Harnessing 3D cameras and deep learning for on-the-fly automated body condition and mobility analysis to improve cattle welfare

不断增长的世界人口和气候变化正在给粮食供应带来压力。高动物福利和卫生做法比以往任何时候都更重要，以满足社会对畜牧业的需求。对奶牛使用精密监测仪器是优化生产同时保持动物健康的关键。在这个博士项目中，将使用最新的英特尔摄像机的3D视频技术，以不引人注目的方式提供无压力的监测，监测奶牛个体的流动性和身体状况的增量变化，目的是了解观察到的跛行之前的行为线索，以改善奶牛的健康、福利和生产力，从而提高牛奶生产的气候和环境可持续性。这些特征目前是通过人工视觉评估来测量的，需要很高的技能水平和训练，但仍然对个人的主观性开放，很少捕捉到新行为研究所需的纵向细节。这将实现一个可以移植到农场的系统，而不需要大量的仪器，允许农民和价值链中的其他人，如兽医、营养学家和牲畜顾问，使用更精确、一致和频繁的测量，为提高奶牛的性能和福利创造更大的机会。该奖学金将适合对可持续食品生产感兴趣的数学或计算专业学生，或者希望建立人工智能技能的兽医或生物科学专业人士——无论哪种情况，都将开发量身定制的培训包。学生将学习动物福利评估的关键方面，并利用这一点和尖端的人工智能来构建和应用该系统，以便我们更好地了解跛行。该学生将在两个领先的、地理位置接近的研究所——西英格兰大学的布里斯托尔机器人实验室和布里斯托尔大学的布里斯托尔兽医学院和视觉信息实验室工作，并将受益于由Melvyn Smith教授（机器视觉）和Andrew Dowsey教授（One Health Data Science）领导的广泛的跨学科监督团队。他们在这一领域发表了最先进的作品，将以此为基础（见https://doi.org/10.1016/j.compind.2018.02.011, https://arxiv.org/abs/2006.09205, https://www.biorxiv.org/content/10.1101/2020.08.03.234203v2）。数据收集和验证将利用布里斯托尔兽医学院的约翰·奥德克奶牛生产可持续发展和福利中心，这是一个位于温德赫斯特奶牛场的新研究中心，该中心将包括我们所有185头奶牛的全天候视频覆盖，与生产、排放和兽医评估数据相关（https://www.bristol.ac.uk/vet-school/research/john-oldacre-centre--farm-research-data-platform）。

A growing world population and climate change are stressing food availability. High animal welfare and health practices are more important than ever to satisfy societal demands for the livestock sector. The use of precision monitoring instrumentation for dairy cattle is key to optimising production while maintaining animal health. In this PhD project, 3D video technology with the latest Intel cameras will be used to unobtrusively provide stress-free monitoring of incremental changes in individual cow mobility and body condition with the aim of understanding behavioural cues preceding observed lameness to improve cow health, welfare and productivity and hence increase the climate and environmental sustainability of milk production. These traits are currently measured by manual visual assessment, requiring high skill levels and training, but are nevertheless open to the subjectivity of individuals and rarely capture the longitudinally detail needed for novel behaviour research. This will realise a system that can be transplanted into farms without extensive instrumentation, allowing farmers and others in the value chain such as vets, nutritionists and livestock advisers to make use of much more precise, consistent and frequent measurements, creating greater opportunities to improve cow performance and welfare. The studentship would suit either a mathematical or computational student interested in sustainable food production, or someone with veterinary or biosciences expertise who wishes to build up artificial intelligence skills - in either case a tailored training package will be developed to suit. The student will learn the key facets of animal welfare assessment, and use this and cutting edge AI to build and apply the system across the studentship timeline so we can develop a better understanding of lameness. The student will be based 50%/50% at two leading, geographically close institutes - Bristol Robotics Laboratory at the University of West of England, and Bristol Veterinary School & Visual Information Laboratory at the University of Bristol, and will benefit from a broad cross-disciplinary supervision team, led by Prof Melvyn Smith (Machine Vision) and Prof Andrew Dowsey (One Health Data Science), who have published state-of-the art work in this area that will be built upon (see https://doi.org/10.1016/j.compind.2018.02.011 , https://arxiv.org/abs/2006.09205, https://www.biorxiv.org/content/10.1101/2020.08.03.234203v2). Data collection and validation will harness the Bristol Veterinary School's John Oldacre Centre for Sustainability and Welfare in Dairy Production, a new research centre based at our Wyndhurst dairy farm, which will include blanket 24/7 video coverage of all our 185 cows linked to data on production, emissions and veterinary assessments (https://www.bristol.ac.uk/vet-school/research/john-oldacre-centre--farm-research-data-platform)