A brief review of the application of machine vision in livestock behaviour analysis

Abstract

It is desirable to increase the frequency between livestock welfare assessments to enhance problem identification and consumer confidence in livestock welfare management. However, animal welfare is difficult to monitor in practice, due to the inefficiencies involved in manually documenting and determining, animal behaviour, social interaction and health condition of large numbers of animals. Furthermore, the effectiveness of a welfare assessment relies on the intuition of the observer which may vary considerably between assessors. Hence, this review investigates the application of machine vision systems to recognise and monitor the behaviour of animals in a quantitative manner. Behaviour-recognition concepts, techniques, and current behaviour monitoring systems are reviewed. Findings indicate that further research is required to develop systems that can monitor the behaviour and welfare of animals' more efficiently and effectively in commercially realistic environments. Introduction: Livestock Behaviour and Vision SystemsMeasuring and assessing the behaviour of livestock is important as it can be used to indicate their welfare status (Dutta et al., 2015;Porto et al., 2014). Behaviour is formed from an animal's continuous interaction with its environment (Figure 1). Animal behaviour is a response to an internal stimuli (physiological) such as hunger or an external stimuli such as climate. If the goal of the behaviour is not, or cannot, be achieved the animal may change its behaviour or physiological response (Busse et al., 2015). There is potential for welfare problems to arise when there are inadequate environmental enrichments to support the behavioural needs of livestock.It has been argued that the husbandry methods used in intensive livestock production have resulted in the deprivation of some naturally occurring behaviours (Fraser, 1983). Producers have a strong interest in maintaining the welfare of their livestock from both economic and ethical perspectives. As often proactive welfare management has a positive effect on a farm's production efficiency and the quality and marketability of the end product (Kashiha et al., 2013).The observation processes currently used to measure livestock behaviour are subjective, as farm workers perform the welfare assessment. The worker's involvement in these tasks are necessary and therefore increase the demand on the labour and costs associated with monitoring animal behaviour. These factors have potential to influence the level of attention each animal receives (Oczak et al., 2013;Pereira et al., 2013). Thus, behavioural measurements are open for interpretation and have potential to be overlooked.Given the problems mentioned above, this review investigates the application of machine vision systems to recognise and monitor the behaviour of animals in a quantitative manner. Specifically in this article, behaviour-recognition concepts, techniques, and current behaviour monitoring systems are reviewed. Applying machine vision systems to the behaviour recognition taskTo over...

ASAS-NANP SYMPOSIUM: Applications of machine learning for livestock body weight prediction from digital images

“…An alternative way to acquire morphometric and behavior-based biometric measurements of livestock animals consists in using contactless optical systems, which overcome difficulties arising from direct measurements. Different types of 2D and 3D optical sensors have been successfully used mostly for morphometric measurements and in a limited way for biometric identification of animal behaviors ( Porto et al, 2013 ; Banhazi and Tscharke, 2016 ; Nasirahmadi et al, 2017 ; Li et al, 2019 ). The 2D sensors include regular 2D digital cameras, thermal cameras ( Stajnko et al, 2008 ), and systems of cameras capable to extrapolate 3D models from a series of 2D images.…”

A Literature Review of Modeling Approaches Applied to Data Collected in Automatic Milking Systems

“…Furthermore, there is the potential for real-time analysis of this data [ 94 ]. The application of machine vision systems to recognize and monitor the activity and behavior of animals in a quantitative manner could become the solution needed [ 95 ]. Guzhva et al [ 96 ] used top-view cameras to automatically detect social interactions (head pressing and body pushing).…”

Deep learning image recognition of cow behavior and an open data set acquired near an automatic milking robot

“…Several technical solutions have been used to monitor animal welfare, including the use of digital video cameras (Porto et al 2015, Ardo et al 2017, Ter-Sarkisov et al 2017, depth sensor cameras (Nasirahmadi et al 2017), sound (Schirmann et al 2009);three-dimensional accelerometers (Müller et al 2003, Steeneveld and Hogeveen 2015, Gardenier et al 2018, Shen et al 2020; and infrared thermography (de Sousa et al 2018, Cuthbertson et al 2019, Xudong et al 2020, Anagnostopoulos et al 2021. Computer vision approaches employing video cameras are scalable and low-cost solutions (Banhazi and Tscharke 2016) and have been successfully used to monitor physiological and behavioral parameters related to pre-slaughter stress (Jorquera-Chavez et al 2019); to detect hoof disease (Gu et al 2017); to track gait and identify lameness (Gardenier et al 2018, Jiang et al 2019a, Kang et al 2022, to analyze health problems through calculating body condition scores (Zin et al 2018b, body structure b, Liu et al 2020) and faecal monitoring (Atkinson et al 2020); and to detect aggressive behaviors (Chen et al 2019). One often used computer vision method is to segment instances to recognize and track individual cows (Guzhva et al 2018, Ter-Sarkisov et al 2018, Zin et al 2018a, Qiao et al 2019, Shao et al 2019, Li et al 2021.…”