engleski

Discovering the Animal Movement Patterns using Hidden Markov Model

Collars equipped with GPS/GSM device provide automated remote tracking of animal at successive time intervals. Since animal movement is complex and irregular, methods based on repetitiveness of human movement cannot be applied to animals. In this paper we propose a method for animal movement pattern discovery using Hidden Markov model. We present a prototype for pattern discovery which enables arbitrary definition of model states by an expert in the field, as well as marking the patterns for learning phase of the method. Case studies of wolf and bear movements validate the method and show promising results - up to 94.02% accuracy in detecting the patterns among foraging, travelling and bedding for wolfs, and up to 82.14% accuracy in detecting feeding, resting and travelling patterns for bears on intermediate spatio-temporal scale.

moving objects; animal movement; pattern discovery; hidden Markov model

nije evidentirano