A Highly Efficient And Automated Approach To Identifying And Classifying Urban Traffic Images Using Self Organizing Maps and Artificial Intelligence. (c2006)

Abstract

Identifying moving objects and isolating background noise to concentrate on a particular foreground object is a task that is heavily needed by a lot of applications in this day and time. Such a task is not always an easy one and can prove very tedious if done using the manual approach of assigning human beings to watch and enter data for long working hours. This is both a waste of human resources and valuable company time. Previous work on this subject included subtracting background information using conventional imaging techniques. This paper proposes a highly efficient approach using both optimized software and hardware techniques to achieve the best results in terms of correctness and speed. We do so by employing heavily optimized algorithms which divide the load over the CPU (central processing unit) and the GPU (graphical processing unit) and hence extract maximum performance from both processing units simultaneously. This sort of hybrid method has not been observed in proposed solutions to our problem.