Real -Time Tracking for Intelligent Surveillance Systems
Maryam Nabil Zakaria Al-Berry;
Abstract
Intelligent surveillance is very important for security-sensitive fields. Generally, surveillance can be defined as the observation of changing information, activities or behaviors for some purpose. The framework of visual surveillance systems includes environment modeling, motion detection, object classification, tracking as well as behavior understanding and description. Environment modeling is the module responsible for creating and updating dynamic models for the environment. Motion detection is the module responsible for segmenting moving objects from static or irrelevant background. This module is the base for any subsequent processing; thus it must be accurate, robust and fast. A surveillance scenario may contain different types of moving objects, therefore the system must classify the detected objects. The tracking module tracks the classified moving objects from one frame to another and then the behavior of the tracked objects is analyzed and a description of actions/activities is provided.
The first contribution of this thesis is mainly concerned with the problem of motion detection. Two innovative spatio-temporal wavelet-based motion detection techniques are proposed, combining the advantages of wavelets, multi-resolution analysis and data fusion to enhance the performance without raising the complexity. The first proposed technique is based on 3D Stationary Wavelet Transform (SWT), which combines spatial and temporal analysis into a single 3D transform by applying 1D analysis in the x-, y- and t- domains. The second proposed technique is implementing the 3D transform as two separate spatio-temporal analyses. Both of the proposed techniques are compared to the recent techniques using a benchmark dataset. In addition, the proposed 3D technique is compared to another 3D wavelet-based technique using a traffic monitoring dataset. Both of the proposed techniques outperform traditional techniques, especially in the cases of low contrast scenes and those having non-uniform illumination and they succeeded to detect moving objects in bad and time-varying illumination conditions.
The first contribution of this thesis is mainly concerned with the problem of motion detection. Two innovative spatio-temporal wavelet-based motion detection techniques are proposed, combining the advantages of wavelets, multi-resolution analysis and data fusion to enhance the performance without raising the complexity. The first proposed technique is based on 3D Stationary Wavelet Transform (SWT), which combines spatial and temporal analysis into a single 3D transform by applying 1D analysis in the x-, y- and t- domains. The second proposed technique is implementing the 3D transform as two separate spatio-temporal analyses. Both of the proposed techniques are compared to the recent techniques using a benchmark dataset. In addition, the proposed 3D technique is compared to another 3D wavelet-based technique using a traffic monitoring dataset. Both of the proposed techniques outperform traditional techniques, especially in the cases of low contrast scenes and those having non-uniform illumination and they succeeded to detect moving objects in bad and time-varying illumination conditions.
Other data
| Title | Real -Time Tracking for Intelligent Surveillance Systems | Other Titles | التتبع في الوقت الحقيقي في نظم المراقبة الذكية | Authors | Maryam Nabil Zakaria Al-Berry | Issue Date | 2015 |
Recommend this item
Similar Items from Core Recommender Database
Items in Ain Shams Scholar are protected by copyright, with all rights reserved, unless otherwise indicated.