A Sequential Hypothesis Testing Approach to Detecting Small, Moving Objects in Image Sequences

A new algorithm is proposed for the detection of small, barely discernible moving objects of unknown position and velocity in a sequence of digital images. First, statistically robust prewhitening techniques are used to eliminate background structure and transform the image sequence into an innovations representation, modeled as Gaussian white noise. Then, a large number of candidate trajectories, organized into a tree structure, are hypothesized at each pixel in the sequence and tested sequentially for a shift in mean intensity. Underlying the algorithm are new general results in detection theory, including the use of multistage hypothesis testing (MHT) for simultaneous inference, and a new framework for quickest detection of time-varying signals in noise. In addition, exact, closed-form expressions for MHT test performance are derived; these predict the MHT Object Detection Algorithm's computation and memory requirements, where it is shown theoretically that several orders of magnitude of processing are saved over a brute-force approach. Feasibility of a parallel implementation on an MIMD, distributed memory, message-passing architecture is also shown. Results are verified experimentally on a variety of image sequences, including outdoor scenes digitized from videotape, digitized photographs, and digital data gathered by a CCD array at the output of a telescope.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2014

Author dc:creator

• Blostein, Steven David

Contributors dc:contributor

• Huang, Thomas S.

• Mathematics
• Engineering, Aerospace
• Engineering, Electronics and Electrical

Identifier

(UMI)AAI8908625

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/69407