Title Modeliranje ponašanja i algoritmi za predviđanje kretanja divljih životinja Title (english) Behaviour modelling and algorithms for wild animals movement prediction Author Ivana Nižetić Kosović Mentor Krešimir Fertalj (mentor)Committee member Damir Kalpić (predsjednik povjerenstva)Committee member Zoran Kalafatić (član povjerenstva)Committee member Robert Manger (član povjerenstva)Granter University of Zagreb Defense date and country 2012-01-30, Croatia Scientific / art field, TECHNICAL SCIENCES Universal decimal classificationUDC ) 004 - Computer science and technology. Computing. Data processing Abstract Istraživanje pokretnih objekata je posljednjih desetak godina sve pristupačnije i popularnije. Posebnu klasu pokretnih objekata čine divlje životinje koji se kreću slobodno u prostoru, čije je kretanje kompleksno i uvjetovano raznim vanjskim i unutarnjim čimbenicima, koje se u dosadašnjim radovima uglavnom ne uzima u obzir. Metode analize i predviđanja kretanja, koje su se do sada koristile za druge vrste pokretnih objekata, nisu u izvornom obliku primjenjive na kretanje divljih životinja. U doktorskom radu dan je pregled područja istraživanja pokretnih objekata, s naglaskom na divlje životinje. Predložen je generički model za analizu i predviđanje kretanja divljih životinja temeljen na znanju koji uključuje kontekstualne informacije u predviđanje. Opisan je i formaliziran vlastiti postupak predviđanja kretanja divljih životinja, uvažavajući ekspertno znanje i kontekstualne informacije. Postupak uključuje određivanje geografskog područja u kojem se u nekom vremenskom razdoblju životinja može nalaziti, kao i predviđanje u stvarnom vremenu, kada je potrebno na terenu opaziti te po mogućnosti uhvatiti ili snimiti životinju. Predstavljeni su popratni algoritmi potrebni za uspješno odvijanje postupka predviđanja. U okviru rada, izgrađen je prototip koji podupire algoritme analize i predviđanja kretanja divljih životinja, uz pomoć kojeg je evaluiran predloženi postupak. Validacija je provedena usporedbom rezultata algoritama s terenskim podacima o kretanju vukova i medvjeda u Hrvatskoj.
Abstract (english) Recent wide-ranging usage of Global Positioning System devices and wireless communication devices, together with enhancements of supportive technology, induced the expansion of the research on moving objects. By modelling and analyzing moving objects data, we learn about the moving objects behaviour and even become able to predict their future locations. The most common objects of research are humans and vehicles. The less studied class of moving objects are wild animals which are moving freely in space, whose movement is complex and conditioned by various external and internal factors. The movement analysis and prediction models applied for humans and vehicles are not appropriate for animals, therefore new models should be proposed. In most of the related work, contextual and additional information is not considered into the process of movement analysis and prediction. In the thesis introduction is given an overview of research area of moving objects, with emphasis on wildlife. Wild animals are placed in area of moving objects research with its specificities in order to justify the need of setting up a new approach to movement analysis and prediction for this type of moving objects. The existing systems and algorithms that support some part of the analysis and prediction process are also presented. Later on, we propose a generic model for knowledge representation, analysis and prediction of moving objects movement adapted to the wild animals. The model consists of knowledge base and set of algorithms for the analysis and prediction of animal movement. The proposed solution includes expert knowledge, contextual information, historical movement of animal and animal characteristics in analysis and prediction. The model is based on the automatic process of integration of the identified properties of the movement. Movement prediction process is suggested and described taking into account the expert knowledge and contextual information. The procedure involves determining the geographical area in which animals can reside for some period of time, as well as the prediction of the next movement in real time, when it is necessary to observe or to catch the animal. Supporting algorithms necessary for successful performance of prediction procedure such as significant location identification are presented as well. The prediction process meets the requirement for the possibility of prediction in areas where the animal has not been previously, or where there is no historical data on movement in this area. In order to support the process of movement prediction and to evaluate algorithms for the analysis and prediction of movement of wild animals, we have built a prototype. The results of algorithms have been validated by comparison with field data of the movement of wolves and bears in Croatia. The research of moving objects, especially those moving freely in space, is challenging research area still in development. The results of this study will be useful to community members engaged in studies of wildlife, whether they need to analyse complex spatio-temporal data offline or predict the movement of animal at the terrain research in real time. We believe the same generic model could be applied to different kind of animals but also to other classes of moving objects.
Keywords
predviđanje kretanja divljih životinja
prepoznavanje obrazaca kretanja
prostorno-vremensko zaključivanje
ekspertno znanje
informacije o kontekstu
generički model
Keywords (english)
wild animals’ movement prediction
movement patterns recognition
spatio-temporal reasoning
expert knowledge
contextual information
a generic model
Language croatian URN:NBN urn:nbn:hr:168:773608 Project Number: 036-0361983-2022 Project Number: 036-0361983-3137 Study programme Title: Computer Science Catalog URL http://lib.fer.hr/cgi-bin/koha/opac-detail.pl?biblionumber=38017 Type of resource Text Extent 138 str. ; 30 cm File origin Born digital Access conditions Closed access Terms of use Created on 2019-07-16 12:15:34
