TRAP@NCI

Evaluation of Background Segmentation Algorithms on Embedded Devices

Stepanenko, Denis (2015) Evaluation of Background Segmentation Algorithms on Embedded Devices. Masters thesis, Dublin, National College of Ireland.

[img]
Preview
PDF (Master of Science)
Download (1MB) | Preview

Abstract

The CCTV monitoring operators miss up to 95% of intrusions when monitor-ing multiple cameras simultaneously (Hyenkyun, et al., 2010). In order to tackle this problem, many monitoring centres utilize various motion detection approaches, as opposed to constantly watching the screens. These ap-proaches work well in illuminated environments or indoors, but outdoors these tend to generate multiple false alarms. The problem is even worse when the camera utilizes Infra-Red lighting, as noise motion such as drops of rain, cobwebs and flies are much more visible. Since every detection needs to be investigated, operators are able to monitor less cameras than they would otherwise, which increases the monitoring service price.

One potential solution to false detections is the installation of motion sen-sors, such as PIRs. However, these tend to be expensive, require additional wiring and hardware, and can be extremely unreliable.

This research investigates if the latest background segmentation algo-rithms can be utilized to suppress noise objects efficiently enough, to be utilized on embedded devices such as those used by IP cameras. To answer this question, a number of utilities were developed to allow the testing of problematic feeds and analysing results. In addition, a number of post-detection filters were built to suppress false detections even further.

The tests were carried out on a variety of video feeds containing intruders and noise objects. These tests were carried out on a laptop and various em-bedded devices such as Raspberry PI. The research methodology used was quantitative. The analysis of the data shows that a significant amount of noise objects can indeed be suppressed, with acceptable decrease in FPS rate.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Divisions: School of Computing > Msc.: Master of Science in Web Technologies
Depositing User: CAOIMHE NI MHAICIN
Date Deposited: 20 Oct 2015 09:14
Last Modified: 20 Oct 2015 09:14
URI: http://trap.ncirl.ie/id/eprint/2101

Actions (login required)

View Item View Item