Automatic Extraction of Objects/Features from Imagery of Electrical Substations

Abstract

Substations are among the most important components of electrical transmission and distribution systems in the world. Unfortunately, some electrical equipment in substations is suitable place for animals to perch, roost, and hunt leading to their electrocution and possible power outages. Researchers have proposed different solutions to protect these animals from electrocution such as covering the electrical equipment. This solution requires field work to determine the dimensions of electrical equipment within high voltage environments which is potentially dangerous. Remote sensing methods can provide a suitable means to obtain measurements while mitigating such danger, because the field worker can stay a safe distance away from hazardous equipment such as power lines and hot connections. In this work, analysis of images taken from electrical substations is used to obtain measurements of equipment of interest. The main idea of this work is that hot connections are placed somewhere between insulators and power lines or insulators and bus pipes. Thus, if centerlines of insulators and power lines are extracted, the position of the hot connections will be known. Therefore, different feature-based algorithms are developed to extract insulators and power lines which will help in extracting hot connections. These algorithms include: 1) an automated texture/edge-based algorithm for insulator extraction, 2) an automatic edge-based power line extraction, and 3) a semi-automatic snake-based power line extraction. Then, the main objective of this thesis is to extract insulators and power lines from imagery of electrical substations. The extraction of bus pipes is not directly addressed; however, the power line extraction algorithms can be used to extract some of bus pipes, too.