Visual scores or subcutaneous fat thickness measurements, which is better for designing a 3D vision-technology body condition assessment method

Eddie Bokkers, Elles Leijdekkers, X. Song, J.J.P. van der Tol

Research output: Chapter in Book/Report/Conference proceedingAbstract


Assessing body condition (BC) of dairy cows is a way to collect information about their fat reserves, which is essential for effective feeding management. 3D vision-technology might offer the opportunity to automatically and frequently assess BC. Daily monitoring provides a short feedback-loop of changes in fat reserves. Hence, feedings strategies can be optimized which is beneficial from an economic (lower costs), environmental (lower emissions) and animal welfare (lower disease incidences) perspective. This study aimed to determine whether visual scores or subcutaneous fat thickness measurement is better to design
a system using 3D vision-technology to predict BC scores. Cows (n=46) ranging in BC score on a 5-point-scale from 1 (lean) to 5 (fat) were preselected on four farms. For eight body parts,
BC scores (1-5) given by two observers independently via visual observation and subcutaneous fat thicknesses (mm) measured by ultrasound were related with corresponding 3D variables. The three methodologies were conducted twice on two consecutive days. Only two body parts, i.e. sacral ligament and line between spinous process and tips of short ribs could be assessed by all three
methodologies (ultrasound was limiting). The inter- and intra-observer agreement (kappa) for visual observation ranged from 0.52 to 0.71 and 0.50 to 0.69 for the two body parts respectively. Correlation coefficients between days for the two body parts were 0.72 and 0.88 for subcutaneous fat thickness and 0.91 and 0.96 for 3D vision variables. Furthermore, 3D vision variables
of the two body parts correlated better with visual observations (range 0.79-0.88) than with subcutaneous fat thickness (range 0.55-0.69). Despite the high precision of the ultrasound device, this method appeared labour intensive and only delivered useful information for two body parts. Instead visual observations were made on all eight body parts and showed better potential for predictive BC models using 3D vision variables.
Original languageEnglish
Title of host publicationAgEng Conference 2018: New engineering concepts for a valued agriculture
Subtitle of host publicationbook of abstracts
Number of pages1
Publication statusPublished - 8 Jul 2018
EventAgEng Conference 2018: New Engineering concepts for a valued agriculture - Wageningen, Netherlands
Duration: 8 Jul 201812 Jul 2018


ConferenceAgEng Conference 2018
Internet address

Fingerprint Dive into the research topics of 'Visual scores or subcutaneous fat thickness measurements, which is better for designing a 3D vision-technology body condition assessment method'. Together they form a unique fingerprint.

Cite this