Failure detection of solar panels using thermographic images captured by drone

In this work, we compare several methods for detecting failed solar panels, and found that drone detection is the most efficient. We show thermographic images captured by drone and explain various types of panel failures. An examination by deep learning revealed that single shot detection (SSD) is an effective method of automatic detection. We ran tests with the SSD model using 4-MW solar plant data and achieved a mean average precision of 49.11%.