COMPUTERS AND ELECTRONICS IN AGRICULTURE, vol.135, pp.4-10, 2017 (SCI-Expanded)
In this study, a new and non-invasive method was developed to automatically assess the lameness of broilers. For this aim, images of broiler chickens were recorded by a 3D vision camera, which has a depth sensor as they walked along a test corridor. Afterwards, the image-processing algorithm was applied to detect the number of lying events (NOL) based on the information of the distance between animal and the depth sensor of 3D camera. In addition to that, latency to lie down (LTL) of broilers was detected by 3D camera. Later on, the data obtained by proposed system were compared with visually assessed manual labelling data (reference method) and the relation between these measures and lameness was investigated. 93% of NOL were correctly classified by the proposed 3D vision camera system when compared to manual labelling using a data set collected from 250 broiler chickens. Furthermore, the results showed a significant correlation between NOL and gait score (R-2 = 0.934) and a significant negative correlation between LTL and gait score level of broiler chickens (R-2 = -0.949). Because of the strong correlations were found between NOL, LTL and gait score level of broilers on the one hand and between the results obtained by 3D system and manual labelling on the other hand, the results indicate that this 3D vision monitoring method can be used as a tool for assessing lameness of broiler chickens. (C) 2017 Elsevier B.V. All rights reserved.