Browsing by Author "De Silva, GC"

Now showing 1 - 3 of 3

item: Thesis-Abstract
Automation of traffic flow measurement using video images
De Silva, GC; Wickramarachchi, NK
This research focuses on the design and implementation of a system for automating the task of measuring vehicular traffic flow on a motorway, using video image sequences. Such a system is useful when designing traffic management systems such as signal lights at road junctions. It can also function as part of an intelligent transportation system, providing useful quantitative information, such as number and size of vehicles, speed and direction of movement etc, for managing vehicular traffic effectively. The system processes a sequence of images captured by a video camera mounted above a motorway. It performs motion segmentation on the images using background subtraction, and derives traffic parameters using a combination of image processing techniques such as statistical methods and solving systems of equations. While making use of existing algorithms for image processing and image sequence analysis, new algorithms have been developed and employed for background extraction and occlusion modeling. The system has been designed to handle inconsistencies caused by occlusion of vehicles, by predicting the motion of vehicles using the derived parameters of motion. Knowledge about the traffic scene has been used by the system to improve the accuracy of results. The system relies on a centralized 'knowledge base', a source of information related to the traffic scene. This knowledge base is loosely coupled to the system in such a way that the system can be customized to perform well in different traffic scenes by changing only the content of the knowledge base. The system has been implemented to work on an IBM PC based system with an Intel Celeron 450 MHz processor and 64MB Memory. The tools for implementation have been selected after a study of their performance. This automated traffic flow measurement system has been tested using three image sequences from different traffic scenes. A number of tests have been conducted by varying the frame rate, image resolution and the method used to create the background image. The results have been compared with each other, as well as the ground truth, while recording the processing time involved with each test case. The automated system measures traffic parameters with an average accuracy of 90%. The new technique that has been proposed in this work for background construction produces the most accurate result. Processing speed can be improved by reducing the frame rate and the resolution of images, with a slight reduction in accuracy of the results.
item: Conference-Extended-Abstract
Robust human body model acquisition from images in the presence or occlusion
(2003) De Silva, GC; De Silva, CR; De Silva, LC
Automated detection of postures and actions of humans from image/video data is \ aluablc for many applications. Some examples are automated surveillance. effective Human- Computer Interaction and Entertainment such as computer games. An essential step ill performing this task is to detect the presence of a human in a scene and acquire parameters of a predefined body model. Most of the existing research uses multiple cameras to obtain a full view of the human body for modelling. Where monocular images are used. using markers for identifying different parts or joints of the body is common. In most svsterns based on monocular images. it is assumed that the human body is occluded onlv b~ itself Marker-less automated human body model acquisition using monocular video in the presence of occlusion is still a challenging task.
item: Conference-Extended-Abstract
Virtual drum: using computer vision for musical expression
(2004) De Silva, GC
Musical instruments are transducers that convert forces applied by a human being into a c1~~s of .•.:l audio signals. With the availability of electronic devices that not only amplify the sounds generated, but also synthesize music when a small force sufficient to press a button is applied, the notion of 'force' is diminishing; instead, use of gestures to make music has gained more attention in recent research related to musical expression.