Unified modeling of non-homogeneous 3D objects for thermal and visual image synthesis

Abstract

This paper presents a new approach to integrated modeling of 3D objects for generating multi-sensory images. Specifically, the model supports the generation of visual and thermal images under different viewing, ambient and internal conditions. The model also predicts the values of features that are used by multisensory image interpretation systems. A representation known as the volume surface octree is used for object modeling. It is shown that the above representation is well suited for thermal modeling of complex objects with non-homogeneities and heat generation. An improved volume surface octree construction technique is presented. This technique increases storage efficiency without degrading the quality of image synthesis. A technique to incorporate non-homogeneities and heat generation using octree intersection is described. A computationally efficient, implicit finite difference method is described for heat flow simulation of objects with large number of octree nodes and with non-homogeneities. Several examples of the generated thermal and visual images are presented in this paper, which illustrate the usefulness of the approach. The model is designed to be used in a multi-sensor vision system that would use the images and features predicted by this model in a hypothesize-and-verify scheme.