Pattern recognition in a multi-sensor environment

Current pattern recognition systems tend to operate on a single sensor, e.g., a camera. however. the need is now evident for pattern recognition systems which can operate in multi-sensor environments. For example, a robotics workstation may use range finders. cameras, tactile pads, etc. The Multi-sensor Kernel System (MKS) provides an efficient and coherent approach to the specification, recovery, and analysis of patterns in the data sensed by such a diverse set of sensors. We demonstrate how much a system can be used to support both feature-based object models as well as structural models. The problems solved is the localization of a three-dimensional object in 3-space. Moreover, MKS allows rapid reconfiguration of the available sensors and the high-level models.