346 Sensor-Determined Behaviors and Pasture Intake Estimation in Extensive Grazing Systems
| Autor: | Philip Valencia, G. J. Bishop-Hurley, Paul L. Greenwood, R. C. Dobos, Bryce Little, Aaron Ingham, F.A.P. Alvarenga |
|---|---|
| Rok vydání: | 2020 |
| Předmět: | |
| Zdroj: | J Anim Sci |
| ISSN: | 1525-3163 0021-8812 |
| DOI: | 10.1093/jas/skaa278.143 |
| Popis: | Wearable sensor devices to monitor livestock behavior and location in extensive grazing systems can overcome limitations to collection and use of behavior data. These data enable generation of new phenotypes for genetic parameter estimation and decision support tools. Technical challenges, including device hardware and location on-animal, sensor types and modalities, data and power management, and sensor networks to enable measurement of livestock phenotypes in extensive environments, are being addressed. Wearable sensors currently used for behavior classifications include accelerometers, magnetometers and/or gyroscope within inertial measurement units, and pressure and acoustic sensors. We primarily use tri-axial accelerometers because of their reliability and richness of data for feature extraction to classify behaviors. Behavior data combined with GPS also allows location, activity and behavior mapping. Development of livestock behavior classifiers using sensors requires annotation of time-synchronized behavior recordings using video and/or a behavior recording app (e.g. CSIRO AnnoLog). Various analytical methods are used to classify behaviors from sensor data, including supervised machine-learning applied to accelerometer data. Our devices generate data for concurrent classification of behaviors including grazing, ruminating, walking, resting and drinking with reliabilities ≥ 90%. Estimates of pasture intake using behavior data across a range of environments also require validation. We have a facility to concurrently generate benchmark estimates of pasture intake using chemical markers and/or biomass disappearance while recording behaviors using sensors. To date, our R&D using pastures ranging from nutritionally optimal to severely drought affected suggest time spent grazing accounts for up to 60% of variation in pasture intake by individual beef cattle. We are assessing other sources of variation including pasture removal events (bites, tears), classes of cattle, and pasture characteristics to determine if more variation in pasture intake can be explained within extensive grazing systems to enhance development of new traits and applications for precision management. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|