Development of an early detection system for lameness of broilers using computer vision

Abstract

Lameness is one of the most important causes of poor welfare in poultry. Previous studies have documented approximately 30% of the chickens were seriously lame. In this research, a novel technique was developed for early detection of lameness in broilers. For this purpose, broiler chickens with five different predefined gait scores were continuously monitored by a digital camera as they walked throughout a test corridor. Then, image analysis algorithm was applied to detect some feature variables (speed, step frequency, step length and the lateral body oscillation) of broilers. Afterwards, a correlation test was performed to define the coefficient of correlation between the feature variables (step frequency, step length, speed and LBO) obtained by the proposed algorithm and the gait score levels of the birds, which respectively resulted in r=0.831, 0.882 (p<0.05), 0.844, 0.861. Furthermore, each feature variable was investigated to find statistical differences between gait scores (as a measure of lameness) of broilers. It was performed to assess the effects of gait score on speed, step length, step frequency and lateral body oscillations of the broilers. The results showed that all investigated feature variables were efficacious to detect lameness in broilers starting from GS3. Since correlations were found between the feature variables (step frequency, step length, speed and LBO) obtained by the proposed algorithm and the gait score levels of the birds on the one hand and the statistical differences between gait score levels of broilers on the other hand; the results recommend that this fully-automated detection system has potential to be used as a real-time monitoring tool for early detection of lameness in broilers starting from GS3. However, to define lower gait scores than GS3, either new feature variables like foot curls and wing use should be inserted into the proposed system or this system should be combined with other automatic behaviour analysis tools for early detection of lameness in future research. It is very important to detect lameness at an early stage because it allows farmers and veterinarians to take immediate management actions in time.