Methods for object recognition and related arrangements
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06K-009/00
G06T-015/02
G06F-017/30
G06K-009/62
G06T-017/00
G06K-009/46
G06T-015/20
출원번호
US-0050063
(2016-02-22)
등록번호
US-9478063
(2016-10-25)
발명자
/ 주소
Rhoads, Geoffrey B.
Bai, Yang
출원인 / 주소
Digimarc Corporation
대리인 / 주소
Digimarc Corporation
인용정보
피인용 횟수 :
1인용 특허 :
4
초록▼
Methods and arrangements involving portable user devices such smartphones and wearable electronic devices are disclosed, as well as other devices and sensors distributed within an ambient environment. Some arrangements enable a user to perform an object recognition process in a computationally- and
Methods and arrangements involving portable user devices such smartphones and wearable electronic devices are disclosed, as well as other devices and sensors distributed within an ambient environment. Some arrangements enable a user to perform an object recognition process in a computationally- and time-efficient manner. Other arrangements enable users and other entities to, either individually or cooperatively, register or enroll physical objects into one or more object registries on which an object recognition process can be performed. Still other arrangements enable users and other entities to, either individually or cooperatively, associate registered or enrolled objects with one or more items of metadata. A great variety of other features and arrangements are also detailed.
대표청구항▼
1. An image processing method for generating a first non-realistic 2D symmetric silhouette from model data that defines a 3D exterior surface corresponding to a physical reference object, the model data comprising data stored in an electronic memory that identifies a 3D position for each of plural m
1. An image processing method for generating a first non-realistic 2D symmetric silhouette from model data that defines a 3D exterior surface corresponding to a physical reference object, the model data comprising data stored in an electronic memory that identifies a 3D position for each of plural model points on said surface, the resulting non-realistic 2D symmetric silhouette being useful in identifying an unknown physical object from imagery thereof, the method comprising the acts: identifying, with one or more hardware processors, a dominant plane P for the 3D surface defined by the model data, said identifying act comprising (a) identifying a plane tangent to said 3D surface that contains a greatest number of model points, or (b) identifying a plane tangent to said 3D surface that contains three or more of said model points and spans a largest area;determining, with one or more hardware processors, a primary reference axis about which the model data should be rotated, said primary reference axis passing through the modeled object and being perpendicular to the dominant plane P; andestimating, with one or more hardware processors, said first non-realistic 2D symmetric silhouette of the reference object based on rotation of the model data, or a viewpoint, about the primary reference axis, said estimating being based on plural depictions of the object viewed from an orbit around the object and in a plane parallel to the dominant plane P;wherein said symmetric silhouette does not match any silhouette of said physical reference object. 2. The method of claim 1 in which the act of identifying the dominant plane comprises identifying a plane: (a) that is tangent to the 3D surface defined by the model data, and (b) which contains a greatest number of said model points. 3. The method of claim 1 in which the act of identifying the dominant plane comprises identifying a plane: (a) that is tangent to the 3D surface defined by the model data, and (b) which contains three or more of said model points spanning a largest planar area. 4. The method of claim 1 in which the plane P extends in X and Y directions, and the act of determining the primary reference axis comprises: averaging X coordinates for each model point to yield an average X′-coordinate;averaging Y coordinates for each model point to yield an average Y′-coordinate; andlocating the primary reference axis to pass through said average X′ and average Y′ coordinates on the dominant plane P. 5. The method of claim 1 in which the act of estimating the non-realistic silhouette includes the acts: establishing a viewpoint, the viewpoint being established by plural parameters;for each of a plurality of rotational states of the model data about the primary reference axis, projecting a silhouette from the viewpoint onto a 2D plane, said plurality including twelve different rotational states; andaccumulating the plural projected silhouettes to yield a single, accumulated, non-realistic 2D symmetric silhouette. 6. The method of claim 1 in which the act of estimating the non-realistic silhouette includes the acts: establishing a plurality of viewpoints in a plane parallel to the plane P, said plurality including twelve different viewpoints;projecting a silhouette from each viewpoint onto a 2D plane; andaccumulating the plural projected silhouettes to yield a single, accumulated, non-realistic 2D symmetric silhouette. 