Prototype feedlot biometric, gender, and breed identification system

Summary

Effective production of beef cattle requires consistency. Given the diversity of cattle that attend feedlots, there is a need for new tools to assist sorting cattle at arrival, reimplant, or terminus to support profit maximisation. To date, no commercial sorting solution with appropriate embedded intelligence is available to the Australian feedlot industry.

An effective sorting system will require high-quality knowledge of feed intake, carcase growth, and composition of individuals, especially biometric measurements. Advancements in machine vision and learning technologies mean that it ought to be possible to predict growth and carcase composition accurately.

Research into this opportunity is considered vital for Australian lot feeders as accurate performance predictions may bring value through any one of several modes, namely, profit/loss modelling of cattle of different biological types at different carcase endpoints; categorisation of cattle into homogenous marketing groups (if critical mass of cattle present); optimising days on feed of the sorted group to maximise carcase value over production costs; accurate diet formulation to maximise performance of each pen; and/or most simply, improved eating quality of the produced beef for consumers.

Against these considerations, the current investigation enabled the development and validation of a prototype biometric, gender, and breed identification system suitable for use within a feedlot operation. The system was developed against defined objectives focused on autonomy, accuracy,
precision, and update rate. After achievement of prototyping milestones, including development of an appropriate truthing strategy, performances were evaluated within a commercial feedlot environment.

Objectives

This project aimed to develop a prototype automatic biometric measurements system; determine the precision, accuracy, and speed of result of the prototype to predict biometric measurements for a sample of independent cattle from three breeds (British, European, and Brahman) and determine possible biometrics associated with gender and breed identification.

Key findings

- It was demonstrated that the prototype system provided very encouraging results, predicting biometric measurements accurately, repeatedly, and quickly.
- Gender was not assessed during the experiment as only steers were available at the experiment’s host site.
- Breed identification was only moderately accurate; however, this report also considers further improvements that can be explored to further enhance measurement outcomes against this criterion.

Benefits to industry

This project has demonstrated it is possible to objectively capture cattle biometrics and gender to parameterize cattle sorting models currently under development.

MLA action

The results of this study have been shared with the Australian Lot Feeders' Association Research & Development committee to inform strategies for carcase endpoint management.

Future research

Future research will develop calibration data sets to associate cattle biometrics with breed (determined from genetic analysis) and mature size of animals (determined from body composition analysis). Once these relationships are developed this will parameterize cattle sorting models.