Analysis of the detection modes of a human presence detection millimeter-wave radiometer

Detection of non-moving humans is important for physical security, search-and-rescue, homeland protection, and military applications, although it has proven difficult to achieve because traditional techniques have significant drawbacks. For instance, IR sensors do not detect well in warm daytime environments and radars cannot easily discriminate stationary humans. The authors have recently presented a solution by applying millimeter-wave radiometry to the detection of stationary humans in cluttered outdoor environments. MMW radiometers operate in both day and night environments and are affected only by moderate to heavy rain. Fog, smoke, or dust, which render IR sensors inoperable, are also transparent to MMW sensors. The radiometric detection system was designed to operate from a moving platform, making image recognition prohibitive due to the high computational expense of image processing in real-time with a constantly changing background. The presence-detection radiometer operates on onedimensional sensor signals, resulting in lower computation time. The system uses four detection modes: two Ka-band total power radiometers (indicated by tp₁ and tp₂), one Ka-band correlation radiometer (cr), and one W-band total power radiometer (tp_w). The authors´ previous work has demonstrated the feasibility of the detection method and introduced detection algorithms, and a future publication will present the real-time performance characteristics of the method. In this paper an analysis of the detection capabilities of the sensor system is presented by selectively removing sensor modes from the detection algorithms. In this way the dependence of the individual modes on the overall system performance is determined. The performance is analyzed using the receiver operating characteristic.