Unified 3D Models for Multisensor Image Synthesis

Abstract

A method for modeling three-dimensional objects for multi-sensory images and feature prediction is presented. The object representation is a modified volume-surface octree created using volume intersection based on the object's silhouettes. This representation is easy to construct and is shown to be well adapted for the simulations of the physical processes that affect the generation of the different imagery. A novel technique is developed that eliminates false volumes in the model created during the process of backprojecting the silhouettes in the volume intersection algorithm. This technique is shown to be valid for the important class of objects defined by simple engineering drawings. The elimination of false volumes increases the accuracy of the volumetric fit of the simulated object to the modeled object. The improved model supports the generation of realistic imagery in three imaging modes: video, thermal, and laser radar. Thermal image simulation based on the modeling of complex objects with nonhomogeneities and time-varying heat generation is described. A computationally efficient, implicit finite difference method is described for the simulation of energy flow within objects and between the object surface and the environment, The prediction of laser radar (ladar) images uses a statistical model to predict speckle noise, which allows the synthesis of realistic imagery. Examples of the multisensory imagery produced by this scheme are presented. The imagery and features values predicted by this unified modeling scheme may be used to train multisensorbased object recognition systems.