基于中心投影形状特征的车载移动测量系统交通标志自动识别  

Automatic recognition of road traffic sign based on central projected shape feature for vehicle-borne mobile mapping system

作　　者：张卡[1] 盛业华[1] 叶春[1] 李志英[1] 

机构地区：[1]南京师范大学虚拟地理环境教育部重点实验室,南京210046

出　　处：《仪器仪表学报》2010年第9期2101-2108,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金项目(40901200);江苏省高校自然科学基础研究项目(07KJA42005);南京师范大学科研启动基金(2008105XGQ0149)资助项目

摘　　要：针对车载移动测量系统所拍摄图像中的交通标志的自动识别问题,提出了一种基于中心投影特征的交通标志自动识别算法。文中详细研究了基于自适应图像分割和中心投影变换的交通标志图像的形状特征计算方法,利用信息熵解决了中心投影变换的最佳投影个数确定问题。在中心投影特征计算的基础上,利用训练后的概率神经网络实现了交通标志具体含义的自动识别。使用车载移动测量系统所拍摄的实际交通标志图像对本文算法进行了实验,并将中心投影形状特征和不变矩特征及SIFT特征的识别效果进行了对比,实验结果表明基于中心投影特征的识别方法具有最高的识别准确率。Aiming at the problem of automatic traffic sign recognition in natural scene images taken by vehicle-borne mobile mapping system,a new algorithm for traffic sign recognition based on central projected shape feature is proposed.In this paper,the shape feature computation method of traffic sign based on adaptive image segmentation and central projection transformation is studied in detail.The problem of determining the optimal projection number in central projection transformation is solved using information entropy.On the basis of central projected feature computation,automatic recognition of traffic sign is realized by the trained probabilistic neural network.This new algorithm is applied to the real scene images taken by the vehicle-borne mobile mapping system in Nanjing at different time.Moreover,the recognition performances of traffic sign recognition using central projected shape feature and other features such as invariant moments and SIFT feature are compared and investigated.Experiment results show that the proposed new method has the highest recognition rate.

关 键 词：道路交通标志识别 车载移动测量系统 中心投影变换 形状特征 概率神经网络 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]