Image analysis-based automatic detection of transmission towers using aerial imagery

Abstract

Uninterrupted electricity transmission is a critical utility service for any nation. A major component of nation-wide infrastructure carrying electricity are the transmission towers. To give uninterrupted supply, timely maintenance of towers is a must. Due to vastness of power grid, fault detection via aerial inspection and imaging is emerging as a popular method. In this paper, we attend to the problem of automatic detection of towers in specific images. We present a four-stage algorithm for such detection. For a porous, cage like object structure that of a tower, we use gradient density and a novel feature called cluster density to detect pylon blocks. The algorithm was tested against image data captured for many towers along two different power grid corridors. The algorithm demonstrated missed detection of < 1% and complete absence of false positives, which is very encouraging. We believe that our result is far more useful in tower detection, than available previous works.