Development and computer-assisted validation of a radio frequency identification system for tracking individual chicken visits to functional areas

Abstract

Understanding how laying hens interact with functional resources-such as drinkers, feeders, perches, nest boxes, and wintergardens-is essential for meeting their physiological needs and enabling species-specific behaviors. This knowledge is crucial for poultry welfare assessments and precision livestock management. However, traditional ethological data collection methods, including direct observation and manual video analysis, are labor-intensive, prone to observer bias, and impractical for individual-level tracking. To overcome these challenges, we developed and validated an RFID-based system for automated, non-invasive tracking of individual hens' visits to key resources, using an established ArUco-based video annotation system as the reference standard. For validation, twenty-one laying hens were fitted with RFID leg bands and 3D-ArUco markers and monitored over five days in a mobile barn setup equipped with ultra-high-frequency RFID antennas. Alignment between data from the RFID and 3D-ArUco systems allowed calculation of performance metrics such as the F1-score-defined as the harmonic mean of precision and sensitivity-for visit durations and event detections (i.e., entries and exits), and the coefficient of determination (r²) for visit counts. Wintergarden showed the highest performance (84 % F1-score, 93 % r²). Metal perch achieved F1-scores of 79 % and 86 % for access and leaving events. Nest boxes showed intermediate performance (78 % F1-score, 77 % r²), while drinkers and feeders were lower (64 % F1-score each; r² values of 69 % and 49 %). These findings confirm RFID's potential for tracking visits to wintergardens, perches, and nest boxes-demonstrating sufficient performance for practical use, though further optimization through antenna positioning remains possible. For feeders and drinkers, however, accurate tracking remains challenging, and complementary technologies may be required, as rapid movements reduce tag dwell time, overcrowding causes signal interference, and open areas increase misreads from nearby surrounding movement. This study highlights RFID's value for behavioral research at the individual level in poultry and supports research-driven innovation in housing equipment design. It also demonstrates how a computer-assisted approach can facilitate validation across diverse behavioral contexts.

Keywords: ArUco markers; Automated behavioral tracking; Camera detection; Precision livestock farming; RFID tracking; Technology validation.

Fig. 1

System architecture of the RFID system to monitor individual chickens' presence within functional areas of key resources. The blue line with two-headed arrows represents the coaxial cables connecting the reader and the antennas (Ant.). The dashed arrows correspond to the ethernet cable connecting the reader and the host computer.

Fig. 2

An RFID tag is attached to a standard chicken leg ring and fixed to the ankle in a flag-like position, perpendicular to the tarsus. The RFID tag and attachment point are marked with a red circle.

Fig. 3

Placement of antennas and expected signal distribution (marked in red) along them to register chickens in the resources' functional areas. The figure shows in A: Antenna covering the drinking line; B: Two antennas in the feeding area; C: Two antennas for registering Barn-to-Wintergarden and vice versa; D: Antenna for the metal perch line; E: Antenna deployed across nest boxes.

Fig. 4

The area of interest (AOI, non-blurred zone) for 3D-ArUco annotations used as the ground truth for validating RFID annotations. Markers on the chickens present are automatically identified, and their IDs are displayed from the moment they enter and throughout their presence in the AOI. The figure shows in A: AOI defined for the drinking line; B: Two AOIs in the feeding area; C and D: AOIs covering Barn-to-Wintergarden and vice versa, respectively; E: AOI defined for the metal perch line; F: AOI defined for nest boxes.

Fig. 5

Effectiveness of bout criteria application on RFID visit data. The average presence duration according to the ArUco-based annotation (73 seconds for the feeder, 153 s for the drinker) was used to reduce fragmented, short-duration visits interspersed with RFID gaps. N_Aruco and N_RFID refer to the number of visits recorded by each approach, while mean_Aruco and mean_RFID represent their respective average durations (in seconds).

Fig. 6

Correlation between total ArUco and RFID visit counts per chicken over the entire data collection period. “r” represents Pearson's correlation coefficient and “r-squared” is the coefficient of determination.