Background modeling of street scenes for intelligent surveillance system /
An increasing number of CCTVs have been deployed in public and crime-prone areas as the demand for automatic monitoring system is increasing as a means to counterbalance the limitation of human or manual monitoring. To have a good monitoring system in such places, a good background model is needed i...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
Gombak, Selangor :
Kulliyyah of Engineering, International Islamic University Malaysia,
2016
|
| Subjects: | |
| Online Access: | http://studentrepo.iium.edu.my/handle/123456789/4365 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | An increasing number of CCTVs have been deployed in public and crime-prone areas as the demand for automatic monitoring system is increasing as a means to counterbalance the limitation of human or manual monitoring. To have a good monitoring system in such places, a good background model is needed in order to reduce the amount of the video processing needed for tracking, classification, counting and many other processes. This research proposes an adaptive background modeling that is able to model a scene under review in real-time. The proposed modeling system is expected to be able to handle dynamic backgrounds and common problems in detection methods. A novel patch-based background reconstruction, based on the highest frequency of occurrences assumption (MOP-BM) and principal component analysis (PCA), is proposed. The proposed algorithm was tested and analytically compared with the dynamic background in street scenes. The main challenges of background subtraction such as illumination changes, geometrical changes, stationary moving object problem and high speed object problem were extensively discussed and subsequently overcome in this research. The experimental results show that the proposed algorithm is able to reconstruct the background model and produce accurate and precise foreground with 99.01% accuracy on average and 89.38% average harmonics mean of precision and recall. From the results obtained, we could notice that the proposed algorithm gives significant contribution to the image processing field, especially in foreground-background segmentation. |
|---|---|
| Item Description: | Abstracts in English and Arabic. "A thesis submitted in fulfilment of the requirement for the degree of Master of Science in (Mechatronics Engineering)." --On t.p |
| Physical Description: | xviii, 145 leaves : ill. ; 30cm. |
| Bibliography: | Includes bibliographical references (leaves 130-135). |