Development of vision-based lane and vehicle detecting systems via the implementation with a dual-core DSP

This paper presents an innovative vehicle intelligent warning device. It consists of a lane detection system (LDS) and a vehicle detection system (VDS) with a single CCD camera. For improving the system efficiency and reliability, the LDS uses an efficient median filter which takes 19 comparisons for two medians in the worst case and an edge-enhancement filter which makes road marks more obvious than background noise such as vehicles, buildings and so on. On the other hand, The VDS detects vehicles ahead by their symmetric characteristic with a modified symmetric area extraction method that can locate the symmetric area efficiently in an image. Besides, the motion vector (MV) estimation approach, which is widely adopted in MPEG image compression applications, is applied to predict the searching windows that lead to have better detecting rate. The LDS and the VDS are successfully implemented in an ADI-BF561 600MHz dual core DSP