Álvaro García
Feed costs represent a significant portion of operational expenses in milk production, with even small savings having the potential to substantially enhance farm profitability. One of the interesting aspects of feed efficiency is its heritability, which opens the door to the possibility of selecting animals for this valuable trait. However, the dairy industry has long grappled with a persistent challenge—the absence of individual cow data recorded on commercial farms, particularly throughout the entire lactation period.
Traditionally, the tracking of individual feed intake has predominantly occurred within academic settings as part of research projects. These efforts often involved labor-intensive methods, such as manually weighing offered and refused feed. In more advanced cases, automated feeding systems were employed to monitor feed consumption. Yet, this landscape is evolving, with emerging technologies like 3D cameras making remarkable strides. These advanced camera systems offer cost-effective solutions for continuous and precise data collection. What sets them apart is their versatility, extending beyond the mere tracking of feed intake. These cameras facilitate enhanced management practices in dairy cattle production, including real-time monitoring of body weight and body condition scores.
One advantage of these innovative technologies is their non-disruptive nature. They minimize stress for cows and maintain their daily routines, making the data collection process seamless and animal friendly. As a result, these technological innovations have the potential not only to revolutionize livestock management but also to reshape the genetic evaluations of dairy cattle.
In a recent experiment conducted by Lassen et al. in 2023, a single 3D camera, in conjunction with a Radio Frequency Identification (RFID) reader, was employed to identify individual cows. These components were strategically placed within a narrow alley, incorporating a time-based trigger system for precise data capture. This system effectively assigned all images to their respective cow ear tags, ensuring that a single reference image was obtained from each cow as they passed through the alley. To minimize the risk of multiple cows passing close to or turning around, the passage was intentionally narrowed beyond the dimensions of a standard exit lane. The 3D camera was strategically positioned directly above the passing cows at a height of 3.4 meters from the floor. Simultaneously, a custom walking scale was installed to measure the individual body weight of each cow. It’s important to note that direct sunlight exposure in the images can pose challenges, emphasizing the need to install these systems indoors under a roof to avoid or minimize direct sunlight interference.
Upon departing from the milking parlor, the electronic ear tags of the cows were read, and 3D images of their backs were simultaneously captured. These images served as a reference for predicting the same cow based on the distinctive contours, color, and patterns of their backs while they were at the feed bunk. This predictive process was accomplished using specialized algorithms. The effectiveness of this approach was rigorously validated across three different cattle breeds: Holstein, Jersey, and Red Dairy Cattle. Specifically, for the Holstein breed, 6575 images from 101 cows were utilized; for Jerseys, 8825 images from 129 cows were employed, and for Red Dairy Cattle, 3897 images from 155 cows were included in the study. Data was meticulously collected from 19 commercial dairy herds in Denmark, encompassing Jersey (6 herds), Holstein (7 herds), and Red Dairy cattle (RDC; 6 herds). The study incorporated a total of 9,142 cows across the Jersey, Red Dairy Cattle, and Holstein breeds, distributed among different herds. The validation process involved a meticulous comparison between the actual cow IDs and those predicted by the algorithm and was successfully executed across all three breeds.
Feed Intake System
The system utilized cameras positioned 2.5 meters apart and situated 4.5 meters from the unoccupied feed bunk, ensuring comprehensive coverage of the feeding area. Each camera’s data underwent initial preprocessing on a computer, involving the calculation of a median picture captured at 5-second intervals. The process for identifying and predicting the cow IDs was triggered when a cow approached the feed bunk. An image of the feed bunk was captured both before and after a cow’s meal, with the height difference of the feed between these two images serving as a metric to quantify the amount of feed consumed. At the end of each day, any remaining feed, accounting for less than 3%, was redistributed to cows based on their time spent eating. Each visit generated a record of five key variables: cow ID, location within the barn, meal initiation and completion times, and the quantity of feed consumed. The feed volume was determined by measuring the height of the feed pile on the barn floor, which was then converted from liters to kilograms by factoring in the feed density. Weekly averages were calculated for both feed intake and body weight based on daily measurements, serving as valuable phenotypic data. The results from a rigorous validation study conducted across all three breeds over a two-week period demonstrated accurate identification in over 99% of visits, regardless of the breed. The system’s performance, however, is contingent upon the quality of the cow’s ear tags, and if necessary, ear tags can be readily replaced to maintain optimal functionality.
Feed Intake Measures
The results showed that mean daily feed intake varied, with higher intakes in Red Dairy Cattle (61.7 kg) and Holstein (64.6 kg) compared to Jersey cows (55.7 kg). Variability in feed intake was mainly due to scaling, with a similar coefficient of variation for all three breeds. Repeatability estimates for daily feed intake were 0.62, 0.65, and 0.63 for Jersey, Red Dairy Cattle, and Holstein, respectively. The repeatability was consistent with previous studies based on research farm data, specifically related to dry matter intake. These systems typically measure intake in kilograms and multiply it by an assumed or measured dry matter percentage. This 3D camera-based system estimated heritability for dry matter intake between 0.23 and 0.34 across breeds. These findings demonstrate the potential of 3D cameras for measuring feed intake in dairy cattle, providing valuable insights into daily intake patterns and repeatability. Average body weight also differed between breeds, with higher values for Red Dairy Cattle (RDC) at 647.9 kg and Holstein at 683.8 kg, compared to Jersey cows at 469.6 kg. These values align with data from research farms. Repeatability estimates for weekly average body weight were 0.83 for Jersey cows, 0.85 for RDC, and 0.88 for Holstein, exceeding previous research farm findings.
The introduction of 3D cameras into dairy farming has the potential to bring about a significant transformation in the industry. These advanced technologies offer precise, non-disruptive solutions for data collection, enabling the monitoring of feed intake, body weight, and body condition scores. The experiment conducted by Lassen et al. in 2023 demonstrated the remarkable accuracy and adaptability of 3D cameras in identifying individual cows and tracking their activities. The value of 3D cameras extends beyond improved livestock management; it also has the power to reshape genetic evaluations for dairy cattle. With data collected from commercial dairy herds encompassing various breeds, these cameras have shown their potential in understanding daily intake patterns, repeatability, and heritability.
© 2024 Dellait Knowledge Center. All Rights Reserved.
