Abstract
Abstract: Lameness in dairy cattle represents a significant problem impacting animal welfare and causing economic losses in the dairy industry. This paper presents an evaluation of the effectiveness of two techniques for the early detection of lameness. In the first methodology, a set of features is extracted from the cow's video, including Leg movement variance, Path regularity standard deviation, Average Step count, Average back curvature, and Cluster ratio. These features are then compared with experimentally obtained threshold values to determine the presence or absence of lameness. In the second methodology, a set of sensors MPU5060 and ultrasonic is used to measure angles and distances. Based on comparing the readings from these sensors with experimentally obtained threshold values to determine the presence or absence of lameness. The results showed that both approaches were highly effective in identifying cases of lameness. The computer vision technique achieved a diagnostic accuracy of 94.44% in identifying affected cases, while the sensor-based system demonstrated an accuracy of 93%. The study proves the feasibility and applicability of computer vision algorithms in farm visual pattern analyzing. The sensors are useful in collecting accurate, quantifiable kinematic information for evaluation. The results confirm and affirm the validity and applicability of both methods as useful tools in lameness recognition. This enhances herd health and productivity.