Validation of HOBO Pendant data loggers for automated step detection in two age classes of male turkeys: Growers and finishers
Applied Animal Behaviour Science
URL with Digital Object Identifier
© 2016 Elsevier B.V. Activity levels can be used as a predictor of health status, physical condition, feed efficiency, and coping style in animals. Small, portable data loggers have been validated as an inexpensive and effective alternative to video or live observation for automated activity detection in several livestock species. Our study aimed to validate the use of Onset HOBO Pendant® G acceleration data loggers for automated step detection in two age classes of male turkeys: growers and finishers. The loggers were attached to one leg in an elastic harness and programmed to record each turkey's leg accelerations every 0.1 s. For the grower turkeys, the activities of 60 males were recorded in a runway twice per week from 9-11 weeks of age. In the finisher trial, the loggers were attached to eight 14 week old male turkeys to record their stepping activities in a trial pen over 54 min. Simultaneous video observation allowed for logger suitability to be determined for reporting walking activity. The 0.1 s data points were identified as either: true positives, false positives, false negatives, or true negatives then the logger sensitivity, false discovery rate, specificity, precision, and accuracy were calculated for each turkey run. Over the 309 grower turkey runs, 4364 steps were observed in the video while the loggers detected 4865 steps. The grower logger results showed a mean sensitivity of 96.02 ± 6.82%, specificity of 99.10 ± 0.86%, and an average false discovery rate of 12.41 ± 10.13% for step detection (mean ± standard deviation). For the grower turkey loggers, the mean precision and accuracy for step detection were 87.59 ± 10.13% and 98.95 ± 0.77%. For the eight finisher turkeys, a total of 2420 steps were seen by the video and the loggers recorded 2691 data points as steps. The finisher turkey logger data showed a mean sensitivity of 89.58 ± 11.75%, mean false discovery rate of 14.73 ± 10.66%, and average specificity of 99.81 ± 0.19%. The finisher logger results had a mean precision of 85.27 ± 10.66% and a mean accuracy of 99.73 ± 0.13%. Although the loggers showed high sensitivity for detecting turkey steps under research settings, future research should try to reduce the loggers' susceptibility to background noise, which would improve their performance for application in commercial conditions.