For rotorcraft in a nap-of-the earth (NOE) flight regime, collision with wires presents a significant threat to safety. Wire detection and avoidance is thus an important part of safe NOE guidance. Wire detection can be difficult because: wires are often not strongly spatially evident, dangerous wires often occur at the focus-of-expansion (FOE), there is often significant background clutter and noise, and the detection task is on-going and must be fast enough to permit evasive maneuvers. This effort examines innovative wire detection and tracking techniques that address these issues. Wires are treated as moving edges in densely sampled imagery, and these edges are then traced using local computational mechanisms that are amenable to fully parallel implementation with VLSI. Tracker control methods will be explored which integrate numeric information from known parameters of camera motion produced by gyros with symbolic and rule-based information based upon aggregates of image features in order to improve tracking and compute motion properties. These tracked edges provide important descriptions of moving wires (e.g., measures of stability and motion) which can be used for avoidance and guidance.An automated wire avoidance system has direct applicability for commercial air traffic safety. This project also provides a basis for developing advanced vision guided robotic capabilities. vision-guided navigation, wire detection and avoidance, moving edge detection, tracking, parallel vision algorithmsSTATUS: Phase I Only
