3D Reconstruction via Stereoscopy for the Study of Mosquito Swarms

Introduction

Male mosquitoes swarm to find mates, but the characteristics of these swarms have not often been quantitatively analyzed in the field. This site includes information on a simple method used to localize individual Anopheles gambiae in a swarm via stereoscopic image analysis of video footage swarms recorded in Doneguebougou, Mali. The methods described represent a streamlining of possibly the most basic method for localizing objects in space in the hope that other researchers will be able to apply the method for studying fundamental questions about flying insects or other taxa. To understand the method and its limitations you should read the "Description of a Method for Localizing Swarming Mosquitoes in 3D Space" and the original Manoukis et al paper ("Structure and Dynamics of Male Swarms of Anopheles gambiae") linked in the References section below.

Software

Manoukis et al used various third-party programs to analyze our data. These are cited in the reference "Description of a Method for Localizing Swarming Mosquitoes in 3D Space" including URLs for download. The researchers also developed two programs to facilitate insect localization, which they make available here: 1) An R function to calculate positions from stereo measurements, called calcxyz.R and 2) a convenience plugin for ImageJ which may be used to easily gather these data (called StereoPointPicker, a small modification of PointPicker by Philippe Thevenaz). The code.zip file contains both programs and a short explanation of how to use them.

Visualization of mosquitoes within swarms

This entry contains visualizations of the estimated locations of Anopheles gambiae mosquitoes filmed in Doneguebougou, Mali, between August and October 2007. You will find two types of visualizations here, described below. The objective is to give the viewer an intuitive understanding of the structure and dynamic of these swarms.

Each swarm filmed and analyzed is shown separately. All the positions shown are relative to the swarm centroid. More details are given in the paper. 

Time lapse movies (FLV)

Two types of flash (FLV) movies are presented here. On the left are the estimated positions of mosquitoes in each stereo image analyzed (each image is separated by an interval of about 15 seconds in real-time). On the right the same data are presented, but the positions of mosquitoes seen are not removed as time progresses. This allows easier comparison of positions between images and over the entire swarming period.

Cumulative object visualizations (VRML)

VRML (Virtual Reality Modeling Language) plots of cumulative mosquito positions for each swarm are included. These models provide complete freedom for the viewer to explore where mosquitoes were seen throughout the swarm. These models are manipulable by the viewer, but you will most likely have to install specialized software for your browser to be able to view them. The NIST VRML Plugin and Browser Detector page (HERE) will detect your browser and platform and list visualization options with download links. Two options that we have tested and can recommend are FreeWRL (for Mac OS X and Linux) and Flux Player (for MS Windows).

NIAID Laboratory of Malaria and Vector Research
May 2008