Toward Automated Utility Pole Condition Monitoring: A Deep Learning Approach

As infrastructure ages, grid operators across the world are becoming more cognizant of the need for monitoring their transmission infrastructure. The geographic extent of the transmission system, however, makes this a difficult and expensive task with thousands of components requiring visual inspection to identify faults which can lead to potentially catastrophic failures. This paper describes the use of Deep Neural Networks to automatically detect areas of concrete damage on utility poles in a European utility from photographs. This is beneficial to reduce time spent in the field as well as variability in between human assessment. The results show that even with a small dataset for training, the network is able to identify new damage with a high level of precision.