交通禁止标志的检测和识别

发布时间：2019-05-28 11:47

【摘要】：道路交通标志的检测和识别系统是智能交通系统中的重要组成部分,它为驾驶员在行车过程中提供指示、警告信息，为安全、便利的驾驶提供可靠的保障。国外对交通标志的识别研究开始于上世纪80年代，而我国在这个领域的研究起步比较晚。在日趋发达的道路交通系统中，交通标志的检测和识别将拥有更广泛的应用，因此，对交通标志自动检测和识别系统的研究具有重要的价值。本文将交通标志中的禁止标志作为研究对象，，首先介绍了禁止标志检测与识别的研究背景、研究意义、国内外的研究状况以及研究过程中存在的技术难点。其次，介绍了采用改进的霍夫变换检测圆的算法对图像进行检测，得到圆心坐标并计算出圆的半径，然后在原图像中标识出禁止标志区域的方法。接着，介绍了如何在HSV颜色空间中根据禁止标志的颜色特征去掉禁止标志中的干扰背景，如何对得到的图像进行各种形态学处理。最后，采用改进的Hu不变矩法提取待识别禁止标志的形状不变矩特征值，采用欧几里德距离计算方法计算待识别图像与样本库中模板图像的相似度，根据相似度来识别禁止标志。本系统使用上述技术能在实际应用中有效的识别禁止标志，通过对100幅分辨率在207×206~1024×776范围内的图像进行测试，识别率为83%，其中，50幅分辨率为1024×776图像的识别率达到了96%，基本实现了预期识别率为90%的设计目标。
[Abstract]:The detection and identification system of road traffic sign is an important part of the intelligent traffic system. It provides an indication and warning information for the driver to provide reliable guarantee for safe and convenient driving. The recognition and study of traffic signs in foreign countries began in the 1980s, and the research in this field was very late in our country. In the increasingly developed road traffic system, the detection and identification of traffic signs will have a wider application, so the research on the automatic detection and recognition system of traffic signs is of great value. In this paper, the prohibition sign in traffic sign is used as the research object, and the research background, research significance, the research situation at home and abroad and the technical difficulties existing in the research are introduced. and secondly, the method for detecting a circle by adopting an improved Hough transform detection circle is introduced, a circle center coordinate is obtained and a circle radius is calculated, Then, how to remove the background of the interference from the color feature of the prohibited mark in the HSV color space is introduced, and how to carry out various morphological operations on the obtained image And finally, adopting an improved Hu constant-moment method to extract the shape-invariant moment characteristic value of the forbidden mark to be identified, calculating the similarity of the image to be identified and the template image in the sample library by using the Euclidean distance calculation method, and identifying the forbidden mark according to the similarity. The system uses the above-mentioned technology to effectively identify the forbidden mark in the practical application, and the recognition rate is 83% by testing the image in the range of 207-1024-776-776 with the 100-frame resolution, wherein the recognition rate of the 50-frame image with the resolution of 1024-776 is achieved. 96%, basically realized the expected recognition rate of 90%
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.52;U495

【参考文献】