An Augmented Navigation System Based On Road Markings Recognition

With the popularity of GPS navigation technology and applications, car travel has become more and more convenient. However, the issues and deficiencies of the existing GPS navigation are becoming increasingly prominent. The new developed vehicle navigation uses sensor, such as camera, and was integrated with computer vision, image processing, graphics and other technical. This technology to enhance the adaptability and practicability of the navigation equipment through being aware of surroundings is called augmented reality navigation. This technology is the future direction of the development of vehicle navigation. The existing augmented reality navigation products can recognize some kind of object, for example vehicles in front, traffic lights and signs of some chain. Nevertheless these products are not capable of lane navigation. Therefore they can not effectively solve the problem that GPS gives the wrong instruction. Augmented reality navigation equipment is not capable of telling the lane direction, so sometimes it may suggest driver the wrong direction into a one-way street or miss the opportunity of telling the driver that he is in the wrong lane. The purpose of this study is to solve the problems above.To solve the issues above, this paper presents a complete set of road markings recognition for iOS devices. First improved Otsu method with the dichotomy was used in image segmentation. Steerable filter was used to detect edge; then Hough transform and linear fitting were used to extract the lane, and histogram was used to determine the region of interest. After correction on the final segmentation of the region, its Haar-like characteristics were extracted. Adaboost classifier was used for classification, and the tracking module was used to reduce the false recognition in the specific environment and to improve the recognition rate finally. Experiments show that this solution not only has good recognition, high robustness but low complexity and real-time preformance.This paper developed a version which run on the PC platform for real driving and complete an iOS application which runs in the iphone. The user interface of this application was also well designed. No matter running in the simulator or in the real machine the application has good recognition effect and fluency. This solution is not only suitable for use in iOS applications, but also very applicable to other mobile embedded devices.