东篱科研大数据发现系统（DRDS）

位置：成果数据库 > 期刊 > 期刊详情页

基于径向对称变换的自适应交通禁止标志检测

ISSN号：1005-0086
期刊名称：《光电子．激光》
时间：0
分类：TP391[自动化与计算机技术—计算机应用技术;自动化与计算机技术—计算机科学与技术]
作者机构：[1]大连理工大学电子信息与电气工程学部,辽宁大连116024, [2]大连民族学院信息与通信工程学院,辽宁大连116605, [3]大连民族学院计算机科学与工程学院,辽宁大连116605
相关基金：国家自然科学基金（81241059,61172108,61139001）、国家科技支撑计划（2012BAJl8806）、中央高校基本科研基金（DCll0313,DCl20101073）和国家民委科研（12DLZ011）资助项目

作者：李厚杰[1,2], 邱天爽[1], 宋海玉[3], 王培昌[2], 贺建军[2]

关键词：区域轮廓特征, 分水岭变换, 径向对称变换, 参数自适应性, 交通禁止标志检测, regional contour feature, watershed transfonm, radial symmetry transform, parameter adapt- ability, traffic prohibitive sign detection

中文摘要：

提出一种基于径向对称变换的自适应交通禁止标志的检测算法。采用RGB归一化阈值分割算法对交通图像进行二值化处理，构造和利用区域轮廓特征矢量，判决和提取候选标志区域。对于标志互相遮挡候选区域，提出采用基于分水岭变换的自适应标志分离算法进行标志分离；然后，对于低维的标志候选区域，根据其自身尺度特征提出一种参数自适应的径向对称圆形检测器和检测算法，最终确定禁止标志。本文算法在标准交通标志数据集（GTSDB）进行了实验验证。实验结果表明，与现有方法相比，本文算法有效提高了算法执行效率和检测性能，降低了算法复杂度；并且，对交通标志尺度、亮度和天气变化、运动模糊以及标志互相遮挡等有着良好的鲁棒性。

英文摘要：

Traffic sign detection is a key technology in driving assistance systems. The effective detection for prohibitive signs is crucial to erasure the driver safety and reduce the traffic accidence. In this paper, we propose an adaptive detection method for prohibitive signs based on the radial symmetry transform. A traffic image is segmented based on RGB normalized threshold algorithm and a binary image is genera ted. Then a regional contour feature vector is constructed and used to determine and extract every candi- date sign region. For the region with possible mutually occluding signs, an adaptive separation algorithm based on the watershed transform is presented to perform signs separation. For each abstracted candidate sign region, the key parameters in the radial symmetry transform are set adaptively according to the scale feature of the region,and an adaptive radial symmetry circular detector and corresponding detection ap- proach are constructed and used to detect the prohibitive signs and reject the non-sign regions. The per- {ormance of the proposed method is tested on the German traffic sign detection benchmark （GTSDB）. Experimental results demonstrate the effectiveness of our method. Our method outperforms the existing algorithms in terms of execution efficiency, detection rate and algorithm complexity, and ours has the ro- bustness to the obiect scale variation,variable lighting condition and weather,motion blurring and mutual occlusion of signs, etc.

同期刊论文项目