Research And Implementation Of Detection And Recognition Technology Of Lane And Zebra Crossing And Road Mark Based On Machine Vision

With the progress of society, cars are becoming an indispensable transportation in our lives. While at the same time, the problem of traffic safety is more and more serious, and it causes widespread attention. It is facing a serious problem that improves safe performances and reduces traffic accidents. Therefore, the research of intelligent transportation assistant driving system is becoming the hotspot. It plays a very important role in the safety of vehicle.In this paper, we study the detection and recognition technology of lane and zebra crossings and road mark. Lane is an important logo on road driving, and zebra crossings and road mark can provide the important information about crossroads, can achieve the purpose of auxiliary driving. In this paper, we study the video collected machine vision. Then lane information is extracted from each frame, the main process of this system is:firstly, image preprocessing operations filter out a lot of interferential information. Preprocessing operations include graying image, smoothing, image enhancement, binarization and edge repair. Secondly, after preprocessing, feature regions of image are extracted. Then, according to the characteristics of lane, feature regions are filtered. The least-square method is adopted to fitting after filtering feature regions. Then lane is identified in the image. After the lane recognition, determine the region of interest in the current lane, scanning the region and according to the characteristics of zebra crossings and road arrow to filtering and detection them. Sequence images have continuity between them. Therefore identified lane in the previous frame image can predict lane in the next frame image. Scanning range is reduced. Real-time performance of the system is improved. It works out tracking and identifying of images. However, due to light or others, extracted the lane is incomplete. In this paper, some work is done to ensure this system robustness.The experimental results show that, the algorithm in this paper can more accurately identify the lane, zebra crossings and road arrow, and realizes the processing to the video. It has good accuracy and real-time performance.