الفهرس 
Dissertation Organization
An Enhanced SOC-Based Sobel Edge DetectorOnly 14 pages are availabe for public view
 |  | from | 141 |  |  |
| | | from | 141 | | |
Abstract
Nowadays, there is tremendous and continuous progress in using digital images in many fields. With the great development in the camera sensors which produces a high-definition video streaming with high resolution and frame rate, the processing of the sequence of digital images in real-time became a great challenge for researchers. Researchers have begun to solve this problem by using different algorithms and techniques. Real-time moving target detection (MTD) using digital images processing is one of its most important areas in research as it has many applications. For example, it helps reduce the phenomenon of terrorism by the early detection of moving objects in the working area scope, and determining if there is a threat or not. Therefore, real-time motion detection (MD) systems must take fast decisions, and hence it must process the large amount of data acquired by camera sensors quickly. However, processing this data sequentially on the serial processor, leads to a slow decision and loss of targets. Here, we treat this problem by applying two methods. First, we process the images in parallel. Second, we use optimized image processing techniques.
Processing the images in parallel reduces the calculations time. To achieve the maximum gain of parallel processing technology, we have used the system on a chip which is divided into two main parts linked to each other by certain protocols that allow for fast and efficacious data transfer. The first part of this system is the serial processor, which is used in tasks management only, the second part is programmable logic circuits, which have the power to process images in parallel using some techniques like pipeline and unrolling techniques.
The second method adds some optimization techniques to the system to make the system capable of implementing all the image processing functions using line buffer and memory window techniques, and manage the design with a new technique which makes the system capable of processing HD video using less amount of on-chip memory.
We also propose an approach to track specific moving targets and speedup their detection and tracking rate by training the system on many images of this target. The output features are stored in a database and they characterize the moving target in all situations. The proposed approach uses some filters to improve the contrast of the image, which improves the quality of detecting targets. On the other hand, we divide the processing frame into sub-windows and process each sub-window individually that supports the system to reduce the amount of memory used. Also, we use some algorithms such as integral images and cascade classifier to lessen the computation time of the processing sub-window features, and early decide the irrelevant features of this sub-window.
We implement both MD system using C++ and System on chip (SOC), and tracking system using C++ and OpenCV tool based on Ubuntu Operating System(OS).