The system generates reports and alarms about detected problems and sends them to a control centre. The system uses visual and thermal video camera data. The visual data are used to detect vegetation and buildings close to power lines and calculate the distance between them and the conductors.
Yan et al. (2007a) detected 20.6 km of the total of 22.5 km of power line. However, they did not report the amount of false detections. Li et al. (2010a) tested their algorithm with more than 4000 images, found all conductors from the images and achieved a correctness (also known as “precision” or “user’s accuracy”) of 99.7%.
With the accumulation of multi-type UAV remote sensing data, a huge amount of powerline observations provide a good foundation for applying deep learning technologies with the goal of achieving high-accuracy automatic fault detection in intelligent inspections, which has become a current research hotspot , , .
Liu et al. (2009) used intensity data and the Hough transform to detect conductors. Zhu and Hyyppä (2014) proposed a method that first identified candidate power line points by using statistical analysis of the point cloud and then refined the results by image-based processing that analysed the 2D geometric properties of objects.
Research literature clearly concentrates on the development of automated methods for power line monitoring applications. Visual and manual inspection methods or semi-automated approaches are not often studied in articles, although they are probably still common in practical work ( Pulkkinen, 2015 ).
Larrauri et al. (2013) described a system called “RELIFO” for automatic and almost real time inspection of power lines from UAV data. The article concentrates on presenting the functionality of the system and algorithms developed for detecting power lines, vegetation and buildings and calculating distances between them.