Automated tracking system for video surveillance system /
In recent years, video surveillance system has emerged as one of the active research area in machine vision community. This thesis intends to integrate machine vision into video surveillance system in order to enhance the accurateness and robustness of video surveillance system. To realize more robu...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
Kuala Lumpur:
Kulliyyah of Engineering, International Islamic University Malaysia,
2013
|
| Subjects: | |
| Online Access: | http://studentrepo.iium.edu.my/handle/123456789/4357 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In recent years, video surveillance system has emerged as one of the active research area in machine vision community. This thesis intends to integrate machine vision into video surveillance system in order to enhance the accurateness and robustness of video surveillance system. To realize more robust and secure video surveillance system, an automated human and objects' tracking system is needed which can detect, classify and track human and objects even when the occlusion occurs. Hence, we proposed an automated human tracking system which includes detection, classification and tracking of human and objects (especially vehicles) in real-time surveillance system and also in solving the problem of partially occluded human by utilizing fast-computation techniques without compromising the accuracy and performance of that particular surveillance system. In this thesis, we presented the use of foreground segmentation based on adaptive background subtraction to extract foreground objects from the image. Then, to obtain correct detection, we apply shadow removal based on global contrast adjustments in RGB colour space. Next, the extracted foreground objects will be morphologically reconstructed before process of classification. In the process of classification, we utilized new set of affine moment invariants based on statistics method together with aspect ratio in order to classify the extracted foreground objects. Then, finally we track the classified human and objects using feature based tracking for five states, which are: entering, leaving, normal, merging, and splitting. The developed video surveillance system can classify human by extracting the human head-shoulder up to 60 – 70 % occlusion with background objects. Besides that, the developed system also can track the human even if occlusion occurs since we used merging and splitting cases in our tracking algorithm. The overall accuracy for our proposed surveillance system in classifying human and objects (especially car) is excellent which is at 97.51%. The overall accuracy for our surveillance system in tracking human and objects (especially car) is fine also which is at 94.74%. |
|---|---|
| Item Description: | "A thesis submitted in fulfilment of the requirement for the degree of Master of Science (Mechatronics Engineering)."--On t.p. |
| Physical Description: | xx, 177 leaves : ill. ; 30cm. |
| Bibliography: | Includes bibliographical references (leaves 134-143). |