Distinguishing Abandoned and Removed Objects

i DECLARATION I hereby declare that this project is entirely my own work and that it has not been submitted as an exercise for a degree at this or any other university ___________________________________ ________________ Name Date ii Permission to Lend I agree that the Library and other agents of the College may lend or copy this thesis upon request. ___________________________________ ________________ Name Date iii Acknowledgements I would like to thank my family and friends for their encouragement throughout the course of this project; in particular I would like to recognise my parents for their continued support. I would also like to extent my thanks and gratitude to my supervisor Dr. Kenneth Dawson-Howe for his guidance and assistance throughout the project. iv Abstract With more cameras watching us now than ever, how this surveillance footage is analysed is very important. While human supervision is used in most circumstances, it is becoming more and more feasible to use computers for this application due to their increased processing power. This project presents a method for reliably detecting foreground events and defining the events as either an object abandonment or object removal using both edge based techniques and area reconstruction techniques. The objective of this project is to demonstrate how computer vision can be used to analyse and extract information from a scene. This is demonstrated by applying various methods within a computer vision system to accurately detect the abandonment and removal of objects from a scene. This project was written in C++ and the OpenCV 2.3.1 library was used for image processing. The project was tested on a laptop running Windows 8 with 6GB of memory and 2.7GHz dual core processor. The system was tested on three datasets, PETS 2006, Candela and Caviar which are the industry standard for examining and testing for abandoned and removal events. The following report details the development of the project and why the chosen approaches were implemented. Similar works in this area and other possible approaches for each stage of the project are also examined. The aim of this project was to detect static foreground events in a scene. These events would then be defined as either an abandonment or a removal and a numerical confidence of this definition would be produced. The foreground events were detected by constructing a background model of the scene and comparing the current scene to this model to discriminate