Design And Implementation Of A Lane Departure Warning System Based On Vision

Traffic safety is the eternal theme in the development of transport, with the continuous increase in the number of vehicles. The number of traffic accidents, particularly severe traffic accidents increases rapidly, and caused tremendous economic losses to all the countries in the world. Severe traffic accidents caused by lane departure have a large proportion in all traffic accidents. In order to reduce the traffic accidents caused by lane departure, the concept of Lane Departure Warning system was put forward.In this thesis, the current research situation of lane departure system is reviewed. A new method using contrast feature between road and lanes for the segmentation of whole candidate region of lane has been put forward to make up for the deficiency of the current existing algorithms in Lane Departure Warning systems. A method using the idea of differential approximation to represent the lane is studied, which has improved the precision of the Lane Departure Warning system, and need not to assign the road model in advance. In addition, by comparison to other models of Lane Departure Warning systems, our model has been decided.In addition, a system design of lane departure warning system has been carried out. The system design mainly depicts the system in two aspects, hardware and software. In hardware design, the ARM series CPU, which has advantages such as low power consumption and high stability, has been chosen. The strategy of data storage and the structure of algorithm have been set up according to the features of the CPU and the data requirements of the system. In software design, A layered architecture is used in order to ensure the extensibility and the stability of the system. Then an analysis of the data flow and the control flow was made, the calling sequence among all the modules was established. Also the core algorithms were optimized in order to meet the performance requirements on the target platform.