Abstract:
This thesis describes a novel system architecture and implementation of a wireless
visual sensor node. The proposed design of the node can be used to extract
traffic information based on the visual description of road. In this research, the
real-time performances and the capability to perform at low power consumption
meanwhile obtaining accurate results were considered as the essential factors since
a large amount of data need to be processed in an embedded level vision system.
At first, a suitable vision algorithm is proposed to harvest the traffic condition
on road. The functionality for each section of the algorithm has been performed
by using carefully selected available vision techniques and image processing algorithms.
The vehicle extraction from the current frame of view and the tracking
process of the vehicle are identified as the most important functions in the algorithm.
The vehicle extraction from the current frame is carried out by the
ViBe algorithm with some modifications in order to acquire promising real time
performances and the tracking process is carried out by a light weight but an
accurate enough particle filtering technique.
Moreover, the complete system is implemented in the FPSoC hardware system
as a hardware and software co-design by considering advantages that can
be obtained from different aspects. The performances of the system have been
evaluated from many aspects for different standard data available from other research
works. The conclusions and suggestions for further development have been
presented at the end of this thesis.
