US6959109B2
Expired / Lapsed
US
⏱️ 0.0 years remaining
System and method for pose-angle estimation
Assignee: Idemia Identity and Security USA LLC
•
Filing: 2002-06-20
•
Granted: 2005-10-25
•
Expires: 2022-06-20
Abstract
Abstract A system and method are disclosed for determining the pose angle of an object in an input image. In a preferred embodiment, the present system comprises a pose estimator having a prototype projector, a regression estimator, and an angle calculator. The prototype projector is preferably adapted to reduce the input image dimensionality for faster further processing by projecting the input pixels of the image onto a Self-Organizing Map (SOM) neural network. The regression estimator is preferably implemented as a neural network and adapted to map the projections to a pattern unique to each pose. The angle calculator preferably includes a curve fitter and an error analyzer. The curve fitter is preferably adapted to estimate the pose angle from the mapping pattern. The error analyzer is preferably adapted to produce a confidence signal representing the likelihood of the input image being a face at the calculated pose. The system also preferably includes two network trainers responsible for synthesizing the neural networks.
Inventors
- 👤 Mohamed Nabil Moustafa
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-09 23:06
CPC Classifications
G; G06; G06T; G06T7/00; G06T7/70; G06F; G06F18/00; G06F18/20; G06F18/21; G06F18/213; G06F18/2137; G06V; G06V10/00; G06V10/20; G06V10/24; G06V10/242; G06V40/00; G06V40/10; G06V40/16; G06V40/161; G06V40/164
Claims Analysis
Total Claims: 53
Claim 1
Independent Claim
224 words
A system for determining a pose angle of an object in an input image in at least one dimension, comprising: a first neural network trained in an unsupervised mode using a first plurality of training images to develop a plurality of weight vectors, each weight vector representing a plurality of the training images; a projector adapted to receive an input image and generate a projection vector comprising a plurality of calculated distances representing distances between the input image and the weight vectors; a second neural network trained in a supervised mode with a second plurality of training images, the second plurality of training images comprising objects at known pose angles, the second neural network comprising a plurality of neurons, each neuron tuned to a particular pose angle and adapted to receive the projection vector and output a value in accordance with a function that is even and monotonically decreasing for changes in the pose angle of the objects from the pose angle to which each neuron is tuned; a curve fitter adapted to receive the values output by the neurons and estimate the pose angle of the object in at least one dimension; and an error calculator that generates a value representative of the error between the values output by the neurons and the calculated values for the function using the estimated pose angle.
Claim 2
Independent Claim
224 words
A system for determining a pose angle of an object in an input image in at least one dimension, comprising: a first neural network trained in an unsupervised mode using a first plurality of training images to develop a plurality of weight vectors, each weight vector representing a plurality of the training images; a projector adapted to receive an input image and generate a projection vector comprising a plurality of calculated distances representing distances between the input image and the weight vectors; a second neural network trained in a supervised mode with a second plurality of training images, the second plurality of training images comprising objects at known pose angles, the second neural network comprising a plurality of neurons, each neuron tuned to a particular pose angle and adapted to receive the projection vector and output a value in accordance with a function that is even and monotonically decreasing for changes in the pose angle of the objects from the pose angle to which each neuron is tuned; a curve fitter adapted to receive the values output by the neurons and estimate the pose angle of the object in at least one dimension; and an error calculator that generates a value representative of the error between the values output by the neurons and the calculated values for the function using the estimated pose angle.
Claim 3
Dependent Claim
12 words
The system of claim 1 , wherein the object is a head.
Claim 4
Dependent Claim
33 words
The system of claim 1 , wherein the function is: cos(α(x)−α k ), where α k is the angle to which a neuron is tuned, and where α(x) is the estimated pose angle.
Claim 5
Dependent Claim
16 words
The system of claim 1 , wherein the curve fitter implements a minimum square-error curve-fitting algorithm.
Claim 6
Dependent Claim
24 words
The system of claim 1 , further comprising an imaging sensor adapted to capture an image in a field of view of the sensor.
Claim 7
Dependent Claim
14 words
The system of claim 5 , wherein the imaging sensor comprises a video camera.
Claim 8
Dependent Claim
14 words
The system of claim 5 , wherein the imaging sensor comprises a digital camera.
Claim 9
Dependent Claim
16 words
The system of claim 1 , wherein the input image is derived form a live subject.
Claim 10
Dependent Claim
15 words
The system of claim 1 , wherein the input image is derived from a photograph.
Claim 11
Dependent Claim
16 words
The system of claim 1 , wherein the input image is derived from a printed image.
Claim 12
Dependent Claim
16 words
The system of claim 1 , wherein the input image is derived from an artists' rendering.
Claim 13
Dependent Claim
31 words
The system of claim 1 , wherein at least some of the first plurality of training images are the same as at least some of the second plurality of training images.
Claim 14
Dependent Claim
25 words
The system of claim 12 , wherein the first plurality of training images and the second plurality of training images comprise the same training images.
