Proceedings of the World Congress on Genetics Applied to Livestock Production, 11.613 Individual cow identification in a commercial herd using 3D camera technology Jørn Rind Thomasen 1 , Jan Lassen 1 , Glenn Gunnar Brink Nielsen 2 , Claus Borggard 2 , Peter René Bolvig Stentebjerg 2 , Rikke Hjort Hansen 2 , Niels Worsøe Hansen 2 , Søren Borchersen 1 1 VikingGenetics, Ebeltoftvej 16, 8960 Randers, Denmark jotho@vikinggenetics.com (Corresponding Author) 2 Danish Meat Research Institute, Gregersensvej 9, 2630 Taastrup, Denmark Introduction Automatic cow identification becomes increasingly important for individual real time monitoring of production, health, behaviour etc., in modern dairy cattle production systems with large herd sizes. Examples of camera based classification of lameness (Viazzi et al., 2013) and conformation traits (Salau et al., 2017) have already been presented. The aim of this study was to identify cows individually at the feeding table using a 3D camera system (Patent no: WO 2017/001538). The purpose of this identification was, with use of the same 3D camera system, to measure the feed intake for the identified cow (Lassen et al., 2017). However, the cow-id identification can also be used in combination with other features. A 3D geometric cow model with corresponding cow-id was used as reference. All cows in this study were labelled with unique cow-id, making it possible to calculate the success rate for 3D cow-id identification system. Material and methods Hardware Setup The input hardware setup consists of two main units. The reference unit responsible for acquisition of the reference cow geometries with a corresponding cow-id. A prediction unit responsible for acquisition of the cow geometry in the area where the identification process takes place. The reference unit consists of a single 3D camera using Time of Flight technology (Microsoft Xbox One Kinect v2) to create a 3D image and a RFID reader (Agrident Sensor ASR550). These were installed in a narrow corridor with a trigger system which ensured that one reference image was obtained from each cow when they pass through. The 3D camera was placed directly above the passing cows 3.4 m above floor level. The prediction unit consisted of 19 3D cameras (Microsoft Xbox One Kinect v2) placed in a line directly above the feeding area with a distance of 4.2 m to the floor. This setup ensured complete coverage of the entire 50 m long feeding table area in the test facility. The prediction unit records an image every 5 seconds. Feature estimation The first step in the identification process is to estimate features from the geometric information in the 3D images, which are useful for separating the individuals. This procedure
Download PDF file