SPIE Membership Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
  • Information for:
SPIE Photonics West 2019 | Register Today

SPIE Defense + Commercial Sensing 2019 | Call for Papers

2019 SPIE Optics + Photonics | Call for Papers



Print PageEmail PageView PDF

Sensing & Measurement

Animal management in the Australian rangelands

Machine vision is used to remotely control water access by wildlife and livestock.
11 January 2008, SPIE Newsroom. DOI: 10.1117/2.1200801.0960

Invasive vertebrates and overabundant native species cause significant economic and environmental damage in the Australian rangelands, which cover more than 70% of the continent. Some estimates put the annual loss in livestock production due to such pest species at more than $700 million.1–3

Access to pastoral watering points has been a major factor in the spread and survival of pest species. While methods of controlling access watering points have been developed,4 they are problematic. Essentially all current trap systems rely on one-way barriers and cannot discriminate between pest and desired species.

To address this problem, we have developed a system of identifying animal species using machine vision, which uses computers to process images and then applies the resultant information to manipulate equipment. In recent years such systems have become more common in remote-sensing applications for all aspects of the agricultural domain5 as well as many areas of wildlife management.6 When used in conjunction with an enclosure and an automated gate, our system can control animals’ access to a limiting resource, like water,7 and thus function to exclude pest species, trap desired species, and otherwise manage livestock. It requires no physical contact with the animals, reducing both system maintenance requirements and animal stress. Computing capabilities have now developed to the point that processing of real-time video data is also feasible.

The concept that governs the system is simple: control the water (or other resource), and you control all large vertebrates (>10 kg) that require water to survive. Three main components achieve this goal.

First, an enclosure with an entry lane surrounds the limiting resource. For the machine vision software to accurately identify animals to the species level, it must view the animals laterally and in single-file arrangement. Our enclosure and entry lane were constructed of galvanized weld mesh (100mm x 100mm x 5mm) at a height of 1.5m to create a barrier against the species most likely to be managed using this technology in the Australian landscape (namely feral pigs, kangaroos, emus, sheep, goats, horses, and cattle).

Second, an automated gate opens or closes to control access to the resource (see Figures 1 and 2). Our automated gates (hinged in a frame 2.2m high and 2.3m long and constructed from 40mm x 40mm galvanized section) were air-operated and based on a truck-brake booster ram. The gates operated on compressed air supplied by a 12V compressor and a 60L air tank compressed to 150psi.

Figure 1. A watering point enclosure excludes some species while allowing access to others. The gate remains open, closing when undesirable animals are detected in the entry lane, then reopening after a predetermined period of time.

Figure 2. A watering point enclosure such as this one can also function as a trap, closing to keep detected desired species inside.

Finally, an intelligent camera system identifies animal species. Machine-vision software analyzes images in real time using a system of algorithms that detect the outline of a moving animal (see Figures 3 and 4).8 A blue background in the entry lane enables the vision software to distinguish a passing animal as whatever component of the image that isn't blue. At night, the software uses movement to detect the outline of the animal. Edge tracking and silhouette encapsulation matches an animal's outline shape against a normalized library. The software processes video input in real time by comparing individual frames to a database.

Figure 3. Video image of a goat (left) and processed image (right) showing the digitized detected edge, the boundaries of the animal, and the matched template underneath.

Figure 4. Video image of a sheep and processed image.

We have developed a stand-alone field camera with low power consumption (0.125A, 6V) and infrared lighting. Our software identifies species with greater than 95% accuracy and could be applied in any habitat in which a resource is limited and can be enclosed for management purposes.

We gratefully acknowledge funding from the Natural Heritage Trust of the Australian Federal Department of Agriculture, Fisheries and Forestry, and technical and financial assistance from RPM Rural Products, Queensland, Australia.

Neal Finch, Peter Murray
School of Animal Studies
University of Queensland
Gatton, Queensland, Australia  

Peter Murray is a senior lecturer in the School of Animal Studies at the University of Queensland. His research and teaching focuses on the management, utilization, and ecology of wildlife, pest, and production animals.

Mark Dunn, John Billingsley
National Centre for Engineering in Agriculture
University of Southern Queensland
Toowoomba, Queensland, Australia