57
Eating and Drinking Behavior Prediction by use of Tri-Axial Accelerometers in Dairy Cattle
Tri-axial accelerometers are often used to measure behavior in cows (e.g. estrus, standing and lying). However, there may be other practical uses for accelerometers in dairy farms. Our objective was to determine if accelerometers placed on a collar around the cow’s neck can be used to monitor feeding and drinking behaviors. For this study, 12 lactating Holsteins (DIM 76 ± 35) were housed in stanchion stalls and continuously recorded for 6 d (Swann Pro-530 night/ day cameras, DVR). Cows were fitted with Onset Pendant G accelerometers on the collar and sampling intervals set at 6s. Video data were watched and evaluated by the same person. Daily Video duration (Video) of each behavior was summarized and compared to daily duration predicted by accelerometers. Three methods were created to evaluate behavior prediction by accelerometers. For method 1 (MET1) dataset was constructed based on the mean for the 3 axes recorded. For method 2 (MET2) dataset was constructed based on the mean plus the standard error for the 3 axes recorded. For method 3 (MET3) dataset was constructed based on the mean minus the standard error for the 3 axes recorded. Four behaviors analyzed were standing and eating with head up (SEHU), standing and eating with head down (SEHD), standing and eating (EAT = SEHU + SEHD), and drinking (DRK). Statistical analysis was performed using the MIXED procedure in SAS. For SEHU there was difference (P < 0 .001) between Video (119.0 min) and MET1, MET2, and MET3 and tendency (P > 0.06) for differences among MET1, MET2, and MET3. For SEHD there was difference (P < 0 .001) between Video (141.7) and MET1, MET2, and MET3; and tendency (P > 0.09) for differences among MET1, MET2, and MET3. For EAT there was difference (P < 0 .001) between Video (260.7 min) and MET1, MET2, and MET3 however, there was no difference (P > 0.48) among MET1, MET2, MET3, and EAT. For DRK there was no difference (P = 0.89) between Video (315.3 min) and MET1, MET2, MET3. The accelerometer under predicted SEHU by 42.6%, 35.4%, and 49.2%; SEHD by 72.0%, 65.3%, and 71.8%; EAT by 57.6%, 51.6%, and 61.05% for MET1, MET2, and MET3, respectively. The accelerometer accurately predicted DRK by 100%, 103%, and 99% for MET1, MET2, and MET3, respectively. In conclusion, accelerometers were successful in predicting drinking behavior.
Keywords: Accelerometer, Intake, Prediction