ACOUSTO-FLUIDIC DETECTION OF INSECT LARVAE IN GRAIN

Abstract

DRT and USDA have entered into a Cooperative Research and Development Agreement (CRADA) to exploit the high sensitivity and signal-to-noise ratio of acousto-fluidic sensors to detect and quantify insects in grain and stored products. ARS has developed a technique for acoustically quantifying insects (including internal feeding larvae) in grain samples. ALFID (Acoustic Location Fingerprinting Insect Detector) detects individual insect sounds (chewing, moving) and cross-correlates them to determine relative time delays between sixteen sensors. This data, arranged in vectors (fingerprints) which indicate sound source locations, is compared and grouped thereby identifying the number of insects. ALFID effectiveness has been limited due to the poor SNR of conventional microphones. Initial experiments with acousto-fluidic sensors, that physically increase the sounds so that they can be detected by low cost electret microphones, have been very promising. SNR improved by 2x and insect detection probability improved by almost 3x. In three separate experiments, more than half the rice weevil larvae (Sitophilus oryzae) detected acousto-fluidically were not discerned conventionally. A 3-stage acousto-fluidic sensor (acoustic gain - 56dB, bandwidth 1 - 7kHz) driving two Tibbetts 251-01 electrets, is able to process sound levels of less than OdB SPL (ref. 0.0002, mbar). Two-week old larvae in a 1 kg grain sample have been detected at distances of over 10cm. Production cost of one sensor will be about $100, including electronics and pneumatic supply (small aquarium pump). Inadvertent export of infested grain (grain visually passed as clean) accounts for multi-million dollar annual losses. It is anticipated that the use of an acousto-fluidic ALFID system should greatly reduce these losses. Monitoring of insect infestations in silos and packaged goods will also be possible. Also, this technology may help advance infestation quantification using large arrays of multiplexed sensors coupled with ARS-developed statistical insect population estimates.