单目视觉的智能电动小车车道线识别与跟踪

发布时间：2018-12-16 11:22

【摘要】：无人驾驶电动汽车集机械技术、电子技术、自动控制技术、人工智能技术、机器视觉技术、新能源技术、新驱动技术等众多高新技术于一体,是计算机科学、模式识别和智能控制技术高度发展的产物,也是衡量一个国家科研实力和工业水平的一个重要标志。自主定位导航和智能行驶控制技术是无人驾驶技术的核心技术,复杂的多变的外界环境使得其成为了无人驾驶技术的重点和难点。本文以园区行驶环境为研究背景,面向低速、低成本、实时可靠的无人驾驶技术,对智能电动小车的运动控制系统软硬件,多传感器数据融合的避障技术以及小车的单目视觉的车道线检测等相关技术进行了研究。文中以车道线图像检测和识别算法为主要内容,其中包括图像的预处理、IPM视图的转换、滤波降噪、霍夫直线的检测、RANSAC和贝塞尔算法的样条曲线拟合、拟合曲线后的扩展等内容。通过提取的车道线确定智能电动小车与车道线的相对位置,对车道发生偏离时发生预警,从而实现对车道线的跟踪。将这些提取的信息与环境感知模块的其它传感器进行数据的融合旨在完成智能电动小车的自主导航。最后对车道线检测算法进行了验证,实测结果表明本文所改进的车道线检测算法比较准确稳定,对于车辆的遮挡、树木的阴影、不清晰的车道线、明显弯曲的车道线等典型不同道路环境情况能够准确检测出车道线,为智能电动小车在园区环境下自主行驶提供了可行性。
[Abstract]:The driverless electric vehicle is a computer science, which integrates mechanical technology, electronic technology, automatic control technology, artificial intelligence technology, machine vision technology, new energy technology, new drive technology and so on. The high development of pattern recognition and intelligent control technology is also an important symbol of a country's scientific research strength and industrial level. Autonomous positioning navigation and intelligent driving control technology are the core technologies of unmanned driving technology. The complex and changeable external environment makes them become the focus and difficulty of unmanned driving technology. In this paper, based on the driving environment of the park, the software and hardware of the motion control system of the intelligent electric car are introduced, which is based on the driverless technology of low speed, low cost and real time reliability. The obstacle avoidance technology of multi-sensor data fusion and the lane line detection of single-eye vision of small car are studied in this paper. The main contents of this paper include image preprocessing, IPM view conversion, filtering and noise reduction, Hough line detection, spline curve fitting of RANSAC and Bessel algorithm, and the algorithm of lane image detection and recognition, including image preprocessing, transformation of IPM view, filtering and noise reduction, detection of Hough line, spline curve fitting of RANSAC and Bessel algorithm. The expansion of fitting curve and so on. The relative position between the intelligent electric vehicle and the lane line is determined by the lane line extracted, and the lane line is tracked by warning when the lane deviates. The data fusion of the extracted information with other sensors of the environment sensing module aims to complete the autonomous navigation of the intelligent electric car. Finally, the algorithm of lane line detection is verified. The measured results show that the improved lane detection algorithm is accurate and stable, which is not clear for the occlusion of vehicles, the shadow of trees, and the unclear lane lines. The typical road environment such as obviously curved lane line can accurately detect the lane line, which provides the feasibility for intelligent electric car to drive independently in the park environment.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】