Data Fusion: Cumulative Effects of Discrete Fusion on Target Detection Probability

This paper describes the culmination of a four-year research, application, and development program towards finding and quantifying a methodology for sensor and data fusion of remotely sensed targets. It builds on previous research reported in IGARSS'02 [1] and IGARSS'04 [2]. Here, we examine the effectiveness of the data fusion methodology, specifically, the impact of iterative data collection on the effective probability of detection of targets. Through statistical (Bayesian) combination of sensor iterations, low confidence sensors (those with moderate probability of detection and moderately low probability of false alarms) can provide high detection performance. This can be extended to multiple data source types. The goal is to make use of higher coverage data products which have only moderate detection performance in the detection and tracking of targets. This is made possible through a combination of discrete target data, along with the analysis of parameters from the respective remote sensing technologies. This process requires the existence of a reasonable maneuvering model or sufficient processing resources.