The use of image analysis to determine the number and position of cattle at a water point

• This work assessed the accuracy of animal counts and positions using image analysis.
• Counting accuracy decrease with distance to camera.
• Positional accuracy of independent points was 0.8 ± 0.5 m.
• Positional accuracy did not change with distance to the camera.

This study assessed the application of an image analysis method to accurately determine the number and position of cattle which are critical inputs for enteric methane emission calculations using micrometeorological methods. Animal imagery was collected with three synchronised time-lapse cameras located at 7, 35 and 77 m from a 20 × 30 m water point enclosure containing 20 steers, recorded over three consecutive days. Four independent observers counted the number of animals visible in each of 516 images. The counting error increased with distance from the enclosure (0.1%, 3.7% and 15.4% of total animals) as a result of increased overlapping and decreased clarity of the animals on the image. Animal positions were estimated using a polynomial transformation of image coordinates (pixels) to map coordinates. The average location error (distance between estimated and actual position) of independent targets was 0.8 ± 0.5 m and did not change with distance to the camera. We conclude that the analysis of 12 MP images from time lapse cameras can provide reliable and accurate estimates of the position and the number of animals located within 55 m from the camera.