Automatic Extraction of Droppers in Catenary Scenes

The aim of this paper is to present an automatic, image-based system for catenary maintenance, a novel application of machine vision which has no equivalent today. This study focuses on the detection of droppers in catenary staves. The system takes benefit from the fact that dropper location inside catenary staves follows mounting rules, an information that is integrated into a top-down approach in order to speed up and make reliable the extraction of droppers. Results obtained on a large database of real images are very satisfying and enable for continuing investigations for dropper fault detection.