Abstract: Artificial monitoring has low efficiency in early warning of tachypnea pigs on farms. For automatically detecting sick pigs, the pig ridge-abdomen contour was captured by a machine vision, and the strong correlation between the frequency of fluctuating ridge-abdomen contour was confirmed by automatically calculating and the frequency by estimating. Ten selected pigs, the health and the tachypnea, were scored visually by six trained observers (5-point scale), and then the videos of the side-view pigs which stood still, with resolution of 320 pixels×240 pixels, were recorded. On Matlab simulation platform, recorded videos were split into sequences of gray images. By using subtraction of background, extraction of ridge-abdomen contour and fluctuating ridge-abdomen contour descriptor, the fluctuating frequencies from automatic capture were compared with the ones from manual calculation, and it showed that the mean correlation coefficient of all measurements was 0.947. The recognition precision of fluctuating ridge-abdomen was higher than 85% with the pig-window size ratio of 0.35–0.75, and the positive linear relationship between frequencies of tachypnea pig and the scores from human was validated. Vision techniques have well potential for warning tachypnea pigs.