Motion calculation device and motion calculation method
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
H04N-007/18
G06T-007/20
G06K-009/00
출원번호
US-0007309
(2011-01-14)
등록번호
US-8866901
(2014-10-21)
우선권정보
JP-P2010-135487 (2010-06-14)
발명자
/ 주소
Azuma, Takahiro
Okutomi, Masatoshi
Sugimoto, Shigeki
출원인 / 주소
Honda Elesys Co., Ltd.
대리인 / 주소
Arent Fox LLP
인용정보
피인용 횟수 :
1인용 특허 :
12
초록▼
A motion calculation device includes an image-capturing unit configured to capture an image of a range including a plane and outputs the captured image, an extraction unit configured to extract a region of the plane from the image, a detection unit configured to detect feature points and motion vect
A motion calculation device includes an image-capturing unit configured to capture an image of a range including a plane and outputs the captured image, an extraction unit configured to extract a region of the plane from the image, a detection unit configured to detect feature points and motion vectors of the feature points from a plurality of images captured by the image-capturing unit at a predetermined time interval; and a calculation unit configured to calculate the motion of the host device based on both of an epipolar constraint relating to the feature points and a homography relating to the region.
대표청구항▼
1. A motion calculation device comprising: an image-capturing unit configured to capture an image of a range including a plane and to output the captured image;an extraction unit configured to extract a plane region from the image by selecting image blocks from the image and by incorporating some of
1. A motion calculation device comprising: an image-capturing unit configured to capture an image of a range including a plane and to output the captured image;an extraction unit configured to extract a plane region from the image by selecting image blocks from the image and by incorporating some of the image blocks into the plan region when a difference between a plane orientation of each of the image blocks and that of the plane region is smaller than a predetermined threshold;a detection unit configured to detect feature points and motion vectors of the feature points from a plurality of images captured by the image-capturing unit at a predetermined time interval and to output a set of feature point correspondences; anda calculation unit configured to receive the set of feature point correspondences from the detection unit and to calculate the motion of the host device by simultaneously using both of an epipolar constraint relating to the feature points and a homography relating to the plane region, wherein the motion of the host device comprises a rotation parameter ωT=[ωx,ωy,ωz ] and a translation vector tT=[tx,ty,tz], wherein a plane orientation is expressed by nT=[nx,n y,ny], andwherein the calculation unit minimizes a cost function expressed by Q1(μ)=∑p∈B,B∈∏tαp{It(w(p;μ))-It-1(p)}2+∑(p,q)∈Ψtβp{qT[t]×R(ω)p}2+λ(tTt-v2)2,where It represents an image at time t, (p,q) represents a feature point correspondence, Ψt represents a set of feature point correspondences, B represents a plane region in an image, πt represents a set of plane regions, μ, ω, t, and R(ω) represent motion parameters of the image-capturing unit, v represents speed information of the host device, w(p;μ) represents warp function, and αp, βp, and λ represent weight coefficients. 2. The motion calculation device according to claim 1, wherein the calculation unit is configured to minimize a cost function based on the epipolar constraint and the homography to calculate the motion. 3. A motion calculation method in a motion calculation device which calculates the motion of the host device, the motion calculation method comprising: causing an image-capturing unit to capture an image of a range including a plane and to output the captured image;causing an extraction unit to extract a plane region from the image by selecting image blocks from the image and by incorporating some of the image blocks into the plane region when a difference between a plane orientation of each of the image blocks and that of the plane region is smaller than a predetermined threshold;causing a detection unit to detect feature points and motion vectors of the feature points from a plurality of images captured by the image-capturing unit at a predetermined time interval and to output a set of feature point correspondences; andcausing a calculation unit to receive the set of feature point correspondences from the detection unit and to calculate the motion of the host device by_simultaneously using both of an epipolar constraint relating to the feature points and a homography relating to the plane region, wherein the motion of the host device comprises a rotation parameter ωT=[ωx,ωy,ωz] and a translation vector tT=[tx,ty,tz], wherein a plane orientation is expressed by nT=[nx,ny,nz], andwherein the calculation unit minimizes a cost function expressed by Q1(μ)=∑p∈B,B∈∏tαp{It(w(p;μ))-It-1(p)}2+∑(p,q)∈Ψtβp{qT[t]×R(ω)p}2+λ(tTt-v2)2Q1(μ)=∑p∈B,B∈∏tαp{It(w(p;μ))-It-1(p)}2+∑(p,q)∈Ψtβp{qT[t]×R(ω)p}2+λ(tTt-v2)2,where It represents an image at time t, (p,q) represents a feature point correspondence Ψt represents a set of feature point correspondences, B represents a plane region in an image, πt represents a set of plane regions, μ, ω, t, and R(ω) represent motion parameters of the image-capturing unit, v represents speed information of the host device, w(p;μ) represents warp function, and αp, βp, and λ represent weight coefficients.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (12)
Choe, Tae Eun; Medioni, Gerard, 3-D reconstruction and registration.
Stein, Gideon S.; Shashua, Amnon; Gdalyahu, Yoram; Liyatan, Harel, Fusion of far infrared and visible images in enhanced obstacle detection in automotive applications.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.