • FIG. 10 illustrates a method of improving object tracking performance using preliminary information on the assumption that objects move constantly for a short time and using motion sizes in a specific area of the image.
  • the larger the navigation range is the larger the search range is.
  • the larger the navigation range is the smaller the search range is.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)
  • Traffic Control Systems (AREA)

Abstract

저사양 영상 기기에서의 실시간 영상 인식 방법이 제공된다. 본 발명의 실 시예에 따른 영상 인식 방법은, 영상에서 제1 주기로 관심 객체를 검출하고, 영상 에서 제2 주기로 관심 객체를 추적하며, 제1 주기는 상기 제2 주기 보다 길다. 이 에 의해, 검출기와 추적기의 동작 횟수를 적응적으로 설정하여, 저사양의 임베디드 환경에서도 실시간 영상 인식이 가능해진다.

Description

명세서
발명의명칭:저사양영상기기에서의실시간영상인식방법 기술분야
[1] 본발명은영상인식기술에관한것으로，더욱상세하게는영상기기에서
촬영된영상으로부터관심객체를실시간으로검출하고추적하는영상인식 방법에관한것이다.
배경기술
[2] 동작인식기반의휴먼인터액션과개인정보보호및차량층돌방지둥과같은 기능을수행하기위해서는실시간 (매프레임마다)영상인식이필수적으로 요구된다.
[3] 기존의영상인식알고리즘은 PC와같은고사양기기에서는좋은성능을
보이지만，계산량이많아휴대폰，태블릿 PC,차량용블랙박스등과같은 임베디드환경에서는실시간으로동작하는것이매우어렵다.
[4] 이에，저사양임베디드플랫폼이탑재된차량용블랙박스등에서는，실제주행 환경에서객체 (보행자,얼굴,번호판,차량)들의오검출률이비교적높은편이다.
[5] 이는,영상인식을전제로하는모자이크처리기능이나차량층돌방지기능을 저하시킬수있어,더욱문제된다.이에，차량용블랙박스와같은저사양 영상기기에서실시간으로안정된성능의영상인식을가능하도록하기위한 방안의모색이요청된다.
발명의상세한설명
기술적과제
[6] 본발명은상기와같은문제점을해결하기위하여안출된것으로서，본발명의 목적은，검출기와추적기를적웅적으로동작시켜，저사양의임베디드
환경에서도실시간영상인식이가능하도록함에있다.
[7] 또한，본발명의다른목적은，카메라정보와영상정보를이용하여,검출기와 추적기의동작계획을적웅적으로조정할수있는실시간영상인식방법을 제공함에있다.
[8] 아울러，본발명의또다른목적은，영상내의고유정보 (스킨，경계선둥)，
카메라가바라보는장면의기히학적구조정보및영상의구조적특징으로인한 객체의크기정보와움직임제약정보,추적결과를통해추정된객체의움직임 정보등을참조하여，검색영역을최소한으로줄일수있는실시간영상인식 방법을제공함에있다.
과제해결수단
[9] 상기목적을달성하기위한본발명의일실시예에따른，영상인식방법은， 영상에서제 1주기로관심객체를검출하는단계;및상기영상에서제 2주기로 상기관심객체를추적하는단계;를포함하고，상기제 1주기는，상기제 2주기 보다길다.
[10] 그리고,본발명의일실시예에따른영상인식방법은,상기영상의특징정보 및상기영상을촬영하는카메라의움직임정보중적어도하나를참조하여， 상기제 1주기및상기제 2주가를설정하는단계;를더포함할수있다.
[11] 또한，본발명의일실시예에따른영상인식방법은，영상을구성하는프레임들 중검출과추적이수행되지않은프레임에는보간기법을통해관심객체를 설정하는단계；를더포함할수있다.
[12] 그리고，상기검출단계는，스킨정보및에지정보증적어도하나를이용하여， 얼굴과차량번호판중적어도하나를포함하는관심객체를검출할수있다.
[13] 또한，상기검출단계는，스킨밀도와에지밀도가기준이상인영역에서만， 상기관심객체를검출할수있다.
[14] 그리고,상기검출단계및추적단계는，상기영상을촬영하는카메라가
바라보는장면의기하학적구조정보를참조하여 ,검출및추적영역을설정할 수있다.
[15] 또한，상기추적단계는，영역별객체움직임크기에기초하여，추적영역을 설정할수있다.
