Flying Insect Detection and Classification with Inexpensive Sensors
October 15th, 2014
We proposed a system that uses inexpensive, noninvasive pseudo-acoustic optical sensors to automatically and accurately detect, count, and classify flying insects based on their flying sound.
The overall goal of this procedure is to detect and identify flying insects based on their wing beat sound. This is accomplished by first using a pseudo acoustic sensor to continuously record sound in an experiment chamber. The second step is to automatically detect the insect flying sounds from the raw recording.
The final step is to automatically classify the detected flying sounds to identify the insect species and sex. Ultimately, figures and tables are used to show the classification accuracy. The main advantage of this technical is that it's simple, inexpensive, and it can be made as ubiquitous as current methods such as sticky trips and intercession trips, but with digital advantages such as higher accuracy and real time monitoring.
To begin the experiment, rear mosquito larvae in enamel pans under standard laboratory conditions. Feed them a mixture of ground rodent chow and brewers'yeast in a three to one ratio. Next, collect mosquito pupi in 200 milliliter cups or collect mosquito adults and place them into experimental chambers.
Make sure each experimental chamber contains 20 to 40 individuals of the same species and sex. Eat adult mosquitoes, a 10%sucrose and water mixture, and replace the food once a week, moist and cotton towels twice a week, and place them on top of the experimental chambers to maintain humidity. In addition, keep a 200 milliliter cup of tap water in the chamber at all times.
Maintain the experimental chambers on a 16 to eight hour light to dark cycle at 20.5 to 22 degrees Celsius and 30 to 50%humidity for the duration of the experiment. Connect the experimental chamber to a power supply and turn on the power. Align the laser lights to the photo array by adjusting the magnets until the laser is centered on all the individual photo diodes.
Next, perform two sanity checks by first making sure that the system is powered and the laser is pointing at the photo array. Second, plug the speaker into the audio jack to check if the alignments of the laser and the photo array are capable of capturing sound. Plunge a hand in and out of the cage near the laser source end.
Make sure the laser light is on the hand and listen to see if the sound changes. As the hand goes in and out, attach a string to an electronic toothbrush and turn on the toothbrush. Make sure the laser light is on.
Then move the string into the laser plane and listen to see if the sound changes. After the system is properly set up, add insects to the cage and close the lid. To begin data collection, turn on the recorder and make a voice annotation that includes the following information and pause the recording.
Next, connect the recorder to the system and resume the recording for three days. Download the data from the recorder into a new folder and empty the recorder by deleting the data. Repeat the recording process until the remaining insects have died and there are no more than five insects left alive in the cage.
To process the data, run the circadian WBF detection software for each folder containing data from a recording session, open MATLAB and type in the command window to change the MATLAB working directory to the code directory and run the code circadian WBF. Then press enter to start. Once the algorithm terminates, check the detection results which the algorithm outputs in a new folder, the classifier achieves more than 96%accuracy when classifying no more than five species of insects as demonstrated by the table, which displays the classification accuracy obtained with increasing numbers of classes.
When the number of classes increases to 10, the accuracy is never lower than 79%This tannic can help to fight inated disease such as material by producing real information that can be used to effectively plan the separation program. It can also help the farmers to limit the spread of incident damages on their crops by monitoring the pests in the field.
View the full transcript and gain access to thousands of scientific videos
Overview
This study presents a system that utilizes inexpensive, noninvasive pseudo-acoustic optical sensors to detect, classify, and count flying insects based on their wing beat sounds. The method offers a digital alternative to traditional insect monitoring techniques, providing higher accuracy and real-time data.
Key Study Components
Area of Science
- Insect monitoring
- Acoustic sensing technology
- Species classification
Background
- Traditional methods for insect monitoring include sticky traps and interception traps.
- These methods can be limited in accuracy and real-time data collection.
- Advancements in acoustic sensing technology provide new opportunities for insect detection.
- Understanding insect populations is crucial for ecological studies and pest management.
Purpose of Study
- To develop a system for automatic detection of flying insect sounds.
- To classify detected sounds to identify insect species and sex.
- To demonstrate the advantages of this method over traditional monitoring techniques.
