Adaptive motion detection in video surveillance.
Giam, Darren Chia Kiong.
Date of Issue2009
School of Electrical and Electronic Engineering
Video surveillance is an active real-world application of computer vision as well as other discipline, such as communication and camera technology. This growing demand in surveillance is due to growing requirements from both military applications and the public for improved safety and security for people. A variety of these applications include the monitoring of everyday environments such as residential buildings, parking lots, banks, train stations and airports. Therefore, the main objective of this project is to be able to sense the background and to detect motion/static object. Various techniques include the initial pre-processing such as grayscale conversion, and noise filtering. Subsequently, background subtraction and inter-frame differencing are both methods to derive the analysis. Any illegal parking or intrusion will be sensed and send a trigger alarm to the operation management or the guard personnel. There will be a Graphical User Interface (GUI) for the operator to view and select various settings of the surveillance system. The language used to integrate this system is Borland C++ builder 6 with ImageEN which is a component suite for imaging processing, viewing and analysis. The surveillance system being proposed proves to be versatile that various spots in the background can be selected as the region of interest. For instance, in a single frame of the car park background, the user can select the stretch along the double yellow line to deter anyone from parking there. If it is a human standing there, it will not trigger the alarm as a threshold value will be set to a certain level. The evaluation of this project is able to perform to a certain level of accuracy by having a threshold function. However, due to weather and lighting conditions, it is not ideal. The limitations will be highlighted for future research.
Final Year Project (FYP)
Nanyang Technological University