Chickens on Camera: Op cal Flow and the Sta s cs of Welfare
Marian Stamp Dawkins
Department of Zoology
University of Oxford
Royal Society, July 2014 Climate change, feeding people and the future of agriculture
” “sustainably intensive The Royal Society (2009)
Is animal welfare being pushed off the poli cal agenda? “Agriculture will need to produce more food from the same or less land, using less water, energy and other inputs and reducing waste and adverse environmental impacts including greenhouse gas emissions”. (2013) h p://www.un.org/millenniumgoals/
¥ Limi ng land for livestock ¥ Greater efficiency ¥ Sustainable intensifica on ¥ More technology
Where is animal welfare? Broiler (meat) chickens are already highly ‘efficient’ at conver ng food to meat
Broilers are kept in large sheds in flocks of up to 70,000 birds
Globally, 60 billion chickens killed each year
It takes only 35 days and 2.25 kg of food to grow a chicken to 1.5 kg (FCR of 1.5) This efficiency achieved by selec ve breeding and diet
Selec on for fast growth rate has also led to welfare problems Lameness in is scored from 0= best to 5=worst Bristol gait score 4 (0=best; 5= worst) From Knowles, T.G. et al (2008) PLoS ONE 3(2) e1545
From:T.G. Knowles et al (2008) PLoS ONE 3(2): e1545. Current methods of assessing broiler welfare are mainly post-mortem
GAIT SCORE PODODERMATITIS
% MORTALITY
HOCK BURN
1 5 10 15 20 25 30 35
AGE OF BIRDS (IN DAYS) We aim to assesses welfare throughout life and so an cipate and intervene to achieve healthier flocks Our aim: efficiency and animal welfare
¥ Monitor welfare throughout life ¥ Help producers to manage their flocks for lower mortality, lower levels of disease, higher quality product ¥ Give early warning of health and welfare problems, enabling targeted interven ons and reduced
Our system uses smartphone cameras
The smartphone delivers a welfare ‘verdict’ every 15 mins Op cal flow: rate of change in image brightness
Time Each (320 x 240) video frame is divided into (8X8) pixel squares. If there is no movement between frames, the brightness of all squares remains the same
If movement occurs, there is a change from light to dark or vice versa. Op cal flow compares the pa erns of light and dark in successive images
(a) Image frame at time t (b) Image frame at time t +1
Individual animals are not tracked, but the changing patterns over time (“flow”) give an indication of behaviour at flock level.
Dawkins et al. (2009) Applied Animal Behaviour Science 119: 203-209
24 broiler flocks, each with 34,000 birds, iden cal houses, one site, one manager
Company On-site Optical flow measures measures from video
Mortality Mean Hock burn Gait score Variance Podo- Skewness dermatitis Kurtosis Body wt. Lower welfare flocks had lower mean flow and more unusual movement (higher skew & kurtosis)
Welfare Mean Variance Skewness Kurtosis Measures optical flow optical flow optical flow optical flow % flock -0.42* 0.003 0.42* 0.45* mortality % hock -0.36 0.09 0.57** 0.56** burn % podo 0.05 0.27 0.33 0.32 % poor -0.33 -0.27 0.42* 0.48* gaits
The figures shown are correla on coefficients; *=p<0.05, **=p<0.01 From Dawkins et al (2012) Animal Behaviouri Kurtosis is sensitive to the spread of extremes in the data
Kurtosis Skew and kurtosis are measures of lack of uniformity in the movement
In a race where all the compe tors are equally fit, the ‘flow’ will have a high mean and low skew & kurtosis.
Photo: I.dailymail.co.uk/prx In a race where there is a greater spread of abili es, the ‘flow’ will have a lower mean and higher skew kurtosis
Photo: 400mtogo.com The sta s cs of welfare
¥ Poor welfare: lower mean movement but higher skew and kurtosis: slower and more variable ¥ High welfare flocks: higher mean movement but lower skew and kurtosis: faster and more uniform The so ware can separate flocks with high final mortality from 15 days
Kurtosis devia on from mean
It can also dis nguish flocks with poor gaits
Kurtosis devia on from mean
It can even predict hockburn in flocks as young as 3 days old
PREDICTION OF HOCKBURN
Correlation significant at p<0.01
Roberts, S.J., Cain, R. & Dawkins, M..S. (2012) Royal Society Journal Interface Where we are now
¥ Commercial trials running in the UK, France and the US ¥ Current research on op cal flow in rela on to disease in the chicken flocks (Campylobacter, Clostridium, Salmonella, Coccidia) ¥ Pressure from producers to be able to use the system for themselves The team….
Steve Roberts (Engineering Science, Oxford) Mar n Maiden (Zoology, Oxford) Adrian Smith (Zoology, Oxford) Frances Colles (Zoology, Oxford) Russell Cain (Zoology, Oxford) Tom Nickson (Engineering Science, Oxford) Marian Dawkins (Zoology)
Making animal welfare count ¥ Integrate animal welfare with other concerns (food security, efficiency, disease reduc on, environmental impact) ¥ Help producers to find economic benefits for good animal welfare Precision farming with animal welfare