Just-enough-smart position sensing (JESSE) system
Abstract
Due to the proliferation of video cameras, the use of visual methods for surveillance, traffic monitoring, and optical navigation has dramatically increased. However, computer vision algorithms can produce a processing bottleneck. In hopes of addressing the latter, the search for optimal algorithms has explored many different strategies. Some have been inspired by the efficient natural machine called the fly. Of special importance in those cases is the fly's eye and its capability for motion detection and further its use in aerial navigation. Within this work, a novel method for tracking moving objects loosely based on the fly is presented. Moreover, taking in consideration some concepts observed in this insect's visual system, this dissertation describes a cheaper alternative (in a computational sense) for position detection. Of significant benefit, the model's optical system origin implies that the gross calculations contained in the approach could potentially be done electro-optically. The presented method detects and tags changes in a scene captured by a camera or other sensing device. The algorithm itself, effectively convolves input images contained in a video stream with a convex-shaped kernel, similar to those in the ommatidium (the basis structure of the fly's eye). Via the convolution, overlapped power responses at specific points are obtained. These responses can yield a set of intersecting circles that can possibly reveal target positions. By simple triangulation, a transformation to a set of power-crossing points may be done in order to find a final prediction of target position. An arrayed set of these ommatidial elements can form a complex data structure that may be applied to the problem of multi-target and multi-resolution tracking of moving objects with still and mobile camera in real scenes. Printed circuit board (PCB) inspection, surveillance/security, traffic monitoring, and optical navigation are some applications of these efforts and examples of these applications are provided.
Subject Area
Electrical engineering|Computer science
Recommended Citation
Saenz-Zamarron, David, "Just-enough-smart position sensing (JESSE) system" (2005). ETD Collection for University of Texas, El Paso. AAI3196417.
https://scholarworks.utep.edu/dissertations/AAI3196417