Use this URL to cite or link to this record in EThOS:
Title: 3D video based detection of early lameness in dairy cattle
Author: Abdul Jabbar, Khalid
ISNI:       0000 0004 6497 0268
Awarding Body: University of the West of England
Current Institution: University of the West of England, Bristol
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Lameness is a major issue in dairy cattle and its early and automated detection offers animal welfare benefits together with potentially high commercial savings for farmers. Current advancements in automated detection have not achieved a sensitive measure for classifying early lameness; it remains to be a key challenge to be solved. The state-of-the-art also lacks behind on other aspects e.g. robust feature detection from a cow's body and the identification of the lame leg/side. This multidisciplinary research addresses the above issues by proposing an overhead, non-intrusive and covert 3-Dimensional (3D) video setup. This facilitates an automated process in order to record freely walking Holstein dairy cows at a commercial farm scale, in an unconstrained environment. The 3D data of the cow's body have been used to automatically track key regions such as the hook bones and the spine using a curvedness feature descriptor which operates at a high detection accuracy (100% for the spine, >97% for the hooks). From these tracked regions, two locomotion traits have been developed. First, motivated by a novel biomechanical approach, a proxy for the animal's gait asymmetry is introduced. This dynamic proxy is derived from the height variations in the hip joint (hooks) during walking, and extrapolated into right/left vertical leg motion signals. This proxy is evidently affected by minor lameness and directly contributes in identifying the lame leg. Second, back posture, which is analysed using two cubic-fit curvatures (X-Z plane and X-Y plane) from the spine region. The X-Z plane curvature is used to assess the spine's arch as an early lameness trait, while the X-Y plane curvature provides a novel definition for localising the lame side. Objective variables were extracted from both traits to be trained using a linear Support Vector Machine (SVM) classifier. Validation is made against ground truth data manually scored using a 1–5 locomotion scoring (LS) system, which consist of two datasets, 23 sessions and 60 sessions of walking cows. A threshold has been identified between LS 1 and 2 (and above). This boundary is important as it represents the earliest point in time at which a cow is considered lame, and its early detection could improve intervention outcome, thereby minimising losses and reducing animal suffering. The threshold achieved an accuracy of 95.7% with a 100% sensitivity (detecting lame cows), and 75% specificity (detecting non-lame cows) on dataset 1 and an accuracy of 88.3% with an 88% sensitivity and 92% specificity on dataset 2. Thereby outperforming the state-of-the-art at a stricter lameness boundary. The 3D video based multi-trait detection strives towards providing a comprehensive locomotion assessment on dairy farms. This contributes to the detection of developing lameness trends using regular monitoring which will improve the lack of robustness of existing methods and reduce reliance on expensive equipment and/or expertise in the dairy industry.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available