240 likes | 302 Views
Validation of a color automated tracking system for activity and pen location of group housed weanling pigs.
E N D
Validation of a color automated tracking system for activity and pen location of group housed weanling pigs J.W. Dailey*1, N. Krebs2, J.A. Carroll1, J.J. McGlone2; Livestock Issues Research Unit, Agricultural Research Service-USDA1, Department of Animal and Food Sciences, Texas Tech University2
Introduction • Behavioral data are included in more and more research projects. • Scan samples: not validated • Continuous observation: time-consuming • Automated Tracking System (ATS): • Eliminates human observer • Increases rate of scoring tapes • But: limited range of behavioral description (location mainly)
Objective • Compare: • a human observer using the Observer 5.0 (HOB) • an automated tracking system (ATS), Ethovision, using the ′Color′ tracking method • Model: group-housed weaned pigs, location within a pen
Materials and methods • Automated Tracking System-previous work: • Validation of the grey scale tracking • Validation of ‘zones’
Comparison of Human Observer vs Automated Tracking Systembased on the zones*means significantly different at P<0.05 r2 =0.83 SE =3.23 r2 =0.63 SE =1.79 r2 =0.37 SE = .72 r2 =0.13 SE = .37 * *
Conclusion for Grey Scale • The only behaviors which were accurately scored by the ATS: • overall activity • ′in the feeder zone′ • Regression: • r2=0.83 for overall activity • r2=0.63 for the feeding zone
Materials and methods • Current focus: color tracking • Improved efficiency and accuracy • More specific coordinate location is plotted for every data point • 4 to 8 animals
Materials and Methods • 40 nursery pigs, group housed, 4 per pen (1.22 m X 2.44 m), fed ad libitum • Each pig in the pen marked with a different color tape around its shoulders • Blue • Yellow • Pink • Orange
Materials and Methods • Each pig was scored by a human observer for 1h • Time spent on the right of the pen
ATS set up • Preliminary set up: ‘identify what and where to track’ • Definition of a general arena • Definition of specific zones within the arena • Definition of colors • Definition of the movement criteria
Definition of the movement criteria • Where in the pen is the animal? • Minimum size of object: 10 vs. 100 pixels • ‘When do we consider the pig has moved?’: • Calibration; several known measurements determine how big the pen is • Averaging intervals; last 99 positions
Statistical Analysis • Experimental design: • Experimental unit = individual pig • Time spent in each zone (presented in %) over the 1h-period • Proc GLM in SAS: equality of the means • Proc REG in SAS: simple linear regression: • equality of the slopes • coefficient of determination
Results: Comparison of ATS vs. HOB on the percentage of time spent on the right with 10 or 100 pixels % of time on the right
Results: Regression analysis • 100 PIX: • HOB = 0.86633 ATS (± 0.06633) +8.033 (± 3.85) • R2= 0.8182 • 10 PIX: • HOB= 0.98707 ATS (± 0.03569) +2.3 (± 2.06) • R2= 0.9514
Conclusion • The ATS set up on 10 pixels, avg 99 latest positions is a reliable tool for the location/position in the pen of newly weaned pigs grouped by 4 • For one hour of data collection: • Human observer: 1 pen = 4 pigs = 80 min • ATS: 1 pen = 4 pigs = 15 min (this was done at 30 frames per second)
Discussion • Need to validate system for each animal model: • Size of animal • Color: brightness, elasticity of the tape, lighting in the room • Some Issues: • Feeding zone: head in feeder or feeder-sitter do not mean ‘eating’: location does not necessarily correlate with behavior ! • Colored pigs
Thank You ! Questions ?
Results: Comparison of Human Observer versus Automated Tracking Behavior Data