Yakov Diskin



Download Project (3.7 MB)


Cost effective persistent wide area surveillance is a challenging real-world problem that research has not sufficiently tackled yet. At present, surveillance corporations spend millions on human analysts to monitor live or recorded video feeds. Depending on the application, the analysts may be looking for unauthorized activities, suspicious behavior, or a more specific sequence of events. Human performance is costly and is often affected by ambiguous definitions of anomalies as well as natural factors such as fatigue. We present a fully automatic 3D change detection technique designed to support persistent overhead surveillance in changing environmental conditions. The novelty of the work lies in our approach of creating an intensity invariant system tasked with detecting changes in a changing environment. Although previous techniques have proven to work in some cases, these techniques fail when the intensity of the scene significantly changes between the capture of the datasets. Our techniques leverages our 3D reconstruction capabilities to overcome the intensity variation challenges. We present several proof of concept experiments conducted in a laboratory setting, in which we study the effects of model noise and scene illumination on the proposed volumetric changed detection algorithm.

Publication Date


Project Designation

Graduate Research

Primary Advisor

Vijayan K. Asari

Primary Advisor's Department

Electrical and Computer Engineering


Stander Symposium project


Arts and Humanities | Business | Education | Engineering | Life Sciences | Medicine and Health Sciences | Physical Sciences and Mathematics | Social and Behavioral Sciences

Determining Volume Changes from Overhead Video Surveillance