[16] 한편,본발명의다른실시예에따른，영상인식방법은，영상에서제 1주기로 관심객체를검출하는검출기;상기영상에서제 2주기로상기관심객체를 추적하는추적기；및상기제 1주기를상기제 2주기보다길게설정하는 설정부;를포함한다.
발명의효과
[17] 이상설명한바와같이，본발명의실시예들에따르면，검출기와추적기의동작 횟수를적웅적으로설정하여,저사양의임베디드환경에서도실시간영상 인식이가능해진다.
[18] 또한,영상내의고유정보 (스킨,경계선둥)，카메라가바라보는장면의
기히학적구조정보및영상의구조적특징으로인한객체의크기정보와움직임 제약정보，추적결과를통해추정된객체의움직임정보등을참조하여 ,검색 영역을최소한으로줄일수있어，계산량을최소화할수있음은물론,객체의 오검출률을획기적으로줄일수있게된다.
도면의간단한설명
[19] 도 1은본발명의실시예에따른영상기기의설명에제공되는도면，
[20] 도 2는영상인식동작계획설정결과를예시한도면,
[21] 도 3은스킨정보를이용한얼굴검출결과를예시한도면，
[22] 도 4는스킨정보를이용한검색영역설정결과를예시한도면，
[23] 도 5는에지정보를이용한검색영역최소화결과를예시한도면，
[24] 도 6은에지정보를이용한번호판검출결과를예시한도면，
[25] 도 7은영상장면의기하학구조정보를이용한번호판검색영역설정결과를 예시한도면，
[26] 도 8은영상장면의기하학구조정보를이용한차량및보행자검출을위한 검색영역설정결과를예시한도면，
[27] 도 9는슬라이딩원도우검색및결과로구한유사도맵을예시한도면,그리고， [28] 도 10은객체의움직임정보와장면에의한움직임정보를예시한도면이다. 발명의실시를위한최선의형태
[29] 이하에서는도면을참조하여본발명을보다상세하게설명한다.
[30]
[31] 1.저사양 « 베디드 ·랫폭화경에서심시? 영상이식
[32] 본발명의실시예에따른실시간영상인식방법은，카메라를통해촬영된
영상에대해，저사양플랫폼에서관심객체를실시간으로검출 /추적한다.
[33] 특히 ,본발명의실시예에따른실시간영상인식방법은,차량용블랙박스에서 개인정보보호 (얼굴과차량번호판의모자이크처리)，휴먼인터액션시스템， 범죄차량인식,보행자 /전방차량인식을통한차량층돌방지등에
웅용가능하며，한종류의객체검출 /추적이아닌모든종류의객체에대한 검출 /추적을추구한다.
[34] 본발명의실시예에따른실시간영상인식방법은,상대적으로계산량이많은 검출기는최소한으로동작시키고，계산량이적은추적기와보간을통해이를 보완하여，영상을구성하는모든프레임들에대해관심객체인식이수행되도록 하는방식이다.
[35] 도 1은본발명의실시예에따른영상기기의설명에제공되는도면이다.본 발명의실시예에따른영상기기는,도 1에도시된바와같이,동작계획 설정부 (110)，검출기 (120)및추적기 (130)를포함한다.
[36] 검출기 (120)는객체검출알고리즘을이용하여카메라를통해촬영된
영상으로부터관심객체를검출하고，추적기 (130)는객체추적알고리즘을 이용하여검출기 (120)에의해검출된관심객체를추적한다.
[37] 동작계획설정부 (110)는검출기 (120)와추적기 (130)의동작주기를설정한다. 검출기 (120)와추적기 (130)의동작주기설정시，동작계획설정부 (110)는영상의 특징정보 (스킨정보，에지정보등)와카메라정보 (위치정보，속도정보등)을 참조할수있다.
[38]
[39] 2.영상인식동작 (객체검출 /추적)계획섬정
[4이 영상을구성하는모든프레임에대해，검출기 (120)로관심객체를검출하면 검출성능은좋을수있지만,계산량이높아저사양플랫폼에서는블가능한 일이다.
[41] 이에,동작계획설정부 (110)는최대한적은횟수로검출기 (120)를동작시키고， 계산량이상대적으로매우적은추적기 (130)를보조적으로사용하여속도와 검출률을모두를도모한다.
[42] 또한,검출기 (120)와추적기 (120)가모두동작하지않는프레임들에대해서는 추적결과들로부터산출가능한움직임정보로관심객체위치를보간하여채워 넣는다.
[43] 도 2는영상인식동작계획설정결과를예시한도면이다.도 2에서는，
검출기 (120)는 30프레임마다한번씩동작하고,추적기 (130)는 3프레임마다한 번씩동작하여검출기 (10)의역할을대신하며，추적기 (130)가동작하지않는 프레임들은움직임정보기반의링크로관심객체위치를보간하여채워넣도록 설정된것을확인할수있다.
