Image Analysis-Based Automatic Utility Pole Detection for Remote Surveillance

In case of disasters such as cyclones, earthquakes, severe floods etc., widespread damages to infrastructures such as power grid, communication infrastructure etc. is commonplace. Especially to power grid, the damages to various structures are typically spread out in wide areas. Usage of drones to do fast remote survey of damage area is gaining popularity. From the remote surveillance video of any wide disaster area that is fairly long, it is important to extract keyframes that contain specific component structures of the power grid. The keyframes can then be analyzed for possible damage to the specific structure. In this context, we present an algorithm for automated detection of utility poles. Specifically, we show robust detection of poles in frames of videos available from various sources. The detection is performed by first extracting 2D shapes of poles as analytically defined geometric shape, quadrilateral, whose edges exhibit noise corruption. A pole is then detected as a shape-based template, where one long rectangular trapezium, is perpendicularly intersected by at least one trapezium representing a crossarm that suspends the conductors. Via testing and comparison, our algorithm is shown to be more robust as compared to other approaches, especially against highly variable background. We believe such detection, with limited false negatives, will form stepping stone towards future detection of damages in utility poles.