Subject: Identification of biomechanical fragility in cattle using non-conventional radar imaging for early assessment of lameness. Design and Implementation of a real-time system.
Context & objectives
Livestock farming plays a crucial role in the shift towards regenerative agriculture, necessitating farmers to employ technical tools for identifying deviations in their herds. The absence of a standard for normality or abnormality requires the evolution of livestock farming to minimize ecological footprints. Environmental concerns related to beef and sheep production primarily stem from certain production systems, prompting the establishment and support of ecologically conscious systems worldwide. Cattle, as herbivores with low energy conversion rates when fed on energy-rich cereals and protein-rich soya, are well-suited for environments where alternative agriculture is challenging, such as mountainous regions, leading to low-density farming. However, this poses a challenge for farmers who must visually observe their herds for early symptom detection to enact curative actions.
Lameness, a significant welfare issue in dairy cattle, particularly affects hooves, leading to altered gait, impaired fertility, and reduced milk production. The associated costs for non-specific lameness problems can be prohibitive per animal. While various detection methods have been proposed, they often suffer from accuracy, reliability, equipment cost, and time consumption issues. The proposed project aims to use radar technology for analysing gait abnormalities in cattle. Previous work has demonstrated the feasibility of radar for lameness detection, but challenges in machine learning labeling persist. The project aims to conduct a longitudinal analysis of individual cows, focusing on long-term temporal sequences derived from radar imagery to identify pathological cases. Collaboration with Bovinéo is anticipated to enhance the field of animal welfare through this innovative approach.
Animal welfare is one of the objectives of Sustainable Development Goal (SDG3 and 15). The Smart_Farm research project aims to automate the detection and classification processes of lameness in cattle using radar technology. To achieve this, we will first assess the applicability of the radar in combination with the computational power of artificial intelligence (AI) and machine learning (ML) algorithms in identifying subtle variations in cattle mobility that signal early-stage lameness. This issue needs to develop biomechanical models of the animal with the collaboration of vet, and to follow the parameters over time. Variations and deviations in these parameters will help to identify a change in behaviour. The next step is the evaluation of the generalisability of the proposed system within the broader context of animal welfare by analysing long-term temporal sequences collected from the radar signatures. Data on the pathological cases associated with lameness will be collected from farms collaborating with Bovinéo, one of the largest commercial groups in the west of France dealing with cattle. The overarching aim is to develop an efficient and real-time automated monitoring system for identifying lameness in cattle and contribute to advancing sustainable livestock farming practices.
Traditional methods of detecting lameness involve visual scoring of cattle in the herd after exiting the milking parlour. This approach, while widely used, has a few limitations. Visual scoring is subjective, influenced by individual biases that cause variations in the outcomes (Schlageter-Tello et al., 2014), and is inherently time-consuming and labour-intensive, especially with larger herds. Therefore, due to these logistical constraints, continuous monitoring becomes impractical, and it can be implemented only a few times per year, even though several studies have demonstrated that lameness, particularly in severe cases, adversely impacts animal productivity (Booth et al., 2004; Green et al., 2002; Morris et al., 2011), and affects the profitability of stakeholders (Dolecheck & Bewley, 2018; Liang et al., 2017). The oversight of subtle changes in mobility could lead to a delay in treatment, which in turn makes a continuous monitoring system a necessity.
Several technological advancements have been proposed for lameness detection, including video surveillance for kinetic analyses (Anagnostopoulos et al., 2023), infrared cameras (Alsaaod et al., 2015), force plate systems (Pastell et al., 2008), and accelerometers attached to cattle (Balasso et al., 2021). However, none of these systems is widely adopted as they present limitations. Infrared cameras, for instance, require calibration before every use based on ambient temperature, and wearable devices are herd size dependent, increasing the cost and also contributing to an environmental footprint. Other systems, like cameras for kinetic analysis, even though they are more promising, exhibit accuracy limitations such as high specificity (90%) but low sensitivity (51%) (Anagnostopoulos et al., 2023). The experimental conditions for monitoring herds limit the use of cameras in the visible range. They require regular maintenance and cleaning of the optical parts due to, for example, the projection of mud by the animals and environmental conditions (fog, rain, light, …). Moreover, these systems do not permit monitoring of the animals on a 24-hour basis, especially during the night.
Radar technology provides the possibility to overcome most limitations by requiring only one device (easily scalable to the herd size), offering contactless monitoring, and operating independently of environmental conditions—making it ideal for diverse farm environments. Previous publications (Busin et al., 2019; Shrestha et al., 2018) support the efficiency of the system, showing high accuracy (cattle: ACC> 0.83, SP=0.81, SE=0.85) in identifying lame animals (cows and sheep) and its potential as an effective tool for improving lameness detection and classification in livestock. Our hypothesis is that the radar system has the capability to detect deviations from the normal mobility patterns of cows and classify the severity of lameness. Lameness in cows is often associated with an altered gait pattern, indicative of potential hoof lesions or issues affecting other parts of the limbs, with approximately 90% of the lesions located in the hooves (Clarkson et al., 1996). We expect that the proposed automated system will enable the processing of long-term temporal sequences collected from radar signatures, and with the use of machine learning and AI for processing and analysis, we will achieve continuous, real-time monitoring and identification of behavioural limitations associated with lameness.
Objective of the application
The radar system will automate the detection and classification of lameness in cattle and enable continuous monitoring of their mobility behaviours.
Target population: dairy milking cows with varying mobility statuses ranging from sound (normal mobility behaviour) to severely impaired/high degree of lameness (cow cannot move forward).
Sensing modalities: The proposed system will use radar and video recordings, as well as visual assessments and physical examinations, along with a biomechanical model of the cattle as references
The University of Glasgow and University Cergy-Pontoise have been developing a bespoke real-time platform for human activity recognition in the context of assisted living that will allow in the context of animal welfare to improve on the separation of the limbs for a better fidelity in gait analysis. This radar V1 has now been tested and the radar V2 is currently being developed for deployment on farms.
The objectives of the post-doc position are in the field of the development and validation of real-time Monitoring
System Using Radar Technology :
- Real-time experiments in vivo, in the farm.
- Collect data and upgrade the models
- Implementation of SW for the post-treatment of the collection of data
- Promote the new results by journal and conference.
The position is open for starting as soon as possible.
How to candidate?
- Send a CV and motivation letter to : olivier.romain@cyu.fr and Julien.LeKernec@glasgow.ac.uk