Aerial image recognition and matching for inspection of large-scale photovoltaic farms

With the rapid development of solar photovoltaic power generation, inspection for PV plants based on UAV platforms has become prevalent. Despite the obvious advantages in efficiency, cost, accuracy and so forth, a new challenge should be noticed that how to match PV strings between images correctly in UAV images' processing when detecting the failures in solar panels or inspecting installed capacity of a PV farm. Conventional image matching algorithms are usually invalid under these circumstances. This paper proposed a new method to solve this problem by recognizing PV modules and structuring an appropriate feature description vector according to PV modules' different colors. Also, median filtering and morphological method are utilized in image pre-processing and recognition. By comparing feature description vectors of two PV strings in different images respectively, it can be confirmed whether they are the same one in reality. Experiments clearly demonstrated the efficiency and accuracy of the proposed method, which makes a contribution to PV strings' image processing in inspection for large-scale PV farms.