US5459636A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Position and orientation estimation neural network system and method
Assignee: Raytheon Co
•
Filing: 1994-01-14
•
Granted: 1995-10-17
•
Expires: 2014-01-14
Abstract
Abstract Disclosed are a system and method for determining the pose (translation, rotation, and scale), or position and orientation, of a model object that best matches a target object located in image data. Through an iterative process small adjustments are made to the original position and orientation of the model object until it converges to a state that best matches the target object contained in the image data. Edge data representative of edges of the target object and edge data representative of the model object are processed for each data point in the model object relative to each point in the target object to produce a set of minimum distance vectors between the model object and the target object. A neural network estimates translation, rotation, and scaling adjustments that are to be made to the model object. Pose of the model object is adjusted relative to the target object based upon the estimated translation, rotation, and scaling adjustments provided by the neural network. Iterative calculation of the minimum distance vectors, estimation of the translation, rotation, and scaling adjustments, and adjustment of the position and orientation of the model object is adapted to reposition the model object until it substantially overlays the target object. Final position of the model object provides an estimate of the position and orientation of the target object in the digitized image.
Inventors
- 👤 Allen Gee
- 👤 David M. Doria
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-08 18:07
CPC Classifications
G; G06; G06T; G06T7/00; G06T7/70; G06T7/73; G06T7/75; G06V; G06V10/00; G06V10/70; G06V10/74; G06V10/75; G06V10/751; G06V10/7515
Claims Analysis
Total Claims: 16
Claim 1
Independent Claim
233 words
A system for determining the position and orientation of a target object in a digitized image, said system comprising: an edge detector for processing the digitized image to produce edge data representative of edges of the target object; model storage means for storing model data corresponding to a model object of the target object that is to be detected, which model data comprises a plurality of data points disposed along edges of the model object; minimum distance vector calculation means for processing the edge data and the model data to produce a set of minimum distance vectors between the target object and the model object; a neural network, responsive to the minimum distance vector calculation means, for estimating distortions of the model object relative to the target object; and position and orientation adjustment means, responsive to the neural network and the model storage means, for adjusting the position and orientation of the model object relative to the target object based upon the distortions estimated by the neural network; wherein iterative calculation of the minimum distance vectors, estimation of the distortions, and adjustment of the position of the model object is adapted to reposition and reorient the model object until the model object substantially overlays the target object, whereby the final position and orientation of the model object provides an estimate of the position and orientation of the target object in the digitized image.
Claim 2
Independent Claim
233 words
A system for determining the position and orientation of a target object in a digitized image, said system comprising: an edge detector for processing the digitized image to produce edge data representative of edges of the target object; model storage means for storing model data corresponding to a model object of the target object that is to be detected, which model data comprises a plurality of data points disposed along edges of the model object; minimum distance vector calculation means for processing the edge data and the model data to produce a set of minimum distance vectors between the target object and the model object; a neural network, responsive to the minimum distance vector calculation means, for estimating distortions of the model object relative to the target object; and position and orientation adjustment means, responsive to the neural network and the model storage means, for adjusting the position and orientation of the model object relative to the target object based upon the distortions estimated by the neural network; wherein iterative calculation of the minimum distance vectors, estimation of the distortions, and adjustment of the position of the model object is adapted to reposition and reorient the model object until the model object substantially overlays the target object, whereby the final position and orientation of the model object provides an estimate of the position and orientation of the target object in the digitized image.
Claim 3
Dependent Claim
20 words
The system of claim 1 wherein the neural network comprises a three layer perceptron trained by a back-propagation learning algorithm.
Claim 4
Dependent Claim
34 words
The system of claim 1 wherein the neural network comprises 2 n input neurons, fifty hidden neurons, and four output neurons, where n is the number of model points in the sampled contour model.
Claim 5
Dependent Claim
25 words
The system of claim 3 wherein each output of the output neurons is interpreted in a trinary manner by the position and orientation adjustment means.
