Use of a tri-axial accelerometer for automated recording and classification of goats’ grazing behaviour
Publication Type: |
Journal Article |
Year of Publication: |
2009 |
Authors: |
Maëg Moreau, Stefan Siebert, Andreas Buerkert, Eva Schlecht |
Publication/Journal: |
Applied Animal Behaviour Science |
Keywords: |
acceleration data logger, activity level, motion sensor, mounting system, spatial location |
ISBN: |
0168-1591 |
Abstract:
The suitability of an inexpensive tri-axial accelerometer for the automated recording of goats’ activities at pasture was tested on a slightly undulating pasture in Central Germany (52h of registry) and on a rugged mountainous pasture in northern Oman (70h of registry). The logger was either mounted onto a chest belt, a dog harness or a neck collar. The device registered the animals’ acceleration and changes in head inclination every second (Germany) or every two seconds (Oman). To calibrate and validate the logger’s registries, an observer simultaneously recorded the goats’ activities, distinguishing between walking, resting and eating; the latter was further subdivided into grazing (head-down) and browsing (head-up). Merged with the observation data, the accelerometer recordings were imported into a specially designed computer programme that calculated moving averages for the transformed accelerometer data and selected threshold values to distinguish resting from eating and eating from walking. Calibration functions established from data sets of a first goat were validated with data from a second goat fitted with the same harness type. The true recognition of activities detected by the accelerometer and the corresponding programme ranged from 87% to 93% for eating, 68% to 90% for resting and 20% to 92% for walking. It was affected, for resting and walking, by the type of mounting system used for logger fixation (fixed effect; P<0.001) and, for resting and eating, by the number of observations (covariable; P<0.01). Using a dog harness, the programme correctly recognized head-up and head-down positions in 75–82% and in 61–71% of the observed cases, respectively. With solid data sets for the calibration, a reliable automated classification of goats’ activities is possible across different individuals and across husbandry systems, provided that the same harness type is used.