Quantification of simulated cow urine puddle areas using a thermal IR camera

In Europe, National Emission Ceilings (NEC) have been set to regulate the emissions of harmful gases, like ammonia (NH₃). From NH₃ emission models and a sensitivity analysis, it is known that one of the major variables that determines NH₃ emission from dairy cow houses is the urine puddle area on the floor. However, puddle area data from cow houses is scarce. This is caused by the lack of appropriate measurement methods and the challenging measurement circumstances in the houses. In a preliminary study inside commercial dairy cow houses, an IR camera was successfully tested to distinguish a fresh urine puddle from its background to determine a puddle's area. The objective of this study was to further develop, improve and validate the IR camera method to determine the area of a warm fluid layer with a measurement uncertainty of <0.1 m². In a laboratory set-up, 90 artificial, warm, blue puddles were created, and both an IR and a colour image of each puddle was taken within 5 s after puddle application. For the colour images, three annotators determined the ground truth puddle areas (A_p,GT). For the IR images, an adaptive IR threshold algorithm was developed, based on the mean background temperature and the standard deviation of all temperature values in an image. This IR algorithm was able to automatically determine the IR puddle area (A_p,IR) in each IR image. The agreement between the two methods was assessed. The A_p,IR underestimated the A_p,GT by 2.53% for which is compensated by the model A_p,GT=1.0253·A_p,IR. This regression model intercepted with zero and the noise was only 0.0651 m², so the measurement uncertainty was <0.1 m². In addition, the A_p,IR was not affected by the mean background temperature.