Automated assessment of broiler chicken welfare using optical flow patterns in relation to behaviour, disease risk, environment and production.

Real improvements to farm animal welfare could be made if there were practical, objective, easy to use ways of assessing welfare routinely on commercial farms, not as a substitute for good stockmanship, but as an extension to it, when there is no-one around. Now that we have inexpensive camera technology available off the shelf, it might seem a simple matter to install cameras on farms, but how do we make sense of the mass of information cameras provide to enable us to assess the welfare of the animals?
In the case of broiler (meat) chickens, we have shown that a promising way forward is to use a computer to monitor the camera images and to pick out changes in the 'optical flow' patterns caused by movements of the chicken flocks. 'Optical flow' works by detecting the rate of change in light and darkness in different parts of an image and so is particularly good at picking up movement. However, what seems to be most revealing about welfare is not the amount of movement (quiet flocks can have just as good welfare as active ones) but the heterogeneiity or mixture of movement that they show.
One of the biggest welfare problems in broiler chickens is that many of them become lame and have great difficulty walking. The optical flow patterns of a healthy flock (all walking well) are much more uniform than those shown by flocks with a large proportion of lame birds because lame birds walk more slowly and there is a greater range of walking speeds in flocks with poorer welfare. The optical flow system picks up the greater range of movement in unhealthy flocks, thus giving a direct connection between a major welfare concern (lameness) at the individual level and the optical flow patterns seen at flock level.
Our preliminary trials have been so successful at picking out broiler flocks with welfare problems (% mortality, walking ability, condition of feet and legs) that we now want to test it out on a much wider range of broiler farms than we have attempted so far and to relate what the optical flow patterns are showing more closely to what is causing disturbances to health and welfare. So in addition to continuing to look at lameness and the leg damage that broilers can be subject to, we want to see whether changes in optical flow could also reveal the presence of disease, perhaps before the birds are showing clinical signs. We will focus on diseases that are of particular concern to poultry producers (Salmonella, necrotic enteritis and coccidiosis) as well as diseases that are carried by poultry but primarily affect humans (Campylobacter). All of these organisms alter the gut of chickens, often making their faeces runny, which makes the litter messy, which in turn gives the birds ulcerated feet and damaged legs ('hock burn'). This connection between gut health and external damage that could affect walking behaviour makes gut diseases a prime candidate for showing up as disturbances of optical flow at flock level. If we can show that a connection does exist between disease levels and optical flow disturbances, this could be important in allowing farmers to detect disease and poor welfare at an early stage and so intervene before they become real problems.
To ensure our results are relevant to the way most broilers are farmed today, all our work will be done on commercial farms and in collaboration with a leading chicken breeder and a major producer. Funding by the BBSRC will, however, make sure that our results are seen as independent of commercial pressures.
Our ultimate goal, extending beyond the life of this project, is to develop the optical flow system so that it is not just a way of assessing chicken welfare but also becomes an important management tool for producers, enabling them to reduce disease levels and manage their flocks more efficiently as well. If producers can see the commercial advantages of managing flocks with low disease and high welfare, everyone gains, especially the animals.

Our goal is to develop a practical, objective, easy to use way of assessing animal welfare that can be applied routinely on commercial farms. With broiler (meat) chickens, we have already shown that a camera/computer system can be used to detect disturbances to 'optical flow' patterns caused by movements of a flock inside commercial broiler houses. Changes in the skew and kurtosis of optical flow are correlated with key outcome measures such as final flock mortality, mean gait score (indication of the % of lame birds), and the numbers of birds with damaged hocks. The aim of this project is to build on these basic findings by using optical flow technology to develop a new management and welfare monitoring tool for broiler producers, inexpensive enough to be widely used on commercial farms and informative enough to represent a step-change in animal welfare assessment.
The objectives are
1. To validate the use of optical flow as a method for assessing the welfare of commercial broiler chickens in a wider range of housing types, genotypes and environments than has so far been attempted. We will test the hypothesis that variations in optical flow patterns are associated with variations in established welfare indicators such as % mortality, gaits score and leg/foot health.
2. To test the hypothesis that variations in optical flow patterns are associated with variation in disease burden, particularly those diseases and zoonoses that affect gut health (Coccidia (Eimeria), Salmonella, Campylobacter and Clostridium) and are of particular concern to the poultry industry. If the hypothesis is true, then husbandry interventions in real-time on farm become possible and could make a substantial contribution to reducing the burden of chicken and human disease.
3. To test the hypothesis that variations in optical flow are mediated through changes in the behaviour of individuals associated with increased risk of infection associated with poor litter quality.

Beneficiaries include:
The poultry industry (breeders, producers, retailers) who will benefit from having access to a new inexpensive means of assessing the welfare of poultry that is objective and gives them continuous information about the state of their flocks. They will have a management tool that will enable them to rear chickens to higher standards of health and welfare and to achieve economically desirable evenness in their birds. Farm managers and veterinarians will have a way of detecting health problems before they become serious and while the birds are still young enough for remedial steps to be taken.

The industrial beneficiaries will gain in at least two ways. Firstly they will have available a new way of assessing their flocks even when a stockman is not even present. This will enable them to give a clear indication to their customers that improving animal welfare is a priority for them and for the companies to set higher standards for chicken health and welfare. This in turn could have important economic consequences for them as they can use this in the marketing of their birds. Secondly, the algorithms developed for the camera/computer system pick out variation and unevenness in flocks and therefore will help producers to grow their birds in a standard way. This is commercially important because companies need to deliver a predictable product of chickens of the same weight. Evenness of body size also makes flock management easier (e.g. in raising drinkers to a height that all birds can reach) and slaughter house management more efficient and humans (e.g. in setting equipment to deal optimally with the whole flock)

Animal welfare scientists will benefit from having a research tool that will enable the to collect welfare data on a much larger scale than has been possible before and thus to base their conclusions on much higher quality evidence. The application of the optical flow camera/computer system to broiler chickens, will pave the way to implementing the same for other species and to other welfare issues such as predicting outbreaks of tail-biting in pigs before serious damage is done.

Policy makers (Governments, EU, NGOs, supermarkets etc) will benefit from having access to higher standards of evidence and so be able to make sounder, more evidence-based decisions. The debate on how to make livestock production more efficient while still giving priority to animal welfare will be informed by much better evidence of what actually does improve chicken health and welfare.

The general public will benefit from knowing that something is being actively done to improve the welfare of an animal that arouses the concern of many people and has been the focus of considerable media attention.

The main pathways to impact will be through working directly with a large chicken breeder company and a major chicken producer, obtaining feedback from farmers and using their links with customers throughout the poultry industry to show-case what camera/computer system can do for them. We will also engage with the public through open days, seminars, popular articles in the farming press and online, and with academic researchers through conferences and papers in the scientific literature.