Received: 7 September 2016 Revised: 6 March 2017 Accepted: 9 March 2017 DOI: 10.1002/rob.21720 REGULAR ARTICLE The nested k-means method: A new approach for detecting lost persons in aerial images acquired by unmanned aerial vehicles Tomasz Niedzielski Mirosława Jurecka Magdalena Stec Małgorzata Wieczorek Bartłomiej Miziński Department of Geoinformatics and Cartography, Faculty of Earth Sciences and Environmental Management, University of Wrocław, Wrocław, Poland Correspondence Tomasz Niedzielski, Department of Geoinforma- tics and Cartography, Faculty of Earth Sciences and Environmental Management, University of Wrocław, pl. Uniwersytecki 1, 50-137 Wrocław, Poland. Email: tomasz.niedzielski@uwr.edu.pl Abstract A new method, named as the nested k-means, for detecting a person captured in aerial images acquired by an unmanned aerial vehicle (UAV), is presented. The nested k-means method is used in a newly built system that supports search and rescue (SAR) activities through processing of aerial photographs taken in visible light spectra (red-green-blue channels, RGB). First, the k-means clas- sification is utilized to identify clusters of colors in a three-dimensional space (RGB). Second, the k-means method is used to verify if the automatically selected class of colors is concurrently spa- tially clustered in a two-dimensional space (easting-northing, EN), and has human-size area. The UAV images were acquired during the field campaign carried out in the Izerskie Mountains (SW Poland). The experiment aimed to observe several persons using an RGB camera, in spring and winter, during various periods of day, in uncovered terrain and sparse forest. It was found that the nested k-means method has a considerable potential for detecting a person lost in the wilder- ness and allows to reduce area to be searched to 4.4 and 7.3% in spring and winter, respectively. In winter, land cover influences the performance of the nested k-means method, with better skills in sparse forest than in the uncovered terrain. In spring, such a relationship does not hold. The nested k-means method may provide the SAR teams with a tool for near real-time detection of a person and, as a consequence, to reduce search area to approximately 0.5–7.3% of total terrain to be visited, depending on season and land cover. KEYWORDS nested k-means, search and rescue (SAR), statistics, unmanned aerial vehicle (UAV), wilderness search and rescue (WiSAR) 1 INTRODUCTION Recent advances in search and rescue (SAR) activities are associated with geographic information systems (GIS) and the related operational systems such as the SARPlan 1 or SAR Map 2 as well as with new platforms for on-demand data acquisition, the examples of which are unmanned aerial vehicles (UAVs), informally known as drones. There are numerous examples of how UAVs can be used to assist SAR mis- sions. They include: use of airborne video and thermal cameras to iden- tify human shapes, 3–5 use of low quality aerial imagery in the human identity recognition, 6 person’s tracking, 7 processing UAV-acquired images being combinations of a few channels of spectrum, 8–10 applica- tion of synthetic aperture radar for target detection, 11,12 transporta- tion of first need goods to victims. 3,13 Remotely piloted aerial systems, which are nowadays commonly used by geoscientists, become powerful tools for supporting medical activities in a sense that they may reduce search time. Since the time remains a critical issue in SAR missions, 14,15 the use of popular UAVs in a near real-time fashion may have a significant impact on a probability of surviving. Indeed, survival time – which is controlled by many fac- tors such as air temperature, wind and clothing 16,17 —may drop below 24 h if a worn lost person is exposed to low temperatures (less than –20˚C). Finding a lost person alive, particularly in the wintery condi- tions, requires rapid implementation of SAR activities. Therefore, UAVs may serve as efficient tools for on-demand observation of terrain and may reduce search time and increase a probability of surviving. A key issue in the time reduction is associated with on-demand access to drones which may be limited for SAR purposes. There are several attempts to overcome this problem, and in this context a note should be given on the volunteer network known as Search With Aerial Rc Multi-rotor (SWARM) or SARDrones (sardrones.org) which pro- vides access to UAVs for SAR activities. The SWARM is a worldwide J Field Robotics. 2017;1–12. c  2017 Wiley Periodicals, Inc. 1 wileyonlinelibrary.com/journal/rob