A Night-time Anomaly Detection System of Hog Activities Based on Passive Infrared Detector

Abstract

Abstract. The amount of daily activity can be used as important data for the analysis and evaluation of the health, diseases, and environmental conditions of hog farms, which in turn can affect fertility rate and productivity. In this article, a monitoring system based on a passive infrared detector (PID) is proposed to analyze daily hog activity and abnormal behaviors. The hardware includes a high-accuracy acquisition system, which uses a 24-bit ADS1256 chip as its A/D conversion and signal input channel, and a PID, which ensures that the signal can be obtained uninterruptedly day and night. Based on the LabVIEW software platform, a real-time data acquisition, display, and storage system was programmed in which the activity curve can be displayed, and the system parameters can be modified if necessary. A simulation experiment was performed in a test laboratory (7 × 17 m) with a larger size than a typical hog room (7 × 15 m), and the appropriate orientation of the sensor, the installed position, and the lens were selected. Data for 90 days (day and night) were collected in a hog room to establish the model of daily activity. To find the abnormal behaviors during the night, an improved K-means clustering was constructed. The results indicated that the improved K-means clustering method performed satisfactorily in clustering and anomaly detection. The developed system for daily activities monitoring and night-time anomaly detection could be a potential technique to assist research in hog behavior detection and animal welfare improvement. Keywords: Animal activity, Hog, Motion sensor, PID, Signal processing.