雾霾天气下交通标志的检测与识别

发布时间：2018-04-10 11:45

本文选题：交通标志 + 雾天图像　；参考：《天津大学》2016年硕士论文

【摘要】：伴随着智能交通系统的迅速发展,道路交通标志识别作为智能交通系统的重要组成部分,逐渐成为当前研究热点。然而,在我国华北和东北地区经常出现的雾霾等恶劣天气,使现有的交通标志检测识别率严重降低,无法满足现实需求。为了提高雾霾天气下交通标志的检测识别率,本文设计了一套交通标志检测识别系统。首先,研究雾霾天气下标志识别系统的第一个环节——雾霾检测。通过研究大量同一场景清晰图像与雾霾图像的区别,得出雾霾图像的四个特征:图像亮度、图像对比度、图像辨识度、雾霾区域单一性,设计了一个基于图像亮度和图像对比度两个特征的Fisher分类器,来判别当前图像是否为雾霾图像。其次,图像被认证为雾霾图像后,需进行清晰化复原。雾天图像清晰化技术研究主要包括两个方向:一个是基于数学模型的图像复原技术,另一个是基于人体视觉的图像增强技术。概述了两个研究方向的经典算法,并完成基于MATLAB平台的试验仿真,总结仿真结果。文章采用图像复原技术,建立大气光衰减模型和环境光模型,两个模型的相互作用生成了大气雾霾图像成像的数学模型,模型反推恢复清晰图像。最后,进行交通标志的研究,包括两个环节:一个是标志的定位检测,另一个是标志的分类识别。在交通标志定位检测算法上,提出了一个改进的基于颜色信息分割和形状特征定位的交通标志检测算法;在标志识别环节中采用减小搜索区域来缩短时间消耗的模板匹配算法,完成交通标志的识别工作。文章设计的雾霾天气下交通标志检测与识别系统,能显著提高系统在雾霾环境中标志的检测识别准确率,同时系统也适用于晴天环境。
[Abstract]:With the rapid development of intelligent transportation system, road traffic sign recognition, as an important part of intelligent transportation system, has gradually become a hot research topic.However, severe weather such as haze often occurs in North and Northeast China, which makes the existing traffic sign detection and recognition rate seriously reduced, and can not meet the actual needs.In order to improve the detection and recognition rate of traffic signs in haze weather, a set of traffic sign detection and recognition system is designed in this paper.Firstly, the first step of haze weather identification system-haze detection is studied.By studying the differences between a large number of clear images of the same scene and haze images, four characteristics of haze images are obtained: image brightness, image contrast, image identification, and the uniqueness of haze region.A Fisher classifier based on image brightness and image contrast is designed to determine whether the current image is a haze image.Secondly, after the image is certified as a haze image, it needs to be clear and restored.The research of fog image sharpening mainly includes two directions: one is image restoration based on mathematical model and the other is image enhancement based on human vision.In this paper, two classical algorithms are summarized, and the simulation results are summarized based on MATLAB platform.In this paper, the atmospheric light attenuation model and the ambient light model are established by using image restoration technology. The interaction of the two models generates the mathematical model of atmospheric haze image imaging, and the model recovers the clear image.Finally, the study of traffic signs includes two steps: one is the location detection of signs, the other is the classification and recognition of signs.In the traffic sign location detection algorithm, an improved traffic sign detection algorithm based on color information segmentation and shape feature location is proposed.Complete the identification of traffic signs.The traffic sign detection and recognition system under haze weather can improve the detection and recognition accuracy of the system in haze environment, and the system is also suitable for sunny weather.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U495;TP391.41

【相似文献】