The nested k‐means method: A new approach for detecting lost persons in aerial images acquired by unmanned aerial vehicles

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 classification 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 spatially 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 wilderness 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.

A real‐time field experiment on search and rescue operations assisted by unmanned aerial vehicles

“…SARUAV is the two‐module system dedicated to support SAR activities targeted at finding missing persons (Jurecka & Niedzielski, ; Niedzielski et al., ). The first module of the system uses GIS tools with the results of studies on lost person behavior (Doherty et al., ; Heth & Cornell, ; Koester, ).…”

“…One of such systems is the Search and Rescue Unmanned Aerial Vehicle (SARUAV) developed at the University of Wrocław (Poland). The fundamentals of the system are the mobility model (Doherty, Guo, Doke, & Ferguson, ) and the ring model (Heth & Cornell, ; Jurecka & Niedzielski, ; Sava et al., ) as well as the nested k ‐means method (Niedzielski, Jurecka, Stec, Wieczorek, & Miziński, ). The SARUAV system may support standard search procedures, which use POA that describes the chance of a missing person to be in a specific area (Koester, ).…”

“…Although the nested k ‐means method, as a part of SARUAV, has been verified and validated using a set of photographs taken by a UAV at different seasons and periods of day (Niedzielski et al., ), the entire SARUAV system has not been tested so far, in particular in a real‐world search scenario, under time pressure, and in severe weather. The objective of this paper is therefore to report on the field experiment carried out on March 2–3, 2017, in the Izerskie Mountains (southwestern Poland) in which the SARUAV system was applied to guide ground searchers in finding persons who pretended to be lost.…”

Detection of bodies in maritime rescue operations using unmanned aerial vehicles with multispectral cameras

“…More specifically, UAVs and multispectral cameras are usually used for search and rescue tasks of lying bodies outdoors (Rudol & Doherty, ) where detection and localization are performed to train rescue routines in gravel roads, asphalt, and grass. Missing and lost people (standing and lying) can also be searched for by emergency services in natural environments such as uncovered terrain and forest (Niedzielski, Jurecka, Miziński et al, ; Niedzielski, Jurecka, Stec, Wieczorek, & Miziński, ), and in desert and chaparral (Coulter et al, ).…”

“…There are some previous works addressing people detection on land or in indoor environments with traditional machine learning methods such as Andriluka et al (). For example, Amanatiadis, Bampis, Karakasis, Gasteratos, and Sirakoulis () used Chevyshev image moments and thresholding techniques, Niedzielski, Jurecka, Stec, et al () used color information and a ‐means method, Avola et al () implemented an red‐green‐blue local binary pattern (RGB‐LBP) descriptor, Blondel et al () and Rudol and Doherty () combined local visual features such as histogram of gradients (HOG) to be used as input of a cascaded Haar classifier, or combined with an support vector machines (SVM) by Portmann et al (), speeded up robust features (SURF) and fast library for approximate nearest neighbors (FLANN) by Symington, Waharte, Julier, and Trigoni () and local image texture was used by Coulter et al ().…”