Unattended baggage detection using deep neural networks
As the world becomes ever more attuned to potential security threats, the need for sophisticated surveillance system is increasing to monitor and detect any potential threats. Sophisticated surveillance system should functions as an intuitive “robotic eye” for accurate and real-time detection of thr...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/79477/1/OngYiWeiMFKE2018.pdf |
| _version_ | 1796863052753141760 |
|---|---|
| author | Ong, Yi Wei |
| author_facet | Ong, Yi Wei |
| author_sort | Ong, Yi Wei |
| collection | ePrints |
| description | As the world becomes ever more attuned to potential security threats, the need for sophisticated surveillance system is increasing to monitor and detect any potential threats. Sophisticated surveillance system should functions as an intuitive “robotic eye” for accurate and real-time detection of threats. Unattended baggage has become a critical need for security personnel at airports, stations, malls, and in other public or crowded areas. However, an effective system for detection of objects like baggage and people with a real-time video input requires high processing power and storage to just process the video frames using the typical digital image processing technique. This will require a very high development cost and time in order to make the system work which is impractical for commercial use. Moreover, manual configuration is needed which is not flexible to be for multiple application. Therefore, the objective of this thesis is to improve the object detection accuracy and flexibility compared to existing digital image processing techniques. This proposed system uses deep neural networks approach through collection of datasets thus providing a more accurate detection and flexible application. Tensorflow framework is used as the deep neural network framework for the development of this system. This system utilizes the Single Shot multibox Detection detection algorithm to the ’MobileNet’ neural network architecture which is optimized to provide a promising performance even in embedded system. This project is developed by implementing the Tensorflow Object Detection Application Programming Interface (API). This method enables 4 main classes of detection which are suitcase, backpack, handbag and person. The datasets used for benchmarking are surveillance video sample that contain unattended baggage scenario used by most existing works like AVSS2007, PETS2006 and ABODA. The overall accuracy and flexibility of the proposed system improved up to 43% thus unattended baggage is able to be detected. The system is able to be applied in various environment due to the excellent flexibility of the system. |
| first_indexed | 2024-03-05T20:20:54Z |
| format | Thesis |
| id | utm.eprints-79477 |
| institution | Universiti Teknologi Malaysia - ePrints |
| language | English |
| last_indexed | 2024-03-05T20:20:54Z |
| publishDate | 2018 |
| record_format | dspace |
| spelling | utm.eprints-794772018-10-31T12:41:35Z http://eprints.utm.my/79477/ Unattended baggage detection using deep neural networks Ong, Yi Wei TK Electrical engineering. Electronics Nuclear engineering As the world becomes ever more attuned to potential security threats, the need for sophisticated surveillance system is increasing to monitor and detect any potential threats. Sophisticated surveillance system should functions as an intuitive “robotic eye” for accurate and real-time detection of threats. Unattended baggage has become a critical need for security personnel at airports, stations, malls, and in other public or crowded areas. However, an effective system for detection of objects like baggage and people with a real-time video input requires high processing power and storage to just process the video frames using the typical digital image processing technique. This will require a very high development cost and time in order to make the system work which is impractical for commercial use. Moreover, manual configuration is needed which is not flexible to be for multiple application. Therefore, the objective of this thesis is to improve the object detection accuracy and flexibility compared to existing digital image processing techniques. This proposed system uses deep neural networks approach through collection of datasets thus providing a more accurate detection and flexible application. Tensorflow framework is used as the deep neural network framework for the development of this system. This system utilizes the Single Shot multibox Detection detection algorithm to the ’MobileNet’ neural network architecture which is optimized to provide a promising performance even in embedded system. This project is developed by implementing the Tensorflow Object Detection Application Programming Interface (API). This method enables 4 main classes of detection which are suitcase, backpack, handbag and person. The datasets used for benchmarking are surveillance video sample that contain unattended baggage scenario used by most existing works like AVSS2007, PETS2006 and ABODA. The overall accuracy and flexibility of the proposed system improved up to 43% thus unattended baggage is able to be detected. The system is able to be applied in various environment due to the excellent flexibility of the system. 2018 Thesis NonPeerReviewed application/pdf en http://eprints.utm.my/79477/1/OngYiWeiMFKE2018.pdf Ong, Yi Wei (2018) Unattended baggage detection using deep neural networks. Masters thesis, Universiti Teknologi Malaysia, Faculty of Electrical Engineering. |
| spellingShingle | TK Electrical engineering. Electronics Nuclear engineering Ong, Yi Wei Unattended baggage detection using deep neural networks |
| title | Unattended baggage detection using deep neural networks |
| title_full | Unattended baggage detection using deep neural networks |
| title_fullStr | Unattended baggage detection using deep neural networks |
| title_full_unstemmed | Unattended baggage detection using deep neural networks |
| title_short | Unattended baggage detection using deep neural networks |
| title_sort | unattended baggage detection using deep neural networks |
| topic | TK Electrical engineering. Electronics Nuclear engineering |
| url | http://eprints.utm.my/79477/1/OngYiWeiMFKE2018.pdf |
| work_keys_str_mv | AT ongyiwei unattendedbaggagedetectionusingdeepneuralnetworks |