Apple's US 8,064,685: 3D object recognition
The abstract states:
A method, device, system, and computer program for object recognition of a 3D object of a certain object class using a statistical shape model for recovering 3D shapes from a 2D representation of the 3D object and comparing the recovered 3D shape with known 3D to 2D representations of at least one object of the object class.
AND the first claim:
A method for object recognition of a three dimensional (3D) object, the method comprising at an electronic device, maintaining a statistical 3D-shape model used to express 3D-shapes of object features in terms of a median 3D-shape, .mu., and a base of 3D-shapes, W, where each 3D-shape, t.sub.3D(i), is expressed in terms of an associated latent characteristic, u(i), as t.sub.3D(i)=W*u(i)+.mu., where 3D-shape vectors t.sub.3D and .mu. are d-dimensional, a matrix of the base of 3D-shapes is (d.times.q)-dimensional, and latent characteristic vectors u are q-dimensional, and where the object features include points, lines and contours corresponding to objects of an object class; accessing a profile corresponding to at least a known object of the object class, the profile stored on persistent storage communicatively coupled with the electronic device, the profile including a latent characteristic, u(k), associated with a 3D-shape, t.sub.3D(k), of the known object, in accordance with the statistical 3D-shape model, t.sub.3D(k)=W*u(k)+.mu.; detecting, by the electronic device, object features, t.sub.2D(o), of an object depicted in a two dimensional (2D) image, where a vector, t.sub.2D, representing representation of said detected object features is e-dimensional; determining, by the electronic device, a 3D-shape, t.sub.3D(o), corresponding to the detected object features, t.sub.2D(o), such that a projection model f(t.sub.3D) applied to the determined 3D-shape, t.sub.3D(o), results in the detected object features t.sub.2D(o)=f[t.sub.3D(o)], said determining the 3D-shape, t.sub.3D(o), comprising: optimizing the projection model, f[t.sub.3D(i)], applied to 3D-shapes, t.sub.3D(i), expressed in terms of respective latent characteristics, u(i), in accordance with the statistical 3D-shape model, t.sub.3D(i)=W*u(i)+.mu., said optimizing over the respective latent characteristics u(i), and selecting a latent characteristic u(o) corresponding to the optimized projection model, f[t.sub.3D(o)], such that the optimized projection model applied to a 3D shape expressed in terms of the selected latent characteristic, u(o), results in the detected object features t.sub.2D(o)=f[W*u(o)+.mu.)], where the 3D-shape expressed in terms of the selected latent characteristic, u(o), represents the determined 3D-shape, t.sub.3D(o); comparing, by the electronic device, the selected latent characteristic u(o), corresponding to the detected object features, t.sub.2D(o), at least with the latent characteristic included in the stored profile, u(k), and associated with the 3D-shape, t.sub.3D(k), of the known object; selectively recognizing the object depicted in the 2D image as the known object based on said comparing.
The text beginning -- the method comprising at an electronic device-- is odd form for a method claim.
Also odd is the priority claim: This U.S. non-provisional application claims benefit of priority under 35 U.S.C. .sctn.119 of Swedish Patent Application No. 0402048-3, filed on Aug. 19, 2004, and U.S. Provisional Application No. 60/603,266 filed on Aug. 23, 2004, the entire contents of both of which are incorporated herein by reference.
The US case, 11/201,419 , was filed in 2005, with the patent issuing more than six years later.
Cited within the issued patent:
Blanz, Volker et al., Face Recognition Based on Fitting a 3D Morphable Model, (Sep. 2003), IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 25, No. 9, pp. 1063-1074.
A method, device, system, and computer program for object recognition of a 3D object of a certain object class using a statistical shape model for recovering 3D shapes from a 2D representation of the 3D object and comparing the recovered 3D shape with known 3D to 2D representations of at least one object of the object class.
AND the first claim:
A method for object recognition of a three dimensional (3D) object, the method comprising at an electronic device, maintaining a statistical 3D-shape model used to express 3D-shapes of object features in terms of a median 3D-shape, .mu., and a base of 3D-shapes, W, where each 3D-shape, t.sub.3D(i), is expressed in terms of an associated latent characteristic, u(i), as t.sub.3D(i)=W*u(i)+.mu., where 3D-shape vectors t.sub.3D and .mu. are d-dimensional, a matrix of the base of 3D-shapes is (d.times.q)-dimensional, and latent characteristic vectors u are q-dimensional, and where the object features include points, lines and contours corresponding to objects of an object class; accessing a profile corresponding to at least a known object of the object class, the profile stored on persistent storage communicatively coupled with the electronic device, the profile including a latent characteristic, u(k), associated with a 3D-shape, t.sub.3D(k), of the known object, in accordance with the statistical 3D-shape model, t.sub.3D(k)=W*u(k)+.mu.; detecting, by the electronic device, object features, t.sub.2D(o), of an object depicted in a two dimensional (2D) image, where a vector, t.sub.2D, representing representation of said detected object features is e-dimensional; determining, by the electronic device, a 3D-shape, t.sub.3D(o), corresponding to the detected object features, t.sub.2D(o), such that a projection model f(t.sub.3D) applied to the determined 3D-shape, t.sub.3D(o), results in the detected object features t.sub.2D(o)=f[t.sub.3D(o)], said determining the 3D-shape, t.sub.3D(o), comprising: optimizing the projection model, f[t.sub.3D(i)], applied to 3D-shapes, t.sub.3D(i), expressed in terms of respective latent characteristics, u(i), in accordance with the statistical 3D-shape model, t.sub.3D(i)=W*u(i)+.mu., said optimizing over the respective latent characteristics u(i), and selecting a latent characteristic u(o) corresponding to the optimized projection model, f[t.sub.3D(o)], such that the optimized projection model applied to a 3D shape expressed in terms of the selected latent characteristic, u(o), results in the detected object features t.sub.2D(o)=f[W*u(o)+.mu.)], where the 3D-shape expressed in terms of the selected latent characteristic, u(o), represents the determined 3D-shape, t.sub.3D(o); comparing, by the electronic device, the selected latent characteristic u(o), corresponding to the detected object features, t.sub.2D(o), at least with the latent characteristic included in the stored profile, u(k), and associated with the 3D-shape, t.sub.3D(k), of the known object; selectively recognizing the object depicted in the 2D image as the known object based on said comparing.
The text beginning -- the method comprising at an electronic device-- is odd form for a method claim.
Also odd is the priority claim: This U.S. non-provisional application claims benefit of priority under 35 U.S.C. .sctn.119 of Swedish Patent Application No. 0402048-3, filed on Aug. 19, 2004, and U.S. Provisional Application No. 60/603,266 filed on Aug. 23, 2004, the entire contents of both of which are incorporated herein by reference.
The US case, 11/201,419 , was filed in 2005, with the patent issuing more than six years later.
Cited within the issued patent:
Blanz, Volker et al., Face Recognition Based on Fitting a 3D Morphable Model, (Sep. 2003), IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 25, No. 9, pp. 1063-1074.
posted by Lawrence B. Ebert at 5:50 AM
0 Comments:
Post a Comment
<< Home