Detection of porcine respiration based on machine vision

Abstract

A machine vision method is presented to identify porcine health in real-time by detecting porcine breath. Porcine images in top view are extracted by constructing video image acquisition system, and porcine contour is obtained by serious of image pretreatment. Concave-convex recognition method is used to determine the waist corner and scapular endpoint on one side of ventral lines. The length of the line between two points is measured using improved chain code algorithm. The data of length distribution are recorded to draw time-position figure, and the fluctuation of the target curve approximately reflects the porcine breath in frame sequences. So the breath rate could be expressed as the frequency of the curve. Compared with manual observation, the relative error of the result in this paper is about 6.05% in detecting respiratory rate. Therefore, machine vision-based method is effective for detecting porcine breath.