Jump to navigation
Moving object detection in sequences distorted by atmospheric turbulence
Blomgren, PeterLiu, Xiaobai
In this thesis we review methods of object detection and tracking in video sequences distorted by atmospheric turbulence. Modern Horn-Schunck optical flow methods are used to estimate the displacement field between two consecutive video frames. It is natural to assume that movement from turbulence will be present in addition to movement from the foreground objects that we want to detect. We investigate using complex curvelet decomposition to separate the oscillating and geometric components of the flow field. We find that this method does greatly improve the detection accuracy of optical flow methods. Additionally, we investigate the following: methods of pre-processing images, using vector magnitude to segment pixels into foreground and background, and a comparison of background subtraction methods to optical flow.
Applied MathematicsDynamical Systems
Mathematics and Statistics
San Diego State University
Master of Science (M.S.) San Diego State University, 2020
© 2015 SDSU Library & Information Access. All Rights Reserved.