Automated Kinematic Extraction of Wing and Body Motions of Free Flying Diptera

dc.contributor.advisorHumbert, James Sen_US
dc.contributor.authorKostreski, Nicholas Ivanen_US
dc.contributor.departmentAerospace Engineeringen_US
dc.contributor.publisherDigital Repository at the University of Marylanden_US
dc.contributor.publisherUniversity of Maryland (College Park, Md.)en_US
dc.description.abstractIn the quest to understand the forces generated by micro aerial systems powered by oscillating appendages, it is necessary to study the kinematics that generate those forces. Automated and manual tracking techniques were developed to extract the complex wing and body motions of dipteran insects, ideal micro aerial systems, in free flight. Video sequences were captured by three high speed cameras (7500 fps) oriented orthogonally around a clear flight test chamber. Synchronization and image-based triggering were made possible by an automated triggering circuit. A multi-camera calibration was implemented using image-based tracking techniques. Three-dimensional reconstructions of the insect were generated from the 2-D images by shape from silhouette (SFS) methods. An intensity based segmentation of the wings and body was performed using a mixture of Gaussians. In addition to geometric and cost based filtering, spectral clustering was also used to refine the reconstruction and Principal Component Analysis (PCA) was performed to find the body roll axis and wing-span axes. The unobservable roll state of the cylindrically shaped body was successfully estimated by combining observations of the wing kinematics with a wing symmetry assumption. Wing pitch was determined by a ray tracing technique to compute and minimize a point-to-line cost function. Linear estimation with assumed motion models was accomplished by discrete Kalman filtering the measured body states. Generative models were developed for different species of diptera for model based tracking, simulation, and extraction of inertial properties. Manual and automated tracking results were analyzed and insect flight simulation videos were developed to quantify ground truth errors for an assumed model. The results demonstrated the automated tracker to have comparable performance to a human digitizer, though manual techniques displayed superiority during aggressive maneuvers and image blur. Both techniques demonstrated non-intrusive methods for establishing reference flight kinematics, which are being used to develop flight dynamics models in future work.en_US
dc.subject.pqcontrolledAerospace engineeringen_US
dc.subject.pqcontrolledComputer scienceen_US
dc.subject.pquncontrolledcomputer visionen_US
dc.subject.pquncontrolledflight dynamicsen_US
dc.subject.pquncontrolledkalman filteren_US
dc.subject.pquncontrolledmotion reconstructionen_US
dc.titleAutomated Kinematic Extraction of Wing and Body Motions of Free Flying Dipteraen_US


Original bundle
Now showing 1 - 3 of 3
Thumbnail Image
5.44 MB
Adobe Portable Document Format
No Thumbnail Available
84.21 MB
Unknown data format
No Thumbnail Available
10.69 MB
Unknown data format