Abstract:
In this work we explore the use of cheap sensors to monitor the activities
of dairy cattle with the aim of using these sensors as part of a system to detect
important events such as when a cow is ill or on heat. This draws on advances in
human activity recognition (HAR) where wearable sensors are used to collect data
and infer human activity. Our sensor system is based on the Raspberry Pi
microprocessor interfaced to an accelerometer sensor. We explore the use of simple
machine learning techniques to infer activity from the data we collect and show that
our simple system has the potential to detect different animal activities such as
walking, standing and feeding. We also test the system on detection of human
activities collected under controlled conditions to demonstate the potential use of the
system. We envision an internet of things (IoT) system with cows in a herd mounted
with appropriate sensors which relay information to servers over the internet.
Farmers are then able to access information about their cattle at any time and take
appropriate action when events of interest are detected.
