Recipient Organization
VECTECH, LLC
1812 ASHLAND AVE STE 100
BALTIMORE,MD 212051546
Performing Department
(N/A)
Non Technical Summary
Mosquitoes are the deadliest animal in the world, infecting over 350 million people each year with a range of diseases. In farming communities, mosquitoes can further spread animal-borne pathogens and threaten livestock. Due to a lack of effective vaccines for mosquito-borne viruses, integrated control of mosquito populations remains the primary strategy for disease mitigation. Effective mosquito control is informed by vector surveillance - monitoring an area to understand mosquito species composition, abundance, and spatial distribution. Unfortunately, conventional practices rely on the manual distribution of mosquito traps and routine visits to collect and identify the specimens, a resource-intensive and expensive process. As a result, many county and municipal departments of health, particularly those in rural communities, do not have the capacity and capability to conduct routine vector surveillance.Here, we propose to develop the first automated counting trap for mosquito surveillance with image-recognition. Our approach will include the selection of camera sensors for integration with a mosquito trap, creation of an image-dataset of medically-relevant mosquito genera imaged within the trap, development of software for detecting and distinguishing between mosquitos and non-mosquitoes, and identifying the electronics requirements for deployment of the camera and software as a remote sensing system. Ultimately, successful development of these approaches will establish image-recognition as a reliable remote sensing modality for detection and monitoring of mosquitoes. Eventually translated into a low-cost system to remotely transmit mosquito abundance information to public health decision makers, mosquito control organizations will benefit from reduced operational costs, and a standardized system to mitigate mosquito disease risk.
Animal Health Component
45%
Research Effort Categories
Basic
10%
Applied
45%
Developmental
45%
Goals / Objectives
Mosquitoes are the deadliest animal in the world, infecting over 350 million people each year with a range of diseases. Driven by climate change and insecticide resistance, this burden is expected to grow. Due to a lack of effective vaccines for mosquito-borne viruses, integrated control of mosquito populations remains the primary strategy for disease mitigation. Mosquito surveillance - monitoring an area to understand mosquito species composition, abundance, and spatial distribution - is critical to informing decisions about what control strategies will be most effective in specific locations and is necessary to determine if interventions are effectively decreasing mosquito populations. Conventional surveillance practice relies on manual distribution of mosquito traps and routine visits to collect the specimens. Due to the resource-intensive nature of vector surveillance, many county and municipal departments of health, particularly those in rural communities, do not have the capacity and capability to conduct routine surveillance. This project aims to develop the first computer vision driven automated counting trap for mosquito surveillance. This investigation will focus on the following technical objectives:Specify, design, and validate a low-cost optical configuration for object-detection and genus classification compatible with a mosquito trap.Acquire a relevant image database, develop object detection capability, and implement computer vision domain adaptation for genus classificationDefine hardware specifications for power, communications, and ruggedization of a mosquito trap with integrated optical systemImage recognition techniques will focus on object detection in the context of the new optical design, and hardware specifications will be considered with the aim of transitioning development into a low-cost system to remotely transmit abundance information to community public health agencies and support actionable biosurveillance. These technical objectives will build on Vectech's core competencies in image-recognition and optics, and establish feasibility for phase II research and development.
Project Methods
Conventional surveillance practice relies on manual distribution of mosquito traps and routine visits to collect the specimens. This practice is slow, expensive, and highly resource intensive. The proposed evaluation here will assess the feasibility of automating this practice with computer vision driven automated counting traps for remote assessment of mosquito abundance. While significant efforts have been made to develop remote sensing technologies for mosquito counting, only a computer vision approach has evidence of reliably counting and distinguishing mosquitoes from bycatch. The methods required for this evaluation include: creation of an image dataset of the three most medically-relevant mosquito genera, aedes, anopheles, and culex; optics design and validation to resolve distinctive mosquito morphology from images; development of object-detection and genus classification models for the imaging system design; and electromechanical design for integration with a passive mosquito trap.Key milestones of this proposal include:Report on proposed optical configuration and design justification based on imaging requirements for mosquito counting and genus classificationImage database includes at least two-hundred specimens of each of the three most medically-relevant mosquito genera: aedes, anopheles, and culex.Object detection achieves 90% accuracyGenus classification achieves 90% accuracyReport on hardware specifications and modifications necessary for integration of optical configuration with a mosquito trap