Vision-based remote meter reading system in transformer stations

Remote sensing techniques are vital to replacing the human workers in dangerous environments. This paper presents a real-time remote circular meter reading system, which is equipped on our CIMR platform (Cruising and Inspecting Mobile Robot) for a transformer station environment. The tasks of malfunction detection and daily monitoring are carried out by a mobile robot, which cruises along a dynamically planned route and carries an onboard camera to capture Electro-Optic/Infrared (EO/IR) images and read meters in the transformer environment. With the wireless communication, the control center makes global planning for, and collects the feedback from, the inspecting robot. Due to a localizing error in navigation, the robot needs to detect the geometric shape, specifically circles or ellipses, of the designated meter in a large search space. This challenges on most existing circle/ellipse detection algorithms. In this paper, we introduce a constraint on continuity of the contour to divide the circle/ellipse detection into some smaller subproblems which can efficiently deal with multiple circles/ellipses. Moreover, we use the historical scene information as heuristic information to guide the searching procedure. This system now runs successfully in a 500KV transformer station, which demonstrates its real-time performance and high reliability.