Power Line Detection from Mobile Devices

Overhead power lines are a common structure used to transmit and deliver electricity to consumers. The structures, comprised primarily of power poles and power lines, are frequently upgraded, repaired, and maintained by line workers. Often, this maintenance involves making measurements of the structures. Mobile applications such as LineSmarts can be used for this purpose by making measurements through photographs, but are inconvenient since the conductors must be manually identified in the image. Identifying the power lines is tedious, and an automatic method for finding the power lines would greatly improve the usefulness of this application. Power line detection has been well researched, but the literature only covers power line detection from aerial images, and assumes that the power lines are straight. In images taken from ground-level, this assumption is not always fair, as the power lines can have large amounts of curvature due to wire sag. In this thesis, three methods for detecting power lines in ground-based images are implemented and compared using Hough Transform, and Line Tracing techniques. The algorithms showed positive results, with the Line Tracer detecting 76.7% of power lines in the dataset with 2.13 pixel accuracy