[44] 한편，검출기 (120)와추적기 (130)의동작주기를더욱효과적으로설정하기 위해，영상의특징정보 (스킨정보，에지정보둥)와카메라정보 (위치정보，속도 정보둥)를참조하여，검출기 (120)와추적기 (130)의동작주기를상황에따라 적웅적으로설정할수있다.
[45] 예를들어，카메라 (차량)의위치가고속도로이거나카메라 (차량)의속도가 일정속도이상일경우검출기 (120)의얼굴검출기능을정지시키는것이다. 고속도로또는카메라 (차량)의속도가빠른경우,빠른속도로인한영상 blur가 발생하여얼굴정보가심하게왜곡되어나오기때문에정보보호등을얼굴 검출이불필요하기때문이다.
[46] 한편，영상의복잡도가높은경우에는그렇지않은경우보다검출주기를길게 설정하여,계산량을즐여주는것이가능하다.또한,장면전환이빠른구간 (예를 들면，회전)시점에서는검출주기를짧게설정하여，검출의정확도가
높아지도록하는것도가능하다.
[47]
[48] 3. 상특 을이용하검새 < 역축소
[49] 객체검출을위해영상전체영역을검색하는것은불필요한계산량을늘릴뿐 아니라오검출발생올높이는또다른원인이된다.계산량을줄이며오검출률도 최소화하기위해서는객체를학습한데이터뿐만아니라배경으로부터객체를 구별할수있는객체의독립적인특징정보를활용할필요성이있다.
[50] 영상인식의오검출률을최소화하는동시에검출기 (120)의속도향상을위해 영상에서각객체의고유특징정보를활용한다.구체적으로는，얼굴의경우 공통적으로갖는칼라정보인스킨색 (피부색)정보를활용하며，번호판，차량, 보행자의경우객체주변에존재하는에지정보를활용한다.
[51] 검출기 (120)는영상의전체영역을검색하지않고,관심객체가존재할만한 스킨정보와경계선정보가있는영역만을검색하게되어，계산량을크게 줄이면서도오검출률을줄일수있다.
[52] 스킨색정보를활용할경우，도 3에제시한바와같은장면에서,도 4에제시된 바와같이스킨색이아닌배경을영역을효과적으로제거할수있으며이에따라 얼굴검출영역이현저히줄게되어계산량을낮출수있게된다. [53] 에지정보를활용할경우，도 5에제시된바와같이배경을영역을효과적으로 제거할수있으며이에따라객체검출영역이현저히줄게되어계산량을낮출 수있게된다.도 6은에지정보를이용한번호판검출결과를예시하였다.
[54] 배경으로지정된영역을검색하지않으므로배경영역에서발생하는
오검출률을크게줄일수있다.한편，오검출률을더욱줄이기위해각검출 결과에대한스킨 /에지밀도를추가적으로고려할수있다.
[55] 스킨 /에지밀도는，검출객체에서스킨색 /에지픽셀수를검출객체의넓이로 나눈값이며，이값이임계치이하인검출결과는신뢰도가낮은것으로 취급하여배제시킨다.
[56]
[57] 4.명상장며의기하학구조정보름이용하 색영역축소
[58] 검출기 (120)와추적기 (130)는카메라가바라보는장면의기하학적구조정보를 이용하여효과적인검색영역을설정한다.즉,도 7과도 8에제시된바와같이， 장면의구조로인해영상에서객체크기에따라영상내에객체가위치하는 영역이대략적으로정해질수있다.이를이용하여모든영역에서다양한크기를 고려하여객체를검색하는것이아닌,영역에맞는객체크기만을고려하여 검색에소모되는불필요한계산량을없애는것이다.
[59]
[60] 5.객체의외형모템과특징정보를이용한검색영역축소
[61] 추적기 (130)는검출된결과를이용하여,도 9에도시된바와같이，일정영역 내에서객체의외형모델을이용한검색을수행한다.도 9에도시된바와같이， 술라이딩원도우 (sliding window)검색후얻은유사도맵에서유사도가제일 높게나온픽샐의위치가추적결과이다.
[62] 또한,객체의외형모델외에객체의특징정보도함께활용하여추적성능을 향상킬수있다.얼굴의경우,얼굴검출시구한스킨밀도정보를이용하고， 번호판등의에지밀도정보를이용하여，추적시배경영역으로발산하여추적 실패가발생하는것을방지한다.
[63] 도 10에는객체들의움직임이짧은시간동안일정하게움직인다는가정과 영상의특정영역에서의움직임크기를사전정보로이용하여객체추적성능을 향상시키는방식이도시되어있다.도 10에도시된바와같이 ,영상의좌 /우 영역쎄서는큰움직임이발생하므로탐색범위를크게하고，영상의증앙 영역에서는큰움직임이발생하지않으므로탐색범위를작게하는것이다.
[64]
[65] 또한，이상에서는본발명의바람직한실시예에대하여도시하고
설명하였지만，본발명은상술한특정의실시예에한정되지아니하며， 청구범위에서청구하는본발명의요지를벗어남이없이당해발명이속하는 기술분야에서통상의지식을가진자에의해다양한변형실시가가능한것은 물론이고，이러한변형실시들은본발명의기술적사상이나전망으로부터 개별적으로이해되어져서는안될것이다ᅳ

