A motion and image analysis method for automatic detection of estrus and mating behavior in cattle

Abstract

Estrus detection in cattle is very important for breeding management. The detection results help determine the optimum time for artificial insemination. The success of estrus detection not only increases conception rates but also raises milk production for the herd. The proposed image processing method to detect cows in heat from video is passive and therefore does not require devices to be attached to the cow’s body. The observed scene is a commercial dairy farm, where Holstein cattle are reared in roofed cowsheds.The proposed method is based on the inherent behavior of cows in heat which is characterized by a cow following another cow before mounting it for a short period of a few seconds. A region of interest was first determined around image locations where high levels of motion occurred across the video frames. Foreground segmentation then follows to partition the moving target object from the region of interest in each image frame. The following and mounting behaviors are identified based on the changes of moving object lengths. When an estrus event is declared, the detected event is recorded into the computer to be verified later. The detection system observed 20 cows in a cowshed from the top view. Over the period of 800h (80days at 10h each day), only 0.33% of video frames detected were false positive indicating the system was reliable at detecting mating and estrus behavior from the video frames. Using the method the farmer is required to view the recorded video frames for approximately 2min per day to confirm that the detection results are true estrus events. The proposed method is capable of real-time detection of mating and estrus behavior speeds with a processing rate of 35fps.