Prototype on-farm 3D camera system to assess traits in Angus cattle

Summary

Why the work was done
This project builds upon research work carried out in a 'proof of concept' (B.BSC. 0339) study of 3D comers to assess hip height, P8 fat and muscle score and a pasture fed compliance project (B.SBP.0108) assessing market specifications. The development of a 3D camera technology has the potential to revolutionize the beef industry by providing real time objective measurements to assist producers improve management decisions. The new 3D technology will also provide the beef industry with a method of collecting large quantities of data to assist in making breeding decisions.

How it was done
Working in collaboration with the University of Technology Sydney we have developed software that creates a feature vector from 3D images to assess hip height, P8 fat and muscle score on live cattle within 15 seconds. A machine learning technique is employed on a ground truth dataset of ultrasound P8 fat (mm) and assessed muscle score by a trained assessor. Additional feature vectors were evaluated along with a Convolution Neural Networking technique to improve the assessment of P8 fat.

What was achieved?A light weight auto-calibration apparatus called the “GUBE” was developed.Detection of stability of 3D image generally found within 1 minute.Demonstration that software has been extensively developed to perform the feature extraction and assessment of P8 fat and muscle score from machine learning.Improved the assessment of P8 fat through:Exploring alternative feature vectorsExploring Convolution Neural NetworksDeveloped a prototype measurement chuteConducted two trials on cattle that had not been seen by the cameras.

Industry benefits – results and implications70% of assessed P8 fat using the 3D camera technology lies within an accuracy of ± 1.5mm97% repeatability (n = 30) across 4 sessions for assessment of P8 fat within ± 1.5mm77% of assessed muscle score using the 3D camera technology lies within an accuracy of ± 1 on a 15 point scale80% repeatability (n = 30) across 4 sessions for assessment of muscle score within ± 1 on a 15 point scale3D technology reduces variability of subjective measurementsOne off cost for 3D technology – quick return for investmentObjective measurements on live cattle to revolutionize the beef industry