Study On Lane Departure Warning System Theory And Algorithms Based On Machine Vision

With the rapid increase of the number of cars, the vicious traffic accidents especially thelane departure accidents increased significantly. The main reason is that the driver fatigue ordistraction of their mind. Lane departure warning systems (LDWS) can estimates the momentthat the vehicle running out of the current lane in advance. It is also can apperceive whetherthe driver is intentional deviation. If not, the LDWS will alert to the driver. In turn, lanedeparture accidents can be reduced greatly. Therefore, the lane departure warning system hasbecome the focus of research and development all over the world.In the LDWS, accurate to obtain road information which is the lane marking is thepremise and foundation of accurate warning. Thus, the paper conducted a comprehensiveanalysis of the characteristics of structured road lane marking. Then choose the straight-linemodel to match straight lane and multi-segment linear model to match curved lane. At last,use improved Hough transform to detect lane marking. Considering that the lane markings arelocated in the lower part of the road image, the paper creatively proposed a method ofdynamic set regions of interest (ROI). The method can both reduce the amount ofcomputation to improve the real-time of the algorithm, and can ensure the accuracy of lanedetection.Analysis the advantages and disadvantages about the lane departure decision algorithmbetween the one based on camera calibration and the one based on image information onlywithout the need for camera calibration. This article chooses an algorithm without cameracalibration, only relies on the angle of the right lane marking and the left one to determinewhether the vehicle deviates. Dynamic track more than800frames of video image left andright lane marking angle offered by China FAW Technology Center, and statistic the changetrend. It shows that the algorithm without camera calibration can well describe the processoccurs when the vehicle deviates. At the same time, the algorithm has good robustness ofbumpy roads.The research and test shows that the proposed lane detection algorithm based onimproved Hough transform has good effect, and can be applied to the actual project ofintelligent vehicle safety assistance systems. It can provide the theoretical basis and designbasis for enterprise to develop LDWS and other active safety products.