Traditional inspection methods of pollution deposition on ceramic insulators are impractical and are hence avoided by utilities due to the need for extended power system outage. This paper presents a novel method of pollution level estimation that can be performed on live power lines with the aid of image processing and pattern recognition. The proposed solution is tested under both fixed and variable testing conditions with four different classifiers; all yielding very high classification accuracies. Thereafter, the resilience of the system to a variation of operating conditions is demonstrated. © 2018 IEEE.