High-precision scanning system for complete 3D goat udder and teat imaging, and analysis of morphological traits - INRAE - Institut national de recherche pour l’agriculture, l’alimentation et l’environnement
Journal Articles Small Ruminant Research Year : 2024

High-precision scanning system for complete 3D goat udder and teat imaging, and analysis of morphological traits

Abstract

Currently, udder scoring for goat udder phenotyping in France is made by a specialized technician trained to score 5 traits for the udder and 4 for teats, only one of which being measured (teat length). It is a hard work in big flocks, and only done one time in the life of goats. Recent works underlined a clear degradation of udder/teat shape with parity and an increasing proportion of morphological and functional unbalance of udders. Thus, for improved genetic selection, we must increase the number and quality of quantitative udder traits measured throughout the productive lifespan of dairy goats. To achieve this goal, we aimed to design a 3D Scanner device to scan the entire udder and to produce high definition numeric images. The device developed is easy to transport in a van and consists of a portable corridor cage in which the goats enter freely. Goats are then blocked less than 2 min while the images were taken. We used 4 Intel Realsense depth D415 cameras mounted on a mobile trolley situated 60 cm under and on the two sides of the cage and sliding from the front to the back of the animal in 10 s. The D415 has a baseline of 55 mm which allows the depth error to be less than 2% at 2 m. The device has an auto calibration internal software that ensure good evaluation of distances with one-millimeter depth precision at 50 cm. The reconstruction of the 3D Image from these 4 sources was done by the RecFusion (R) software. The image resolution obtained is 1280 * 720 pixels and real-time scene capture is done at up to 90 frames per second (900 images/camera/run/goat). The size of voxel generated was 2 mm. The first prototype was tested on 260 goats of different parity and udder shape and produced well-defined 3D images on which we can measure dimensions with Metrux2 alpha (R) software after manual positioning of references points. The first validation steps wad done on a soccer ball and on 10 goats for comparison of manual scoring and Image analysis. Machine learning algorithms are currently being developed for the purpose of automatic measurement on these images. The actual precision of the automatic measurement will be assessed following the completion of this final phase. Nevertheless, this scanner shows great promise for future implementation within the goat selection scheme, as it enables swift, objective, and quantitative evaluation of udder and teat characteristics.
No file

Dates and versions

hal-04477561 , version 1 (26-02-2024)

Identifiers

Cite

Marnet Pierre-Guy, Delattre Laurent, Delouard Jean Michel, Luginbühl Thibault, Laluque Thibaut, et al.. High-precision scanning system for complete 3D goat udder and teat imaging, and analysis of morphological traits. Small Ruminant Research, 2024, 231, pp.107194. ⟨10.1016/j.smallrumres.2023.107194⟩. ⟨hal-04477561⟩
23 View
0 Download

Altmetric

Share

More