文献类型：期刊文献

中文题名：基于改进的SIFT算法交通标志牌的检测与识别

英文题名：Traffic Sign Detection and Recognition Based on the Improved SIFT Algorithm

作者：张琪[1];何宁[2];池悦[1];曹珊[1]

机构：[1]北京联合大学北京市信息服务工程重点实验室;[2]北京联合大学信息学院

第一机构：北京联合大学北京市信息服务工程重点实验室

年份：2017

卷号：31

期号：2

起止页码：59-65

中文期刊名：北京联合大学学报

外文期刊名：Journal of Beijing Union University

基金：国家自然科学基金项目(61572077)

语种：中文

中文关键词：交通标志牌;HSV颜色空间;识别;SIFT算法;最近邻分类器

外文关键词：Traffic sign; HSV color space; Recognition; SIFT algorithm; Nearest neighbor classifier

摘要：提出了一种针对自然场景下的标志牌的检测和识别算法,首先对获取的视频进行帧提取,采用图像增强算法对图像进行预处理,进而转换到HSV颜色空间,利用其颜色和形状特征进行检测定位,再根据感兴趣区域的面积特征排除多余目标,最后根据改进的SIFT特征匹配算法,利用最近邻分类器算法进行识别,样本库选用的是自然场景下的道路交通中的数据,通过对比实验发现,该算法在保证检测率的同时大大提高了算法的实时性。
We propose an approach for detecting circular traffic signs from images degraded by motion blur recorded in natural scenes. First, it extracts a key frame from the video image sequence, and uses the image enhancement algorithm for image preprocessing. Then, in HSV color space, testing positioning is made based on the color and shape features. After that, redundant targets will be eliminated according to the area of the region features. Finally, according to the improved SIFT feature matching algorithm, the nearest neighbor classifier algorithm is used for the recognition. The data in the sample library is from the road traffic data in natural scene. Through the contrast experiment, it is found that the real-time performance of the algorithm is greatly improved while ensuring the detection rate.