University of Twente Student Theses

Login

Investigating generic online animal activity recognition across different animal species

Hamelers, L.H. (2018) Investigating generic online animal activity recognition across different animal species.

[img] PDF
624kB
Abstract:Recognizing animal behaviour has proven to be very useful for detecting changes in the environment of the animal and in their wellbeing. Research has mainly focused on optimizing classifiers for one species or using different techniques for training classifiers and testing them separately on different species. In this research a different approach is taken to classifying animal behaviour. The research investigates how generic a specific animal behaviour classifier can be, applied to the group of quadruped animals. Seven different classifiers are trained with data from cows, horses, sheep and goats. The classifiers are trained with the datasets of one, two or three different species and tested on another species, a so-called non-mixed experiment. This research shows that the k-Nearest Neighbor algorithm gives the best results for non-mixed classifiers, and that with increasing of the number of species in the training set, the accuracy of the classifier increases. Further, the recognition of certain activities is investigated. The activities that are recorded are stationary, grazing, walking, trotting and running. The activities stationary and grazing are classified more accurately than the other activities. This work supports further research into non-mixed classifying to eventually develop a generic classifier for quadruped animals.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:50 technical science in general, 54 computer science
Programme:Business & IT BSc (56066)
Awards:Best Paper Award 29th Biannual Twente Student Conference on IT
Link to this item:http://purl.utwente.nl/essays/77099
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page