Claims

청구범위
영상에서계 1주기로관심객체를검출하는단계;및
상기영상에서제 2주기로상기관심객체를추적하는단계;를 포함하고，
상기제 1주기는，상기제 2주기보다긴것을특징으로하는영상 인식방법.
청구항 1에있어서，
상기영상의특징정보및상기영상을촬영하는카메라의움직임 정보중적어도하나를참조하여,상기제 1주기및상기제 2 주기를설정하는단계;를더포함하는것을특징으로하는영상 인식방법.
청구항 1에있어서，
영상을구성하는프레임들증검출과추적이수행되지않은 프레임에는보간기법을통해관심객체를설정하는단계;를더 포함하는것을특징으로하는영상인식방법.
청구항 1에있어서，
상기검출단계는，
스킨정보및에지정보증적어도하나를이용하여，얼굴과차량 번호판중적어도하나를포함하는관심객체를검출하는것을 특징으로하는영상인식방법.
청구항 4에있어서,
상기검출단계는,
스킨밀도와에지밀도가기준이상인영역에서만，상기관심 객체를검출하는것을특징으로하는영상인식방법.
청구항 1에있어서.
상기검출단계및추적단계는，
상기영상을촬영하는카메라가바라보는장면의기하학적구조 정보를참조하여,검출및추적영역을설정하는것을특징으로 하는영상인식방법.
청구항 1에있어서，
상기추적단계는,
영역별객체움직임크기에기초하여，추적영역을설정하는것을 특징으로하는영상인식방법.
영상에서제 1주기로관심객체를검출하는검출기;
상기영상에서제 2주기로상기관심객체를추적하는추적기；및 상기제 1주기를상기제 2주기보다길게설정하는설정부；를 포함하는것을특징으로하는영상인식방법.
PCT/KR2014/012832 2014-12-24 2014-12-24 저사양 영상 기기에서의 실시간 영상 인식 방법 Ceased WO2016104831A1 (ko)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR10-2014-0187887 2014-12-24
KR1020140187887A KR101641647B1 (ko) 2014-12-24 2014-12-24 저사양 영상 기기에서의 실시간 영상 인식 방법

Publications (1)

Publication Number Publication Date
WO2016104831A1 true WO2016104831A1 (ko) 2016-06-30

Family

ID=56150837

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2014/012832 Ceased WO2016104831A1 (ko) 2014-12-24 2014-12-24 저사양 영상 기기에서의 실시간 영상 인식 방법

Country Status (2)

Country Link
KR (1) KR101641647B1 (ko)
WO (1) WO2016104831A1 (ko)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11721026B2 (en) * 2017-01-20 2023-08-08 Sony Corporation Information processing apparatus, information processing method, and information processing system

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102212358B1 (ko) * 2019-05-10 2021-02-04 명홍철 실시간 영상에서의 마스킹 방법 및 장치
KR102183689B1 (ko) * 2020-03-30 2020-11-30 에스큐아이소프트 주식회사 실시간 영상의 객체 마스킹 처리 시스템 및 이를 이용한 마스킹 처리 방법

