Modeling sensor detectability with the VANTAGE geometric/sensor modeler

This paper describes a method of modeling sensor detectabilty and its implementation in the VANTAGE geometric/sensor modeler. Each sensor consists of one or more light sources and TV cameras, which illuminate and observe, respectively, various portions of faces of an object Sensor detectability, which specifies where the sensor can "see" (detect), can be described by a result of AND/OR operations between illuminated and cameraobservable portions of 3D faces. In order to represent this sensor detectability, first we will define a G-source, an abstract sensor component, representing either a light source or a TV camera. Then, we will investigate G-source illumination conditions to describe under what condition each G-source illuminates (or observes) surface portions with respect to its illumination direction. We will show that the sensor detectability can be decomposed into its component G-sources' illumination conditions and AND/OR operations between them. Based on this findings, we will propose the sensor composition tree. The tree consists of G-source illumination conditions as its leaf nodes and set operations which indicate a combination of the component G-source illumination conditions, as its branch nodes. We will also propose several 2D structures to represent detected 2D appearances given by VANTAGE using a sensor composition tree. Finally, we will show how to use these capabilities in model-based vision.