Summary
This project studied alternative technologies for automatically counting live sheep, particularly as they are loaded on and off ships. The technologies investigated included machine vision, radio frequency tags, and races that can singulate sheep mechanically. Australia presently exports about 5 000 000 live sheep each year, of which some 4,000,000 leave from the port of Fremantle.
At the time of this report, all these sheep were manually counted several times, not only at the ports of embarkation and disembarkation but also at feeding lots where sheep are kept for a number of weeks prior to export. Counting at the ports was done by tally clerks. For the technology to be appropriate to the Australian livestock export industry three criteria appeared to be essential:
Accuracy
The industry's principal motivation for automation is not the labour costs of tally clerks, estimated to be about 5 per sheep per count, but the lack of 100% accuracy inherent in human counting of large numbers of animals that superficially look alike. Of particular concern to industry is the discrepancy between the number of sheep counted when they leave Australia and when they arrive at their overseas destination. For an automated counting system to be adopted by industry it must be at least as accurate as the manual system it replaces. Therefore, machine vision systems installed overseas that count pedestrians at a claimed accuracy rate of 90% or 95% are unsuitable for this application, even if they could be adapted to count sheep instead of people. We conclude that only those technologies promising accuracy >99% are suitable for further development.
Cost
In contrast to cattle, horses and companion animals such as dogs and cats, the value of a sheep is low. Technologies that can be utilised cost-effectively to count higher-value animals therefore may be unsuitable for counting sheep. We conclude that only technologies promising a reasonable return on investment with respect to the current annual financial loss of $1 200 000 are suitable for further development.
Speed
Because of the cost of chartering ships and the comparatively high cost of Australian labour, it is imperative that automated counting methods not delay the loading or unloading of ships. We understand that sheep presently are loaded at rates as high as 9,000 sheep per hour = 150 sheep per minute at the port of Fremantle, although unloading overseas takes place at a much slower rate, typically 1,500 sheep per hour = 25 sheep per minute. Therefore, if a single automated counting system cannot accommodate at least 9,000 sheep per hour, multiple systems that work in parallel will be required. Although a single machine vision system potentially can accommodate arrival rates much faster than 25 per minute, or even 25 per second, mechanical systems are inherently more limited in their capacity. The total installation cost must offer an acceptable return on investment. The research reported above has convinced us that, although machine vision conceivably could be developed (with considerable effort) to solve the automated sheep-counting problem, mechanical singulation technology appears to offer, in the short term, the promise of achieving an affordable, accurate solution to this problem. No other technology that we have examined offers this promise. Accordingly, it is our recommendation that the SLAMP Limited system described in the full report be developed and extended, tested and then trialed.