7. The method of claim 1 that further includes generating a second non-realistic 2D silhouette from the model data, by acts comprising: projecting model points into the plane P;applying principal component analysis to identify a line L in the plane P that accounts for a largest amount of variability in the projected model points;determining a secondary reference axis that is parallel to line L and passes through the modeled object; andestimating said second non-realistic silhouette of the reference object based on rotation of the model data, or a viewpoint, about the secondary reference axis. 8. The method of claim 7 that further includes: defining a secondary plane Q that is perpendicular to said identified line L, said secondary plane extending in M and N directions;averaging M coordinates for each model point to yield an average M′-coordinate;averaging N coordinates for each model point to yield an average N′-coordinate; andlocating the secondary reference axis to pass through said average M′ and average N′ coordinates on the secondary plane Q. 9. The method of claim 7 that further includes comparing said second non-realistic 2D silhouette with imagery of an unknown physical object, to determine whether said reference object should be further considered as a possible match to said unknown physical object. 10. The method of claim 9 that further includes comparing said first non-realistic 2D silhouette with imagery of the unknown physical object, to determine whether said reference object should be further considered as a possible match to said unknown physical object. 11. The method of claim 1 that further includes comparing said first non-realistic 2D silhouette with imagery of an unknown physical object, to determine whether said reference object should be further considered as a possible match to said unknown physical object. 12. The method of claim 11 that further includes first and second systems cooperating to gather the imagery of the unknown physical object, the first system comprising a mobile phone device, and the second system comprising an ambient infrastructure camera that has a view of said unknown physical object. 13. Apparatus for generating a first non-realistic 2D symmetric silhouette from model data that defines a 3D exterior surface corresponding to a physical reference object, the model data comprising data stored in an electronic memory that identifies a 3D position for each of plural model points on said surface, the resulting non-realistic 2D symmetric silhouette being useful in identifying an unknown physical object from imagery thereof, the apparatus comprising electronic circuitry including: means for identifying a dominant plane P for the 3D surface defined by the model data, said means comprising: (a) means for identifying a plane tangent to the 3D surface that contains a greatest number of model points, or (b) means for identifying a plane tangent to the 3D surface that contains three or more of said model points and spans a largest area;means for determining a primary reference axis about which the model data should be rotated, said primary reference axis passing through the modeled object and being perpendicular to the dominant plane P; andmeans for estimating said first non-realistic 2D symmetric silhouette of the reference object based on rotation of the model data, or a viewpoint, about the primary reference axis, said estimating being based on plural depictions of the object viewed from an orbit around the object and in a plane parallel to the dominant plane P;wherein said symmetric silhouette does not match any silhouette of said physical reference object. 14. An image processing method for generating a first non-realistic 2D symmetric silhouette from model data that defines a 3D exterior surface corresponding to a physical reference object, the model data comprising data stored in an electronic memory that identifies a 3D position for each of plural model points on said surface, the resulting non-realistic 2D symmetric silhouette being useful in identifying an unknown physical object from imagery thereof, the method comprising the acts: identifying, with one or more hardware processors, a dominant plane P for the 3D surface defined by the model data;projecting, with one or more hardware processors, model points into the plane P;applying principal component analysis, using one or more hardware processors, to identify a line L in the plane P that accounts for a largest amount of variability in the projected model points;determining, with one or more hardware processors, a reference axis that is parallel to line L and passes through the modeled object; andestimating said non-realistic 2D symmetric silhouette of the reference object based on rotation of the model data, or a viewpoint, about the reference axis;wherein said symmetric silhouette does not match any silhouette of said physical reference object. 15. The method of claim 5 in which said plurality includes 18 different rotational states. 16. The method of claim 5 in which said plurality includes 36 different rotational states.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (4)
Nakamura, Akira; Iwai, Yoshiaki; Yoshigahara, Takayuki, Information processing apparatus, information processing method, image identifying apparatus, image identifying method, and program.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.