Claim 6
Independent Claim
221 words
A method for determining the position and orientation of a target object in a digitized image, the digitized image being represented by an electrical signal, said method comprising the steps of: processing the signal of the digitized image to produce edge data representative of edges of the target object; storing a model object of the target object that is to be detected, which model object comprises a plurality of data points disposed along edges of the model object; processing the edge data and the model data for each data point in the model object relative to each point in the target object to produce a set of minimum distance vectors between the model object and the target object; processing the set of minimum distance vectors using a neural network to estimate translation, rotation, and scaling adjustments that are to be made to the model object that are adapted to translate, rotate, and scale the model object relative to the target object; and adjusting the position and orientation of the model object relative to the target object based upon the estimated translation, rotation, and scaling adjustments; iterating the last three steps to successively reposition and reorient the model object until it substantially overlays the object, thus providing an estimate of the position and orientation of the target object in the digitized image.
Claim 7
Independent Claim
221 words
A method for determining the position and orientation of a target object in a digitized image, the digitized image being represented by an electrical signal, said method comprising the steps of: processing the signal of the digitized image to produce edge data representative of edges of the target object; storing a model object of the target object that is to be detected, which model object comprises a plurality of data points disposed along edges of the model object; processing the edge data and the model data for each data point in the model object relative to each point in the target object to produce a set of minimum distance vectors between the model object and the target object; processing the set of minimum distance vectors using a neural network to estimate translation, rotation, and scaling adjustments that are to be made to the model object that are adapted to translate, rotate, and scale the model object relative to the target object; and adjusting the position and orientation of the model object relative to the target object based upon the estimated translation, rotation, and scaling adjustments; iterating the last three steps to successively reposition and reorient the model object until it substantially overlays the object, thus providing an estimate of the position and orientation of the target object in the digitized image.
Claim 8
Independent Claim
97 words
A method for determining pose of a target in a digitized image, said method comprising the steps of: generating signals indicating minimum distance vectors between an edge image of a target and an edge image of a model; determining orientation adjustments to the edge image of the model by processing the signals indicating the minimum distance vectors with a neural network; and adjusting the edge image of the model until the edge image of the model substantially overlays the edge image of the target, whereby the final position of the model indicates the pose of the target.
Claim 9
Independent Claim
97 words
A method for determining pose of a target in a digitized image, said method comprising the steps of: generating signals indicating minimum distance vectors between an edge image of a target and an edge image of a model; determining orientation adjustments to the edge image of the model by processing the signals indicating the minimum distance vectors with a neural network; and adjusting the edge image of the model until the edge image of the model substantially overlays the edge image of the target, whereby the final position of the model indicates the pose of the target.
Claim 10
Dependent Claim
34 words
The method of claim 6, wherein the minimum distance vectors are determined by processing the edge data for the target and model relative to each data point of the edge data for the target.
Claim 11
Dependent Claim
19 words
The method of claim 6, wherein the edge image of the model is adjusted for rotation, translation and scale.
Claim 12
Dependent Claim
26 words
The method of claim 6, wherein the edge image of the model is adjusted from an initial position to a final position in an iterative manner.
Claim 13
Dependent Claim
31 words
The method of claim 9, wherein the step of adjusting includes repeating the steps of determining the minimum distance vectors, estimating the adjustments, and adjusting the edge image of the model.
Claim 14
Dependent Claim
18 words
The method of claim 6, wherein the neural network is initially trained with a plurality of different objects.
Claim 15
Dependent Claim
21 words
The method of claim 11, wherein the output of output neurons of the neural network are interpreted in a trinary manner.
Claim 16
Dependent Claim
18 words
The system of claim 1, wherein the distortions estimated by the neural network include translation, rotation and scale.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1998-10-16 | 1999-04-17 | 1999-10-17 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2002-10-16 | 2003-04-17 | 2003-10-17 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2006-10-16 | 2007-04-17 | 2007-10-17 | $7700.00 | Window Closed / Lapsed |