Handover in a distributed system of UAVs: application to wildlife monitoring
| dc.contributor.advisor | Utete, Simukai Wanzira | en_ZA |
| dc.contributor.author | Marcos, Juliana Thomasia Chakirath | en_ZA |
| dc.contributor.other | Stellenbosch University. Faculty of Science. Dept. of Mathematical Sciences. Division Applied Mathematics. | en_ZA |
| dc.date.accessioned | 2020-12-02T14:47:46Z | |
| dc.date.accessioned | 2021-01-31T19:43:13Z | |
| dc.date.available | 2020-12-02T14:47:46Z | |
| dc.date.available | 2021-01-31T19:43:13Z | |
| dc.date.issued | 2020-12 | |
| dc.description | Thesis (MSc)--Stellenbosch University, 2020. | en_ZA |
| dc.description.abstract | ENGLISH ABSTRACT: Wildlife surveillance is of significant interest for the protection of animals and their habitats. In this study, a distributed system of unmanned aerial vehicles (UAVs) or drones is designed for single-animal tracking in terrestrial settings. The system involves four main components which constitute key contributions of the study. The main component is a visual object tracking approach based on the use of a particle filter that switches between measurements from two sources: a simple and fast approach based on colour image segmentation and a slower but more sophisticated method based on a deep learning object detector, the third version of the You Only Look Once detector (YOLOv3). The particle filter switches between the measurement sources using the structural similarity (SSIM) index from the image-processing literature. The SSIM index is also applied in the study for handover of tracking between a pair of drones. Some of the components of the monitoring system have been simulated using wildlife footage recorded by drone (obtained from an animal behaviour group). Extensive simulation tests were carried out during the study. These demonstrate, amongst other results, that better real-time object detection is obtained by replacing YOLOv3 by techniques such as boosting and channel and spatial reliability tracking (CSRT). The design developed and components tested suggest some directions for single-animal tracking by a distributed system of drones. Keywords: Animal tracking algorithm, boosting, channel and spatial reliability tracking (CSRT), drone, handover, multiple instance learning (MIL), particle filter, structural similarity (SSIM), unmanned aerial vehicle (UAV), You Only Look Once version 3 (YOLOv3). | en_ZA |
| dc.description.abstract | AFRIKAANSE OPSOMMING: Die waarneming van wildslewe is van opmerklike belang vir die konservasie van diere en hul habitatte. Hierdie studie maak gebruik van ’n verspreide stelsel wat bestaan uit onbemande lug voertuie (UAVs), of ’drones,’ wat ontwerp is om ’n enkele dier te agtervolg in aardse instellings. Die sisteem bestaan uit vier hoof komponente, en hierdie komponente is die sleutel bydraers vir die studie. Die belangrikste komponent is ’n visuele voorwerpopsporings tegnologie gebaseer op deeltjiefilter wat wissel tussen metings vanaf twee bronne: ’n eenvoudige en vinnige benadering wat gebaseer is op kleurprintsegmentering en ’n stadiger, maar meer gesofistikeerde, metode wat gebaseer is op ’n diep detektor vir leervoorwerpe, die derde instelling van die ’You Only Look Once’ program (YOLOv3). Die deeltjiefilter skakel tussen die meetbronne deur gebruik te maak van die strukturele ooreenkoms (SSIM) indeks uit die beeldverwerkingsliteratuur. Die SSIM-indeks word ook toegepas in die studie vir die oorhandiging van die opsporings proses tussen hommeltuie. Sommige van die komponente van die moniteringstelsel is gesimuleer met behulp van wild beeldmateriaal wat deur hommeltuie geneem is (verkry van ’n dieregedraggroep). Uitgebreide simulasietoetse is tydens die studie uitgevoer. Dit toon onder andere aan dat beter intydse voorwerpopsporing verkry word as YOLOv3 vervang is met tegnieke soos die versterking en opsporing van kanaal- en ruimtelike betroubaarheid (CSRT). Die ontwerp wat ontwikkel is, en die getoetste komponente, dui op ’n paar moontlike aanwysings vir die gebruik van verspreide hommeltuig sisteeme om enkeldiere te monitor. Sleutelterme: Diereopspoor algoritme, Versterking, kanaal- en ruimtelike betroubaarheidsopsporing (CSRT), hommeltuig, oorhandiging, meervoudige instansie leer (MIL), deeltjie filter, Strukturele ooreenkoms (SSIM), onbemande lugvaartuig (UAV), You Only Look Once version 3 (YOLOv3). | af_ZA |
| dc.description.version | Masters | en_ZA |
| dc.format.extent | xxiii, 152 pages : illustrations | en_ZA |
| dc.identifier.uri | http://hdl.handle.net/10019.1/109292 | |
| dc.language.iso | en_ZA | en_ZA |
| dc.publisher | Stellenbosch : Stellenbosch University | en_ZA |
| dc.rights.holder | Stellenbosch University | en_ZA |
| dc.subject | Wildlife monitoring | en_ZA |
| dc.subject | Drone aircraft | en_ZA |
| dc.subject | Roaming (Telecommunication) | en_ZA |
| dc.subject | Handover (Telecommunication) | en_ZA |
| dc.subject | Distributed systems | en_ZA |
| dc.subject | UCTD | |
| dc.title | Handover in a distributed system of UAVs: application to wildlife monitoring | en_ZA |
| dc.type | Thesis | en_ZA |