Effective approaches to in-flight hyperspectral target detection for surveillance applications

In this work, Spectral Signature-Based Target Detection (SSBTD) as applied to airborne monitoring for surveillance and reconnaissance of ground targets is addressed, and techniques that can help to approach in-flight processing are analyzed from this perspective. In fact, SSBTD is a challenging task from an operating viewpoint, mainly due to the crucial atmospheric compensation step, which is required to make the target measured reflectance comparable to the sensoracquired radiance. Both physics-based radiative transfer modeling techniques and empirical scene-based methods are considered for atmospheric compensation, and their applicability and adaptability to in-flight processing are discussed. Experimental data acquired by a hyperspectral sensor operating in the Visible Near-InfraRed range are employed for analysis. The data consist in multiple images collected during subsequent flights performed over the same scene. Such a situation well reproduces the typical scenario of regularly monitoring an area of interest, and can, therefore, be adopted for examining the aforementioned approaches from an in-flight applicability perspective. Target detection results are analyzed and discussed by examining objective performance measures such as the Receiver Operating Characteristic (ROC) curves.