Automatic reading of Kamars for heat detection in dairy herds

摘要

Most existing heat detection systems for dairy cattle are labour-intensive requiring manual inspection of every animal to maximise submission rates. Our solution is to automate heat detection using Kamar HeatMount Detectors and a camera-based recognition system. The goal of our system is to obtain 99% sensitivity and specificity, ensuring that cows that are on heat are rarely missed and over-drafts are kept to a minimum. Trials of various camera prototypes have been undertaken over the past three years, most recently during the artificial breeding period in the Spring of 2009. Over 70,000 images from four farms were taken during the latest trials; these photographs were used to assess the system. When considering photo observations, the fully automated system had an error rate less than 1% across all sites in the latest trial, and reaches our goal of 99% sensitivity and specificity.