The Basic Research Of Smart Car Vision System Based On ARM

Vision based navigation has become one of the most focused research issue all around world in recent years. General purpose computer used in the research of computer vision navigation is more and more popular. They are, however, not meet the need of low power, small volume and low cost and so on. In responded to this, we have tried to build an Intel XScale based ARM platform which was used as hardware run environment instead of common PC for our vehicle vision navigation system. It is an innovation idea since no ARM attempt was made so far.In this paper, the ARM technology was applied to smart vehicle automatic navigation for the first time. PXA255, an Intel XScale microchip based on ARM10, is used as the core processor. After introduce some relate concepts and background, the design of the hardware platform and interface is discussed in detail. And than analyzed the transplant process of embedded operation system WinCE. Subsequently, an image processing algorithm based on pixel and monocular vision is developed, which can be used to recognize the white strip of a highway and provide valid image information for vehicle automatic navigation. In addition, it can be used to the line traced fields with a very little modify or even though using directly. Experiments show that runtime of the algorithm is less than 70ms, which is perfect for time-critical applications. A transform mapping model is setup to limit the valid region of the algorithm and get the depth information of the road, which is also very useful in obstacle recognition. In the end of the paper, Experiments are conducted with the photos take from Hu-Ning highway Nanjing section. Both straight and curve roadway are demonstrated.This thesis has basically realized the vision platform based on ARM. A lane keeping algorithm was designed for this platform and run results show that the algorithm is fast and accurate enough to meet to the safety of automatic driving.