Parallel Matching of 3d Articulated Object Recognition

Abstract

In dealing with large volume image data, sequential methods usually are too slow and unsatisfactory. This paper introduces a new system employing parallel matching in high-level recognition of 3D articulated objects. A new structural strategy using linear combination and parallel graphic matching techniques is presented for 3D polyhedral objects representable by 2D line-drawings. It solves one of the basic concerns in diffusion tomography complexities, i.e. patterns can be reconstructed through fewer projections, and 3D objects can be recognized by a few learning sample views. It also improves some of the current methods while overcoming their drawbacks. Furthermore, it can distinguish very similar objects and is more accurate than other methods in the literature. An online webpage system for understanding and recognizing 3D objects is also illustrated.

Probabilistic Decisions in Production Net An Example from Vehicle Recognition

“…In contrast to this an aspect is a line graph. Changes in the view that don't change the topology provide the same aspect [11].…”

“…Such a model is described by a directed graph where each basic part is a polyhedron. If the parts have mutual degrees of freedom in rotation such a model is called articulated model [11]. The resulting constraints are used by recognition process.…”

“…truck and trailer systems or tanks). Such objects can be captured by articulated models [11]. The appearance of polyhedrons is affected by self occlusion.…”

“…The appearance of polyhedrons is affected by self occlusion. This may be treated by aspect graphs [4], or by linear combination of characteristic views [11]. We use an equidistantly sampled set of views for each model [8].…”

Assessing the Computational Effort for Structural 3D Vehicle Recognition

“…use articulated 3D models. The projection may also be included in the geometric model, so that finally 2D views -or linear combinations of these -are matched like in [16]. Such modeling may be used, if the camera is directly approaching the target object.…”