复杂城市环境下智能车导航定位方法研究

发布时间：2018-02-01 04:44

本文关键词： 全球导航卫星系统复杂环境地图匹配双目视觉辅助定位系统 Vision/GNSS融合导航定位　出处：《电子科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：具有精确、稳定的定位结果以及合理的价格是未来的智能车辆导航系统的发展趋势。为了达到这个目标,人们建立了多种组合导航模型(GNSS/DR,GNSS/INS,GNSS/MM)。尽管这些模型已在多种不同环境中成功应用,但它们仍有许多缺陷,尤其是在GNSS系统定位精度受到威胁的区域。因而复杂城市环境下智能车辆的导航方法研究也成为了一个关注度高的研究领域。本课题系统地研究了多种城市复杂环境对导航定位结果所产生的影响,并对不同的环境采取相应针对性的导航方法,最终通过实验采集真实数据验证所提方法的正确性及可行性。论文主要研究内容分为六部分:1.将复杂城市环境分为模糊道路环境和城市峡谷环境类型,并针对不同类型的复杂环境建立不同的导航定位方法及模型来提高对应复杂环境下的定位精度。2.提出了一种改进的基于多权重值WΔω,Wd与Wθ的地图匹配算法,通过多种权重值的综合从多条候选路段中选取GPS定位点的最佳匹配路段。在MATLAB和Visual C++上进行仿真将原始具有偏差的轨迹和修正之后的轨迹进行对比,验证了该算法适用于模糊道路环境,实现对导航轨迹偏差的修正。3.研究了传统双目视觉测量的原理,利用经典的张氏标定法对摄像机进行内、外参数的标定。完成了视觉坐标系下的定位算法的设计,对视觉测量误差进行了分析。4.设计了一种基于路标的双目视觉辅助GNSS定位的方法。建立了从路标检测识别到视觉测量,再到辅助定位的整套辅助定位的流程体系:随机霍夫变换RHT用于路标检测,SIFT与K-means算法将用于路标的匹配识别;双目视差计算智能车与路标之间的向量,从而建立辅助定位模型计算车辆的位置。5.设计了双目采集的图像软件,利用实验车在一处复杂环境区域进行实时数据采集,通过计算出的双目视觉定位误差与GNSS定位误差对比分析,验证了该方法在路标可见范围内对GNSS定位结果有明显改善。6.研究了Kalman滤波算法,将视觉信息与GNSS信息进行融合滤波,建立了Vision/GNSS相结合导航的数据融合模型。
[Abstract]:Accurate, stable positioning results and reasonable price are the future trend of intelligent vehicle navigation system. In order to achieve this goal, people have established a variety of integrated navigation models GNSS / Dr. Although these models have been successfully used in many different environments, they still have many drawbacks. Especially in the area where the positioning accuracy of GNSS system is threatened, so the research on the navigation method of intelligent vehicle in complex urban environment has become a research field with high attention. The influence of complex environment on the result of navigation and positioning. And the corresponding targeted navigation methods are adopted for different environments. Finally, the validity and feasibility of the proposed method are verified by collecting real data through experiments. The main content of this paper is divided into six parts: 1. The complex urban environment is divided into fuzzy road environment and urban canyon environment. Different navigation methods and models are established for different complex environments to improve the positioning accuracy. 2. An improved method based on multi-weight value W 螖蠅 is proposed. The map matching algorithm of W d and W 胃. Through the synthesis of multiple weight values, the best matching section of GPS positioning points is selected from a number of candidate sections. In MATLAB and Visual C. The original track with deviation is compared with the modified trajectory. It is verified that the algorithm is suitable for fuzzy road environment, and the correction of navigation trajectory deviation is realized. 3. The principle of traditional binocular vision measurement is studied, and the classical Zhang's calibration method is used to carry out the camera internalization. The calibration of external parameters. The design of location algorithm in visual coordinate system is completed. The visual measurement error is analyzed. 4. A binocular visual aided GNSS positioning method based on the road sign is designed, and the method from road sign detection to visual measurement is established. Then to the auxiliary location of the whole set of auxiliary positioning process system: random Hough transform RHT for road sign detection sift and K-means algorithm will be used for road sign matching recognition; The binocular parallax is used to calculate the vector between the intelligent vehicle and the road sign, and then the auxiliary positioning model is established to calculate the vehicle position. 5. The image software of binocular acquisition is designed. The experiment vehicle is used to collect real time data in a complex environment, and the binocular vision positioning error and GNSS positioning error are compared and analyzed. It is verified that this method can improve the GNSS location results in the visible range of road signs. 6. The Kalman filtering algorithm is studied, and the visual information and GNSS information are fused and filtered. A data fusion model combining Vision/GNSS with navigation is established.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP391.41

【参考文献】