Methods Used
- Continuous recording of sound using pseudo-acoustic sensors in an experimental chamber.
- Automatic detection of insect flying sounds from raw audio recordings.
- Classification of detected sounds to determine species and sex.
- Use of figures and tables to present classification accuracy results.
Main Results
- The system successfully detected and classified various flying insect species.
- Results indicated higher accuracy compared to traditional methods.
- Real-time monitoring capabilities were demonstrated.
- The method is cost-effective and easy to implement.
Conclusions
- This study presents a viable alternative for insect monitoring.
- The proposed system enhances accuracy and efficiency in detecting flying insects.
- Future applications could lead to improved ecological monitoring and pest control strategies.
Frequently Asked Questions
Executive Industry Relevance
This work demonstrates a low-cost, noninvasive sensor system for detecting and classifying flying insects based on wing beat sounds, offering a scalable alternative to traditional entomological monitoring methods. The approach provides high-accuracy, real-time data that can support vector surveillance and pest management programs in agricultural and medical entomology. By enabling continuous monitoring with minimal infrastructure, the system enhances predictive confidence in early-stage target validation for interventions aimed at reducing disease transmission or crop loss.
Strategic Applications in Biopharma R&D
Early Discovery & Target Validation
- Scientific Value: Enables interrogation of insect behavioral phenotypes through continuous, noninvasive monitoring of wing beat patterns.
- Operational Value: Supports hypothesis testing in vector control research by providing quantitative, species- and sex-specific activity data.
- Predictive Value: Improves confidence in target selection by correlating insect activity with environmental or intervention variables.
Screening & Assay Development
- Scientific Value: Generates robust, reproducible acoustic signatures that serve as phenotypic readouts for insect classification.
- Operational Value: Standardizes data collection across experimental chambers, reducing variability in longitudinal studies.
- Scalability: Allows high-throughput monitoring of large insect cohorts over extended periods, as demonstrated in multi-day recordings.
Translational & Preclinical Research
- Translational Continuity: Bridges laboratory-based insect behavior studies with field-applicable surveillance tools for vector-borne disease models.
- Mechanistic De-risking: Enables real-time assessment of intervention impacts on insect populations, supporting go/no-go decisions in preclinical development.
- Disease-Relevant System: Provides a platform to evaluate vector competence and host-seeking behaviors under controlled conditions.
Pipeline & Workflow Integration
The method fits within the discovery continuum from early vector biology to preclinical evaluation of control strategies, particularly where real-time, species-resolved monitoring informs mechanistic understanding and intervention efficacy.
- Discovery Biology: Facilitates hypothesis-driven studies of insect flight behavior, circadian activity, and responses to stimuli.
- Screening: Delivers assay-ready, quantitative outputs (wing beat frequency, amplitude modulation) suitable for high-resolution phenotypic screening.
- Analytics: Enables statistical comparison of insect populations across conditions using classification accuracy and detection rates as key metrics.
- Translational Research: Supports continuity from lab to field by validating monitoring tools that can be deployed in endemic settings.
- Enterprise Reuse: Represents a reusable sensor platform adaptable to multiple insect species and environmental contexts.
Operational & Enterprise Impact
- Scientific Value: Increases predictive confidence in target validation by reducing ambiguity in insect phenotyping and behavioral responses.
- Operational Value: Delivers standardized, real-time monitoring with minimal consumables, lowering operational burden compared to sticky or interception traps.
- Strategic Value: Improves capital efficiency by enabling rapid iteration in vector control screening campaigns through immediate feedback on intervention effects.
- Portfolio Impact: Supports risk-adjusted prioritization of vector control candidates by providing early, measurable biomarkers of population suppression.
Implementation Considerations
- Requires expertise in bioacoustics, signal processing, and entomological assay design.
- Depends on stable optical and audio instrumentation, including laser-photodiode arrays and environmental controls.
- Necessitates cross-team standardization for data annotation, species verification, and classification model training.
- Adaptation across model systems may require recalibration of sensor sensitivity and feature extraction parameters.
- Practical limitations include sensitivity to ambient noise and environmental vibration, which may affect detection fidelity in field settings.
