Topological anomaly detection performance with multispectral polarimetric imagery

Polarimetric imaging has demonstrated utility for increasing contrast of manmade targets above natural back- ground clutter. Manual detection of manmade targets in multispectral polarimetric imagery can be challenging and a subjective process for large datasets. Analyst exploitation may be improved utilizing conventional anomaly detection algorithms such as RX. In this paper we examine the performance of a relatively new approach to anomaly detection, which leverages topology theory, applied to spectral polarimetric imagery. Detection results for manmade targets embedded in a complex natural background will be presented for both the RX and Topolog- ical Anomaly Detection (TAD) approaches. We will also present detailed results examining detection sensitivities relative to: (1) the number of spectral bands, (2) utilization of Stoke's images versus intensity images, and (3) airborne versus spaceborne measurements. Whether utilizing existing capability or designing new capability, the question of imaging targets with either MS, PI, or both is quite relevant. The answer to the question is obviously dependent on target material properties, acquisition geometry constraints, and background properties. In this paper we attempt to examine this problem in the context of trying to locate man-made targets in a background of dense foliage and shadows. More specifically, painted calibration panels and missile launchers are placed in various levels cover within the context of a suburban scene. We chose the Digital Imaging and Remote Sensing Image Generation (DIRSIG) model as a means to simulate remotely sensed spectral-polarimetric imagery from a variety of altitudes to perform a series of trade studies. In an attempt to autonomously find the targets of interest in simulated spectral and/or polarimetric im- agery we utilized the RX1 and TAD2 anomaly detection algorithms. The TAD algorithm is a new approach to anomaly detection and has been shown to exceed the performance of legacy algorithms such as RX on hyper- spectral datasets.2 Through a series of trade studies, we examine the utility of unpolarized spectral, broadband polarimetric, and spectral polarimetric datasets of the same scenario utilizing both the TAD and RX anomaly detection algorithms. In addition, we examine the utility of performing anomaly detection on both the Stoke's polarimetric bands and the lower-level intensity sensed bands (the intensity bands are typically what is directly sensed, while the Stoke's images are merely a linear combination of these).