Crossbow's Mote products have been used for so many diverse and interesting applications. One of the most interesting deployments was done in 2005 using a wireless system to monitor cane toads. You may wonder why anyone would want to monitor toads, so here is some background on this little problem.
The amphibious assault and invasion of the cane toad began in 1935 when the toad (native to South America) was introduced to the sugar cane fields of Queensland in Northern Australia to eat a beetle that was damaging the state's sugar-cane plantations. The experiment was a failure as the cane toads ignored the beetles and began chomping their way through other wildlife from frogs and tadpoles to small lizards. Even worse, the poisonous glands on their backs made them deadly to the crocodiles, mammals, snakes and birds that tried to eat them (a mouthful of toad could be fatal to a dog or cat). Even the tadpoles were poisonous to native animals. The toads adapted quickly to the heat and humidity of tropical Queensland and within decades moved South and West, and finally overran the world famous Kakadu National Park - a World Heritage site. While the region of Darwin was already besieged, millions of other toads were converging on Western Australia. Females could lay up to 35,000 eggs at a time making efforts to head them off difficult. The federal
government looked into cane toad control research through its scientific agency, the Commonwealth Scientific and Industrial Research Organization (CSIRO) to find pathogens or other agents to wipe out the pests.
The toads can not be rounded up like cattle, but must be caught one by one - it is only through volunteer efforts that Australians can stop this amphibian invasion. To aid these efforts, a collaboration between the University of New South Wales, Portland State University and National ICT Australia aimed to design a wireless sensor network that could work unattended. In past deployments, researchers from UNSW used PDA class, disconnected devices. The prototype developed in this project provided a cheap and scalable alternative using networked sensor motes. The wireless acoustic sensor network used automatic recognition of animal vocalizations to census the populations of native frogs and the invasive toads. The wireless sensor network was designed to recognize vocalizations of up to 9 frog species found in northern Australia. The system used Crossbow's Stargate platform and MICA2 Mote products. The MICA2 Motes were used to collect acoustic samples and expand the sensor network coverage while the Stargate were used for resource-intensive tasks such as FFTs and
machine learning.
The development of the frog vocalization algorithm provided a key feature for the deployment. Acoustic features in the time and frequency domains could be used to distinguish the vocalizations of different amphibians. Frog vocalizations are much simpler than human speech, but they must be recognized in very difficult conditions (wind, rain, insects, and other prevalent noise). The algorithm examines each slice of the spectogram to extract attributes of the individual species being targeted by using pre-set classifiers. The use of this algorithm with the wireless sensor network was to pinpoint the regions inhabited by cane toads and to
track their macro movement directions as the system was deployed in boundary regions. Researchers used a hybrid network of Stargate and MICA2 Motes to make the system cost-effective. The MICA2 Motes were scattered to collect acoustic samples while the resource-rich Stargate platform was used to run the FFT algorithm and other functions required. The MICA2 Mote platform performed preliminary processing on the samples to reduce the transmission size and environmental noise of the data sent to the Stargate. The Stargate would then use these inputs to determine the
existence of frogs and pinpoint their location. The macro movements were estimated by comparing the cane-toad existence snap shots at different times by using the location of the sensor device that detected the existence of the species through vocalization. This location information proved to be more than adequate for tracking the cane toad's long-term migration patterns. Although the system would sometimes confuse species, it would never give incorrect results for the cane toad species (the principal species being detected) since it has a very different vocalization compared to the other species found. This type of detection helped volunteers determine which areas they needed to target, and which areas could be left alone allowing for workers to be used more efficiently rather than spending time doing broad sweeps of the area in search of the toad.
Deployments such as these show the vast capabilities and applications into which the Mote platforms may be deployed. Their flexibility and ability to provide a low-cost solution for remote sensing applications presents situations where technology can enable us to understand and monitor our physical environment in greater detail.
Comments