复杂场景下的交通标志识别技术研究

发布时间：2018-01-14 18:03

本文关键词：复杂场景下的交通标志识别技术研究　出处：《合肥工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着城市建设规模的不断扩大,市民生活水平正逐步提高,极大程度地加快了人们的生活节奏,私家车俨然成为众多市民出行的重要交通工具。然而,随着路面机动车辆的不断增加,在给人们带来便利的同时,也造成一些不容忽视的交通问题——道路交通拥堵和交通安全。而这些问题一方面与驾驶员的驾驶行为有着紧密的联系,如酒驾、路怒、驾驶经验和不规范的驾驶行为等;另一方面与我国尚未健全的交通管理措施有关,如何建立规范有效的解决方案已成为世界各国共同克服的难题。目前,有众多学者认为:大规模地普及应用智能交通技术,提高道路交通管理水平和智能驾驶系统运行质量是实现道路交通系统良性发展的一条有效途径。交通标志作为道路中一种重要的基础辅助设施,通过提供指示和禁令等驾驶信息对交通流量的疏导,并且规范驾驶员的交通行为,确保交通秩序井然有序的进行。道路交通标志识别(Road traffic sign recognition.RTSR)作为其智能驾驶辅助系统的关键技术之一,引起众多学者和相关科研机构的重视和研究。对于应用在实际场景下的交通标志识别系统,其所面临的道路背景日趋复杂,而且所设计的识别系统对实时性和高效性的要求较高,这无疑是一个巨大的挑战。本文主要针对复杂环境下的道路交通标志的定位和识别相关算法展开了讨论:首先依据交通标志的形状特征,提出了基于小波低频分量的边缘检测方法,检测出交通标志图像的边缘信息,接着,对其进行灰度形态学和纹理降噪处理,削弱灰度边缘图像中的噪声干扰区域,然后,对降噪后的灰度图像进行二值化处理,随后,采用改良的Hough变换方法分割出路标区域,进而定位出交通标志图像。对自然场景下拍摄的交通标志图像进行算法测试,试验表明:该算法的定位准确率为91.8%,平均定位时间为0.46s;最后,采用Hu不变矩提取样本和定位到的标志图像的特征向量,并使用支持向量机进行分类识别出其含义,同时采用了最小距离分类方法进行了比较。
[Abstract]:With the continuous expansion of the scale of urban construction, the standard of living of citizens is gradually improving, greatly accelerating the pace of people's lives, private cars have become an important means of transportation for many citizens to travel. With the increasing number of road motor vehicles, it brings convenience to people at the same time. It also causes some traffic problems that can't be ignored-road traffic congestion and traffic safety. On the one hand, these problems are closely related to driver's driving behavior, such as drunk driving, road rage. Driving experience and irregular driving behavior; On the other hand, with the unsound traffic management measures in China, how to establish a standardized and effective solution has become a common problem that all countries in the world overcome. Many scholars believe that the application of intelligent transportation technology is widely used in a large scale. Improving the level of road traffic management and the quality of intelligent driving system is an effective way to realize the benign development of road traffic system. Traffic signs as an important basic auxiliary facilities in the road. Through the provision of guidance and ban on driving information to the traffic flow, and regulate the traffic behavior of drivers. Ensure that traffic order is in order. Road traffic sign recognition. RTSR). As one of the key technologies of its intelligent driving assistance system. It has attracted the attention and research of many scholars and related research institutions. For the traffic sign recognition system which is applied in the actual scene, the road background is becoming more and more complex. And the design of the recognition system for real-time and high efficiency requirements. This is undoubtedly a huge challenge. This paper mainly discusses the location and recognition algorithms of road traffic signs in complex environment: firstly, according to the shape characteristics of traffic signs. An edge detection method based on wavelet low frequency component is proposed to detect the edge information of traffic sign image. Then the gray level morphology and texture denoising are processed. The noise interference area in the gray edge image is weakened, then the gray image is binary processed, and then the improved Hough transform method is used to segment the signpost area. Then the traffic sign image is located. The algorithm test of the traffic sign image taken under the natural scene shows that the localization accuracy of the algorithm is 91.8 and the average positioning time is 0.46s; Finally, Hu moment invariant moment is used to extract the feature vector of the image and the support vector machine (SVM) is used to classify and recognize its meaning. At the same time, the minimum distance classification method is compared.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】