Detecting unattended packages in video is a challenging surveillance problem. The goal is to detect packages and their owners, determine relationships between them and the environment, and recognize when an owner abandons a package. Concurrent events….
Contents
1 Introduction
1.1 Unattended Package Detection
1.1.1 Problem Formulation
1.1.2 Previous Work
1.2 Proposed System
1.3 Outline of Thesis
2 Motion-Based Unattended Package Detection
2.1 Overview
2.2 Motion-Based
2.3 Target Classification
2.4 Package Ownership Assignment
2.5 Alarm Conditions
3 Unattended Package Detection using Stochastic Attributes
3.1 Overview
3.2 Detection and Tracking
3.2.1 Video Pre-Processing
3.2.1.1 Measurement Smoothing
3.2.1.2 Shadow and Highlight Detection
3.3 Pattern Specification
3.3.1 Grammars and Languages
3.3.2 Properties of Grammars
3.3.3 Advanced Grammars
3.3.4 Attribute Grammars
3.3.4.1 Handling Concurrent Events
3.3.5 Stochastic Grammars
3.3.5.1 Generating Production Probabilities
3.4 Primitive Extraction
3.4.1 Error Correction
3.4.1.1 Gait Analysis
3.4.1.2 Tracking Errors
3.4.1.3 Shadows
3.4.1.4 Proposed System
3.5 Pattern Recognition
3.5.1 Previous Methods
3.5.1.1 Discrete Finite Accepters
3.5.1.2 Stochastic Automata
3.5.2 Parsing
3.5.2.1 Exhaustive-Search Parsing
3.5.2.2 Earley Parser
3.5.3 Parsing Attribute Grammars
3.5.4 Parsing Concurrent Events
3.5.5 Parsing Stochastic Grammars
3.5.6 Alarm Conditions
4 Experimental Results
4.1 Overview
4.2 System Implementation
4.3 Experiments
4.4 Data Set 1
4.5 Data Set 2
4.6 Data Set 3
4.7 Data Set 4
4.8 Data Set 5
4.9 Data Set 6
4.10 Conclusion
4.11 Future Work
Bibliography
Author: Rosenbush, Gavin E.
Source: University of Maryland
Download URL 2: Visit Now