The purpose of this paper is to design Advanced Driver Assistance Systems(ADAD) and improving method using the smart vehicular camera. This system can provide safe and convenient drive and prevent inattentive driver from car accident.
In this paper we investigate related work likes previous published paper and commercial product to design and implement ADAS. There are some product which is widely used for state-of-the-art automobile manufacturer. But it is almost impossible for developer or researcher to create new function and revise poor factors in commercial ADAS platform. In other words it is based on closed software/hardware platform. Because the general-purpose mobile embedded development kit is not specialized in image processing function, there are some problems in processing image recognition. Straightforwardly, it is slow. In this paper. give the way to overcome closed platform and low performance of the embedded device.
first, we develop ADAS consist of pedestrian detection, car license plate detection, lane departure warning and circular sign detection in accordance with open source platform. That is open source based computer vision library CPU parallelization library and GPU parallelization library.
In point of program function we reduce the frame image area for recognition. Especially two functions, pedestrian detection and car license plate detection are slashed arithmetic operation by proposed method that is putting the image area in a dynamic. Also we could select frame whether to carry out recognition operation or not by means of GPU and acceleration sensor. Furthermore utilization of GPU multi-core and GPU many-core should maximize performance.
A key contribution of this paper is design the ADAS functions using the improved the recognition method based on open source platform and multi-core hardware environment. The system can be perform in real time and save the energy consumption by advanced processing speed.