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20060026233A (ko) * 2004-09-20 2006-03-23 삼성전자주식회사 졸음 감지 장치 및 그 방법
JP2010004261A (ja) * 2008-06-19 2010-01-07 Sony Corp 画像処理装置、及び画像処理方法
KR20120055306A (ko) * 2010-11-23 2012-05-31 현대모비스 주식회사 객체 인식 시스템, 차량용 장애물 인식 시스템 및 차량용 장애물 인식 방법
KR20130007243A (ko) * 2011-06-30 2013-01-18 (주)베라시스 차량 카메라를 이용한 전방 충돌 경보 방법 및 시스템
KR101375765B1 (ko) * 2012-12-14 2014-03-20 (주)오시에스티 칼만 필터를 이용한 이동 물체 추적 장치 및 방법
JP2014186547A (ja) * 2013-03-22 2014-10-02 Toshiba Corp 移動物体追跡システム、方法及びプログラム

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20060026233A (ko) * 2004-09-20 2006-03-23 삼성전자주식회사 졸음 감지 장치 및 그 방법
JP2010004261A (ja) * 2008-06-19 2010-01-07 Sony Corp 画像処理装置、及び画像処理方法
KR20120055306A (ko) * 2010-11-23 2012-05-31 현대모비스 주식회사 객체 인식 시스템, 차량용 장애물 인식 시스템 및 차량용 장애물 인식 방법
KR20130007243A (ko) * 2011-06-30 2013-01-18 (주)베라시스 차량 카메라를 이용한 전방 충돌 경보 방법 및 시스템
KR101375765B1 (ko) * 2012-12-14 2014-03-20 (주)오시에스티 칼만 필터를 이용한 이동 물체 추적 장치 및 방법
JP2014186547A (ja) * 2013-03-22 2014-10-02 Toshiba Corp 移動物体追跡システム、方法及びプログラム

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11721026B2 (en) * 2017-01-20 2023-08-08 Sony Corporation Information processing apparatus, information processing method, and information processing system

Also Published As

Publication number Publication date
KR101641647B1 (ko) 2016-07-22
KR20160078561A (ko) 2016-07-05

Similar Documents

Publication Publication Date Title
JP6873126B2 (ja) 顔位置追跡方法、装置及び電子デバイス
US9405974B2 (en) System and method for using apparent size and orientation of an object to improve video-based tracking in regularized environments
Jung et al. Detecting moving objects using a single camera on a mobile robot in an outdoor environment
EP2570993B1 (en) Egomotion estimation system and method
US9904852B2 (en) Real-time object detection, tracking and occlusion reasoning
CN107688764B (zh) 检测车辆违章的方法及装置
CN104318206B (zh) 一种障碍物检测方法和装置
US20180322640A1 (en) System and method for detecting moving obstacles based on sensory prediction from ego-motion
KR20180042254A (ko) 오브젝트 추적을 위한 시스템들 및 방법들
CN113396423A (zh) 处理来自基于事件的传感器的信息的方法
CN110706247A (zh) 一种目标跟踪方法、装置及系统
EP2700054B1 (en) System and method for video-based vehicle detection
CN106530328B (zh) 一种基于视频图像对运动目标检测和平滑跟随的方法
KR102528002B1 (ko) 탑뷰 영상 생성 장치 및 그 방법
JP2015036980A (ja) 駐車区画占有率判定のための映像および視覚ベースのアクセス制御のハイブリッド方法およびシステム
KR20150067680A (ko) 차량용 제스처 인식 시스템 및 그 방법
US11482007B2 (en) Event-based vehicle pose estimation using monochromatic imaging
Boufares et al. Moving object detection system based on the modified temporal difference and otsu algorithm
WO2016104831A1 (ko) 저사양 영상 기기에서의 실시간 영상 인식 방법
KR20220098677A (ko) 교통사고 예측 방법 및 시스템
CN115035456B (zh) 视频去噪方法、装置、电子设备和可读存储介质
JP7007649B2 (ja) オプティカルフロー推定装置、オプティカルフロー推定方法、オプティカルフロー推定システム、及びオプティカルフロー推定プログラム、並びにヨーレート推定装置、ヨーレート推定方法、ヨーレート推定システム、及びヨーレート推定プログラム
KR101492366B1 (ko) 차량 사고 감지 방법 및 그 장치
Wang et al. Video stabilization based on high degree b-spline smoothing
CN106683113B (zh) 特征点跟踪方法和装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14909145

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 14909145

Country of ref document: EP

Kind code of ref document: A1