We present a general purpose, inherently robust system for object representation and recognition.
The system is model-based and knowledge-based, with knowledge derived from analysis of objects and images, unlike many of the current methods which rely on generic statistical inference. This knowledge is intrinsic to the objects themselves, based on geometric and semantic relations among objects. Therefor the system is insensitive to external interferences such as viewpoint changes (scale, pose etc.), illumination changes, occlusion, shadows, sensor noise etc. It also handles variability in the object itself, e.g.
articulation or camouflage. We represent all available models in a graph containing two independent but interlocking hierarchies. One of these intrinsic hierarchies is based on parts, e.g. a truck has a cabin, a trunk, wheels etc. The other hierarchy we call the "Level of Abstraction (LOA), e.g. a vehicle is more abstract than a truck, a rectangle is more abstract than a door.
This enables us to represent and recognize generic objects just as easily as specific ones. A new algorithm for traversing our graph, combining the advantages of both top-down and bottom-up strategies, has been implemented.
An ATR theory is useful both for developing of ATR algorithms and
evaluating their performance. We present here a model-based ATR theory
based on hierarchies of parts. Objects and parts are represented as
nodes in an attributed graph, while the links between nodes
represent invariant relations between the parts. These can be
either geometric (quantitative) invariants or structural (qualitative)
ones. A metric is used to measure the distance of an object from
known models, based on the distances of the parts. This is different from traditional pixel-based or template-based metrics which are very
sensitive to any variation in the object. Unlike previous graph-based
methods, we do not try to segment the object before recognizing it. Rather, the segmentation is guided by the models and goes hand-in-hand with the recognition process. The theory has been implemented for simple cases.
Object-image relations (O-IRs) provide a powerful approach to performing detection and recognition with laser radar (LADAR) sensors. This paper presents the basics of O-I relations and shows how they are derived from invariants. It also explains and shows results of a computationally efficient approach applying covariants to 3-D LADAR data. The approach is especially appealing because the detection and segmentation processes are integrated with recognition into a robust algorithm. Finally, the method provides a straightforward approach to handling articulation and multi-scale decomposition.
The fundamental problem of smoothing and differentiating of noisy images is an ill-posed problem, and common differentiation filters give very unreliable results. We look at several sources of the errors and show a way to eliminate them. In particular: (1) We show that regularization based filters perform better than the Gaussian, assuming the data changes slowly relative to the noise. (2) Truncation of an infinite filter is very damaging for derivatives, so the common idealized regularization methods cannot be used. We construct finite, discrete regularization based filters using a spline approximation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.