No abstract

No abstract

Enhancement of undersampled imager performance has been demonstrated using superresolution techniques. In these techniques, the optical flow of the scene or the relative subpixel shifts among various snapshots of the scene are calculated, and a high-resolution grid is populated with spatial data using various algorithms. Performance enhancement has been demonstrated for the case of a static image with the undersampled imager compared with a static image that has been acquired through a frame series in a dynamic scene. In this research, the performance is compared for four cases: static image with undersampled imager, static image with superresolution frame sequence, dynamic image with undersampled imager, and dynamic image with superresolution frame sequence.

Sensors and Electron Devices Directorate, ARLApproved for public release; distribution unlimited. ii REPORT DOCUMENTATION PAGE REPORT DATE (DD-MM-YYYY)September 2005 ARL-TR-3639 SPONSOR/MONITOR'S ACRONYM(S) 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES)ARL 2800 Powder Mill Road Adelphi, MD 20783-1145 SPONSOR/MONITOR'S REPORT NUMBER(S) DISTRIBUTION/AVAILABILITY STATEMENTApproved for public release; distribution unlimited. SUPPLEMENTARY NOTES ABSTRACTRecently, object detection based on hyperspectral sensors in support of autonomous robotics navigation has been of great interest. Hyperpspectral sensors have been widely used for automatic target detection in military applications, mainly because a wealth of spectral information can be obtained through a large number of narrow contiguous spectral channels (often over a hundred). The main purpose of this report is to present detection techniques based on hyperspectral sensing that can effectively identify potentially harmful objects to UGV navigation. The hyperspectral detection techniques used are built on the basic premise that the spectral signatures of objects of interest are in general different than background materials, and the objects of interest can be identified from their surrounding background materials based on spectral analysis of the hyperspectral data. In this report, we first present detailed information on two hyperspectral sensors-a dual band hyperspectral imager and an acousto-optic tunable filter imager-that provide hyperspectral data in the infrared and visible bands, respectively. Several anomaly detection and classification techniques newly developed by ARL are then introduced and applied to the hyperspectral data to identify potential obstacles to robotics navigation. Detection performance for each technique is included in this report.