Claim 15
Dependent Claim
25 words
The system of claim 1 , wherein the number of weight vectors is greater than the number of output nodes of the second neural network.
Claim 16
Dependent Claim
25 words
The system of claim 14 , wherein the number of weight vectors is approximately twice the number of output nodes of the second neural network.
Claim 17
Dependent Claim
23 words
The system of claim 1 , wherein the input image is generated by cropping a portion of an image that contains the object.
Claim 18
Dependent Claim
19 words
The system of claim 1 , wherein the projector is adapted to rescale the input image to canonical dimensions.
Claim 19
Dependent Claim
16 words
The system of claim 17 , wherein the canonical dimensions are 20 pixels by 20 pixels.
Claim 20
Dependent Claim
14 words
The system of claim 1 , wherein the at least one dimension is yaw.
Claim 21
Dependent Claim
14 words
The system of claim 1 , wherein the at least one dimension is pitch.
Claim 22
Dependent Claim
14 words
The system of claim 1 , wherein the at least one dimension is roll.
Claim 23
Dependent Claim
20 words
The system of claim 1 , wherein the system is adapted to estimate the yaw, pitch, and roll pose angles.
Claim 24
Independent Claim
218 words
A method for determining a pose angle of an object in an input image in at least one dimension, comprising: training a first neural network in an unsupervised mode using a first plurality of training images to develop a plurality of weight vectors, each weight vector representing a plurality of the training images; projecting the input image onto the first neural network to generate a projection vector comprising a plurality of calculated distances representing distances between the input image and the weight vectors; training a second neural network in a supervised mode with a second plurality of training images, the second plurality of training images comprising objects at known pose angles, the second neural network comprising a plurality of neurons, each neuron tuned to a particular pose angle; processing the projection vector in the second neural network in accordance with a function that is even and monotonically decreasing for changes in the pose angle of the object from the pose angle to which each neuron is tuned to generate a plurality of output values; fitting the output values to a curve to estimate the pose angle of the object in at least one dimension; and calculating a value representative of the error between the output values and the calculated values for the function using the estimated pose angle.
Claim 25
Independent Claim
218 words
A method for determining a pose angle of an object in an input image in at least one dimension, comprising: training a first neural network in an unsupervised mode using a first plurality of training images to develop a plurality of weight vectors, each weight vector representing a plurality of the training images; projecting the input image onto the first neural network to generate a projection vector comprising a plurality of calculated distances representing distances between the input image and the weight vectors; training a second neural network in a supervised mode with a second plurality of training images, the second plurality of training images comprising objects at known pose angles, the second neural network comprising a plurality of neurons, each neuron tuned to a particular pose angle; processing the projection vector in the second neural network in accordance with a function that is even and monotonically decreasing for changes in the pose angle of the object from the pose angle to which each neuron is tuned to generate a plurality of output values; fitting the output values to a curve to estimate the pose angle of the object in at least one dimension; and calculating a value representative of the error between the output values and the calculated values for the function using the estimated pose angle.
Claim 26
Dependent Claim
12 words
The method of claim 23 , wherein the object is a head.
Claim 27
Dependent Claim
16 words
The method of claim 24 , wherein the head includes a face discernable in the image.
Claim 28
Dependent Claim
33 words
The method of claim 23 , wherein the function is: cos(α(x)−α k ), where α k is the angle to which a neuron is tuned, and where α(x) is the estimated pose angle.
Claim 29
Dependent Claim
16 words
The method of claim 23 , wherein the curve fitter implements a minimum square-error curve-fitting algorithm.
Claim 30
Dependent Claim
24 words
The method of claim 23 , further comprising an imaging sensor adapted to capture an image in a field of view of the sensor.
Claim 31
Dependent Claim
14 words
The method of claim 28 , wherein the imaging sensor comprises a video camera.
Claim 32
Dependent Claim
14 words
The method of claim 28 , wherein the imaging sensor comprises a digital camera.
Claim 33
Dependent Claim
16 words
The method of claim 23 , wherein the input image is derived from a live subject.
Claim 34
Dependent Claim
15 words
The method of claim 23 , wherein the input image is derived from a photograph.
Claim 35
Dependent Claim
16 words
The method of claim 23 , wherein the input image is derived from a printed image.
Claim 36
Dependent Claim
16 words
The method of claim 23 , wherein the input image is derived from an artists' rendering.
Claim 37
Dependent Claim
31 words
The method of claim 23 , wherein at least some of the first plurality of training images are the same as at least some of the second plurality of training images.
Claim 38
Dependent Claim
25 words
The method of claim 35 , wherein the first plurality of training images and the second plurality of training images comprise the same training images.
Claim 39
Dependent Claim
25 words
The method of claim 23 , wherein the number of weight vectors is greater than the number of output nodes of the second neural network.
Claim 40
Dependent Claim
25 words
The method of claim 37 , wherein the number of weight vectors is approximately twice the number of output nodes of the second neural network.
Claim 41
Dependent Claim
23 words
The method of claim 23 , wherein the input image is generated by cropping a portion of an image that contains the object.
