Science! By Ryan Whitwam Mar. 19, 2014 5:29 pm
There are untold millions of insect species in the world, many of which we have yet to identify. Entomologists have always had to painstakingly identified their quarry by hand, but a device built at University of California Riverside could change all that. According to lead researcher Yanping Chen, the new laser-based system can identify an insect from several meters away as it flies past.
At the heart of this technology is a recording of the frequency of the insects’ beating wings. Efforts in the past have been able to use wing frequencies for a limited sort of identification, but these techniques always relied on microphones to listen to the sound. The problem being that sound intensity drops by the inverse square law — twice as far away, the noise is only a fourth as intense, for example. Even using more sensitive mics couldn’t make this approach workable. You’d only get more background noise, which makes it impossible to gather enough samples for a data training set.
Chen and colleagues have gotten around this by ditching microphones entirely. Instead, this contraption relies on a laser beam shining onto an array of phototransistors. When an insect crosses the beam, it causes a fluctuation in the signal with the beating of its wings. The signal is then encoded as an MP3 file, and you’ve got a recording of the frequency without using acoustic sound. The laser recording system allows researchers to overcome the largest problem with automated insect identification — getting a large enough data set.
The device was placed in cages with certain species of mosquito to accumulate millions of known samples that can be compared to unknowns later for matching. The picture the system has of each species needs to be very detailed because it’s more than the frequency, it’s harmonics as well. Two different insects might have overlapping frequencies, but the combination of frequency and harmonics makes them much easier to tell apart.
After all the baseline data was gathered, the University of California device was able to correctly identify the six species of test insects with 97% accuracy from several meters away. Scientists hope to use this method to better track insect populations and control mosquitoes without sweeping up harmless insects in the process. The team thinks this is so important, they are handing out a complete system to any research entomologist who wants one.
More...
At the heart of this technology is a recording of the frequency of the insects’ beating wings. Efforts in the past have been able to use wing frequencies for a limited sort of identification, but these techniques always relied on microphones to listen to the sound. The problem being that sound intensity drops by the inverse square law — twice as far away, the noise is only a fourth as intense, for example. Even using more sensitive mics couldn’t make this approach workable. You’d only get more background noise, which makes it impossible to gather enough samples for a data training set.
Chen and colleagues have gotten around this by ditching microphones entirely. Instead, this contraption relies on a laser beam shining onto an array of phototransistors. When an insect crosses the beam, it causes a fluctuation in the signal with the beating of its wings. The signal is then encoded as an MP3 file, and you’ve got a recording of the frequency without using acoustic sound. The laser recording system allows researchers to overcome the largest problem with automated insect identification — getting a large enough data set.
The device was placed in cages with certain species of mosquito to accumulate millions of known samples that can be compared to unknowns later for matching. The picture the system has of each species needs to be very detailed because it’s more than the frequency, it’s harmonics as well. Two different insects might have overlapping frequencies, but the combination of frequency and harmonics makes them much easier to tell apart.
After all the baseline data was gathered, the University of California device was able to correctly identify the six species of test insects with 97% accuracy from several meters away. Scientists hope to use this method to better track insect populations and control mosquitoes without sweeping up harmless insects in the process. The team thinks this is so important, they are handing out a complete system to any research entomologist who wants one.
More...
