Monitoring of sow postural activity from 3D millimeter-wave radar imaging

Abstract

This paper presents a novel approach to monitor the postural activity of sows in farrowing pen using a millimeter-wave radar imaging system. Three-dimensional images of the scene are obtained from a 77 GHz Multiple-Input Multiple Output radar and the mechanical scanning of the radar beam. We show that the processing of radar images based on the Cell-Averaging Constant False Alarm Rate allows detecting the standing and lying postures of sows despite the electromagnetic clutter. The experimental study reported here is performed from the monitoring of 16 sows from the Large White breed for a total of 46 h. 3D positions of the sow are recorded and heatmaps of the sow locomotor activity are proposed. With few training radar data, we perform a classification of standing and laying postures with a mean balanced accuracy greater than 90%. The radar-based technique applied here in a cluttered environment inside farrowing pens may be extended to monitor pigs under other housing condition.