Mosquito-borne diseases pose serious risks to human health. The World Health Organization estimates that millions of people worldwide are infected annually. An essential aspect of controlling and preventing mosquito-borne diseases are reduction of mosquito density. Further, human-assisted mosquito surveillance methods are time- consuming, labor-intensive, and require skilled manpower. Their identification is crucial to take the necessary steps to kill them in an area. Accurate species identification is crucial for effective mosquito control programs. AI-enabled mosquito surveillance framework using our developed robot, named ‘Dragonfly’, which uses the ResNet-50 Deep Neural Network (DNN) algorithm. The Dragonfly robot is designed with a drive mechanism and a mosquito trapping module to attract mosquitoes within the environment.
The Resnet-50 will be trained with mosquito classes, namely Aedes aegypti, Aedes albopictus, and Culex, to detect and classify the mosquito breeds from the mosquito glue trap. The classifier is capable of recognizing 70 mosquito species. The classifier presents the top 5 predictions along with their probabilities. Transfer learning with a ResNet-50 architecture was employed to train, validate, and test the CNN model. Data augmentation, preprocessing techniques, and class imbalance mitigation strategies were applied. The best-performing model, identified through hyper parameter grid search, demonstrated high validation accuracy (91.78%), low validation loss (0.33), and strong average F1-score (90.37%). The efficiency of the mosquito surveillance framework will be determined in terms of mosquito classification accuracy and detection confidence level in offline and real-time field tests in a garden, drain perimeter area, and covered car parking area.
Existing SystemIn the existing system of AI-enabled mosquito surveillance using a dragonfly robot, conventional methods of mosquito surveillance often rely on stationary traps or manual monitoring, which can be inefficient and labor-intensive. These methods may also lack real-time data collection and analysis capabilities, leading to delays in identifying and addressing mosquito breeding grounds or disease outbreaks.
To overcome these limitations, the proposed system integrates artificial intelligence (AI) with a robotic platform inspired by the flight dynamics of dragonflies. The dragonfly robot is equipped with various sensors, cameras, and communication modules to autonomously navigate its environment and collect data on mosquito populations.
Proposed SystemThe proposed system consists of three main components They Are Dragonfly Robot, AI-based Mosquito Detection, and Data Analysis & Visualization.
- Dragonfly Robot: The Dragonfly Robot serves as the autonomous platform for mosquito surveillance. Inspired by the natural flight and behavior of dragonflies, this robotic device is designed for agile and versatile aerial maneuverability. Equipped with lightweight materials and precision-engineered mechanisms, the robot can mimic the flight patterns of real dragonflies, enabling it to navigate complex environments such as urban areas, forests, and water bodies with ease. The robot is equipped with a suite of sensors, including optical cameras, thermal imaging sensors, and microphones, to capture data from its surroundings.
- AI-based Mosquito Detection: The heart of the system lies in its AI-powered mosquito detection capabilities. Utilizing deep learning algorithms trained on extensive datasets of mosquito species and behaviors, the system can accurately identify and classify mosquitoes in real-time based on various characteristics such as size, wingbeat frequency, and thermal signature. The onboard processing unit of the Dragonfly Robot performs the analysis locally, allowing for rapid decision-making and autonomous operation without relying on external computational resources. Additionally, the system can adapt and learn from new data, continuously improving its detection accuracy over time.
- Data Analysis & Visualization: The data collected by the Dragonfly Robot is transmitted to a central server for further analysis and visualization. Advanced data analytics algorithms process the collected data to generate insights into mosquito population dynamics, species distribution, and potential disease transmission risks. These insights are presented through intuitive visualizations and interactive maps, enabling stakeholders such as public health authorities and urban planners to make informed decisions and implement targeted interventions for mosquito control and disease prevention.
This section evaluates the mosquito surveillance framework’s performance in the real-time field trial. As per the literature survey mosquitoes are more active at dusk, evening, nighttime, after rainfall, and in environments such as open perimeter drains, covered car parks, roadside drains, and garden landscapes. Hence, the experiments were carried out during the night (6 p.m. to 10 p.m.) and early morning (4 a.m. to 8 a.m.) in the potential breeding and cluttered environment such as gardens landscapes, open perimeter drains. In this experiment, the mosquito glue trap was fixed inside the trap unit and performed a mosquito trapping and surveillance function. Below figures shows the robot ’Dragonfly’ performing experiments in a different environment. The robot navigated to pre-defined waypoints with the help of Bluetooth control over multiple cycles.
Aedes Japonicus specimen
Psorophora Howardi specimen
The robot paused for 10 min at every waypoint, keeping its trap operational to gather more mosquitoes in the respective location. Once the robot completed its navigation to the final waypoint, it moved to the first waypoint and continued its inspection cycle. it shows a sample of real-time collected mosquito glue trap images from test environments. In this real-time analysis, the mosquito glue trap images were captured by a mobile phone, and images were transferred through IP address to PC for mosquito surveillance .
(a) (b)
Real-time mosquito glue trap (a) Glue Trap 1 (Morning) (b) Glue Trap 2 (Evening)
(a)
(b)
1.High Precision Surveillance
2.Versatile Deployment
3.Autonomous Operation
4.Global Impact
1.Public Health Monitoring
2.Environmental Monitoring
3.Disease Prevention and Control
4.Precision Agriculture
Frequently Asked Questions (FAQs)
Here are some frequently asked questions (FAQs) about the AI-enabled Mosquito Surveillance Using Dragonfly Robot:
- What is the AI-enabled Mosquito Surveillance Using Dragonfly Robot?
- This is a cutting-edge system that combines artificial intelligence (AI) and robotics technology to monitor mosquito populations and potential disease transmission risks. The system employs a dragonfly-inspired robot equipped with advanced sensors and AI algorithms for efficient and effective mosquito surveillance.
- How does the Dragonfly Robot work?
- The Dragonfly Robot is an autonomous aerial platform designed to mimic the flight patterns of real dragonflies. It navigates through various environments, capturing data using onboard sensors such as optical cameras, thermal imaging sensors, and microphones. AI algorithms analyze the collected data to identify and classify mosquitoes in real-time.
- What are the advantages of using the Dragonfly Robot for mosquito surveillance?
- The Dragonfly Robot offers several advantages, including its agility and maneuverability, which allow it to access hard-to-reach areas. It can autonomously collect data in diverse environments, providing real-time insights into mosquito populations and potential disease transmission risks. Additionally, the robot’s AI capabilities enable rapid and accurate mosquito detection without human intervention.
- How accurate is the mosquito detection performed by the AI algorithms?
- The AI algorithms used for mosquito detection are trained on extensive datasets of mosquito species and behaviors, resulting in high accuracy and reliability. These algorithms can identify and classify mosquitoes based on various characteristics such as size, wingbeat frequency, and thermal signature, with minimal false positives or negatives.
- What are some potential applications of the AI-enabled Mosquito Surveillance System?
- The system has numerous applications in public health, environmental monitoring, and disease prevention. It can be used to track mosquito populations, monitor disease transmission hotspots, and implement targeted interventions for mosquito control. Additionally, the data collected by the system can inform policy decisions and resource allocation for public health initiatives.