Three-dimensional intra- and interframe target detection in IR/mmW image sequences for missile seekers

This paper describes a model-based method for the automatic recognition of high value targets in multisensor data. A production net is used to represent the knowledge about target structures. The analysis is carried out by a blackboard-based production system with a database stored in an associative memory. The efficiency of the analysis system is illustrated by an example involving the detection of bridges. For this, sensor data has been interpreted which were recorded with an experimental dualmode sensor (IIR, mmW) mounted to a helicopter while flying over the scenario. Image sequences are taken in oblique view with high frequency. The analysis starts with an intra-frame process by extracting cues in the actual sensor data and by the combination of orthogonal information of IR (intensity, direction) and radar (RCS, range) data to estimate target location in space. To improve detection and reduce false alarms, inter-frame processing is applied to exploit intra- frame results of overlapping images of the scanning sensor system resulting in a higher confirmation of the real target and a better discrimination of false alarms.