https://www.wageningenacademic.com/doi/pdf/10.3920/978-90-8686-940-4_132 - Thursday, September 12, 2024 2:30:55 AM - IP Address:212.97.215.172 132. CFIT – Cattle Feed InTake – a 3D camera based system to measure individual feed intake and predict body weight in commercial farms J. Lassen 1,2* , J.R. Thomasen 1 and S Borchersen 1 1 Viking Genetics, Ebeltoftvej 16, Assentoft, 8960 Randers, Denmark; 2 Center for Quantitative Genetics and Genomics, Aarhus University, Blichers Alle 20, 8830 Tjele, Denmark; jalas@vikinggenetics.com Abstract A 3D camera technology to measure feed intake and body weight on individual cows in commercial farms has been developed. The camera based method for feed intake measurements is validated using scale measures in a commercial farm. In this study the squared correlation between camera data and scale measures on daily intake was 0.90. Body weight is predicted based on images of back couvertures and corresponding scale measures in three dairy breeds: Jersey (301 cows), Holstein (398 cows) and Red dairy cows (103 cows). Root mean square error between observed and predicted weight was 23.5, 31.0 and 22.7 kg and the repeatabilites were 0.86, 0.83 and 0.90 for Jersey, Red dairy cattle and Holstein respectively. Several management tools continue to be developed based on the data for the farmers that have the equipment installed and more traits are expected to be developed based on the 3D methodology in the future. Introduction Feed intake (FI) is the highest variable cost in the dairy farm. Therefore even marginal improvements in the amount of feed that is used to produce an amount of milk and meat will have a huge effect on the turnover for a farmer. Milk and meat production has been core traits in most breeding systems for several decades, but not until recently feed efficiency has been a trait that has been evaluated. This is not due to lack of value of the trait, but more due to lack of data on the trait at an individual level. Individual FI measures has only been available on cows from research farms, and often only for limited parts of lactation and in few lactations per cow. FI is a main driver for cow health and welfare. Most diseases occur in early lactation so if data is not available in this period it is very difficult to provide information that possibly could prevent diseases from developing. The system is currently installed in 20 dairy herds in Denmark where more than 10,000 cows are providing data. FI measures historically has been done using scale based measures. These systems are relatively expensive and labour intensive. Therefore they are also impractical for measurements in commercial herds. Over the recent years both 3D cameras, hardware such as graphic cards and algorithms around image analysis has been under drastic improvements. This includes among others artificial intelligence (AI) algorithms such as the convoluted neural network method called MASK-CNN (He et al. , 2018). Combining all these developments into a working system is very difficult. Since 2016 a 3D camera system has been developed to measure individual FI records and make body weight (BW) prediction at an individual cow level. In this paper the methodology as well as the system will be described and discussed. Materials & methods System and data description. In order to make individual FI and BW measures 3D cameras was used in a top down position (Microsoft Kinect). Each camera is placed 2.5 meter apart and 4.5 meter from the empty feeding table in order to cover the entire feeding table. Also, in the exit corridor of the milking system an ear tag reader and a 3D camera was placed. Three type of images were recorded from the 3D camera: R.F. Veerkamp and Y. de Haas (eds) Proceedings of 12 th World Congress on Genetics Applied to Livestock Production (WCGALP) 577 DOI: 10.3920/978-90-8686-940-4_132, © J. Lassen et al. 2022
Download PDF file