Mark R. Stevens

A histogram-based color consistency test for voxel coloring

Voxel coloring has become a popular technique for reconstructing a 3D scene from a set of 2D images. While many different variants of this technique exist, all rely on a test to determine if each voxel is projecting to regions of consistent color in all views of that voxel. A number of color consistency tests can be used and the specific choice has a large influence on the quality of the reconstruction. Earlier work has used variance or the L/sub 1/ norm. We propose a new form of consistency test based on histograms that: is more robust at reconstructing textured surfaces, 2) deals properly with RGB color information, and 3) does not require fine tuning of parameters.

Integrating Graphics and Vision for Object Recognition

List of Figures. List of Tables. 1. Introduction. 2. Previous Work. 3. Render: Predicting Scenes. 4. Match: Comparing Images. 5. Refine: Iterative Search. 6. Evaluation. 7. Conclusions. Appendices. Index.

Interleaving 3D model feature prediction and matching to support multi-sensor object recognition

The object recognition, system presented combines on-line feature prediction with an iterative multisensor matching algorithm. Matching begins with an initial object type and pose hypothesis. An iterative generate-and-test procedure then refines the pose as well as the sensor-to-sensor registration for separate range and electro optical sensors. During matching, object features predicted to be visible are updated to reflect changes in hypothesized object pose and sensor registration. The match found is locally optimal in terms of the complete space of possible matches and globally consistent in the sense of preserving the 3D constraints implied by sensor and object geometry. Results on real data are presented which demonstrate the algorithm correcting for up to 30/spl deg/ errors in initial orientation and 5 m errors in initial translation.

Three-dimensional visualization environment for multisensor data analysis, interpretation, and model-based object recognition

Model-based object recognition must solve three-dimensional geometric problems involving the registration of multiple sensors and the spatial relationship of a three-dimensional model to the sensors. Observation and verification of the registration and recognition processes requires display of these geometric relationships. We have developed a prototype software system which allows a user to interact with the sensor data and model matching system in a three- dimensional environment. This visualization environment combines range imagery, color imagery, thermal (infrared) imagery, and CAD models of objects to be recognized. We are currently using imagery of vehicles travelling off-road (a challenging environment for the object recognizer). Range imagery is used to create a partial three-dimensional representation of a scene. Optical imagery is mapped onto this partial 3D representation. Visualization allows monitoring of the recognizer as it solves for the type and position of the object. The object is rendered from its associated CAD model. In addition to its usefulness in development of the object recognizer, we foresee eventual use of this technology in a fielded system for operator verification of automatic target recognition results.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Automatic target detection using PMMW and LADAR imagery

The need for air-to-ground missiles with autonomous target acquisition (ATA) seekers is in large part driven by the failure of pilot-guided bombs in cloudy conditions (such as demonstrated in Kosovo). Passive-millimeter wave (PMMW) sensors have the ability to see through clouds; in fact, they tend to show metallic objects (such as mobile ground targets) in high contrast regardless of weather conditions. However, their resolution is very low when compared with other popular ATA sensors such as laser-radar (LADAR). We present an A TA algorithm suite that combines the superior target detection potential of PMMW with the high-quality segmentation and recognition abilities of LADAR. Preliminary detection and segmentation results are presented for a set of image-pairs of military vehicles that were collected for this project using an 89GHz, 18 aperture PMMW sensor and a 1.06 /spl mu/ very-high-resolution LADAR.

Multisensor occlusion reasoning

Most model-based object recognition algorithms attempt to match stored model features to features extracted from imagery. The better the match, the more likely it is that the object is present in the scene. Problems arise when objects are occluded because matches will be incomplete. It is rare for an object recognition algorithm to employ knowledge to explain the absence of occluded features. The work presented here illustrates an approach to object recognition which propagates evidence of occlusion from range to optical sensors and thereby explains missing features.