Claim 42
Dependent Claim
19 words
The method of claim 23 , wherein the projector is adapted to rescale the input image to canonical dimensions.
Claim 43
Dependent Claim
16 words
The method of claim 40 , wherein the canonical dimensions are 20 pixels by 20 pixels.
Claim 44
Dependent Claim
14 words
The method of claim 23 , wherein the at least one dimension is yaw.
Claim 45
Dependent Claim
14 words
The method of claim 23 , wherein the at least one dimension is pitch.
Claim 46
Dependent Claim
14 words
The method of claim 23 , wherein the at least one dimension is roll.
Claim 47
Dependent Claim
20 words
The method of claim 23 , wherein the system is adapted to estimate the yaw, pitch, and roll pose angles.
Claim 48
Independent Claim
219 words
A system for determining a pose angle of an object in an input image in at least one dimension, comprising: a first neural network using a first plurality of training images to develop a plurality of weight vectors, each weight vector representing a plurality of the training images; a projector adapted to receive an input image and generate a projection vector comprising a plurality of calculated distances representing distances between the input image and the weight vectors; a second neural network trained in a supervised mode with a second plurality of training images, the second plurality of training images comprising objects at known pose angles, the second neural network comprising a plurality of neurons, each neuron tuned to a particular pose angle and adapted to receive the projection vector and output a value in accordance with a function that is even and monotonically decreasing for changes in the pose angle of the object from the pose angle to which each neuron is tuned; a curve fitter adapted to receive the values output by the neurons and estimate the pose angle of the object in at least one dimension; and an error calculator that generates a value representative of the error between the values output by the neurons and the calculated values for the function using the estimated pose angle.
Claim 49
Independent Claim
219 words
A system for determining a pose angle of an object in an input image in at least one dimension, comprising: a first neural network using a first plurality of training images to develop a plurality of weight vectors, each weight vector representing a plurality of the training images; a projector adapted to receive an input image and generate a projection vector comprising a plurality of calculated distances representing distances between the input image and the weight vectors; a second neural network trained in a supervised mode with a second plurality of training images, the second plurality of training images comprising objects at known pose angles, the second neural network comprising a plurality of neurons, each neuron tuned to a particular pose angle and adapted to receive the projection vector and output a value in accordance with a function that is even and monotonically decreasing for changes in the pose angle of the object from the pose angle to which each neuron is tuned; a curve fitter adapted to receive the values output by the neurons and estimate the pose angle of the object in at least one dimension; and an error calculator that generates a value representative of the error between the values output by the neurons and the calculated values for the function using the estimated pose angle.
Claim 50
Dependent Claim
21 words
The system of claim 46 , wherein the first neural network is trained in a supervised mode or an unsupervised mode.
Claim 51
Independent Claim
214 words
A method for determining a pose angle of an object in an input image in at least one dimension, comprising: training a first neural network using a first plurality of training images to develop a plurality of weight vectors, each weight vector representing a plurality of the training images; projecting the input image onto the first neural network to generate a projection vector comprising a plurality of calculated distances representing distances between the input image and the weight vectors; training a second neural network in a supervised mode with a second plurality of training images, the second plurality of training images comprising objects at known pose angles, the second neural network comprising a plurality of neurons, each neuron tuned to a particular pose angle; processing the projection vector in the second neural network in accordance with a function that is even and monotonically decreasing for changes in the pose angle of the object from the pose angle to which each neuron is tuned to generate a plurality of output values; fitting the output values to a curve to estimate the pose angle of the object in at least one dimension; and calculating a value representative of the error between the output values and the calculated values for the function using the estimated pose angle.
Claim 52
Independent Claim
214 words
A method for determining a pose angle of an object in an input image in at least one dimension, comprising: training a first neural network using a first plurality of training images to develop a plurality of weight vectors, each weight vector representing a plurality of the training images; projecting the input image onto the first neural network to generate a projection vector comprising a plurality of calculated distances representing distances between the input image and the weight vectors; training a second neural network in a supervised mode with a second plurality of training images, the second plurality of training images comprising objects at known pose angles, the second neural network comprising a plurality of neurons, each neuron tuned to a particular pose angle; processing the projection vector in the second neural network in accordance with a function that is even and monotonically decreasing for changes in the pose angle of the object from the pose angle to which each neuron is tuned to generate a plurality of output values; fitting the output values to a curve to estimate the pose angle of the object in at least one dimension; and calculating a value representative of the error between the output values and the calculated values for the function using the estimated pose angle.
Claim 53
Dependent Claim
21 words
The method of claim 48 , wherein the first neural network is trained in a supervised mode or an unsupervised mode.
References Cited (Backward Citations)
US5239594A
Cited
US5642431A
Cited
US5850470A
Cited
US6144755A
Cited
US6128397A
Cited
US6429699B1
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2008-10-24 | 2009-04-25 | 2009-10-25 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2012-10-24 | 2013-04-25 | 2013-10-25 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2016-10-24 | 2017-04-25 | 2017-10-25 | $7700.00 | Window Closed / Lapsed |