户外交通标志检测与形状识别.pdfVIP

第 14 卷 　第 4 期 中国图象图形学报 Vol. 14 , No. 4 2009年 4 月 Jou rnal of Im age and Grap h ic s Ap r. , 2009 户外交通标志检测和形状识别 许少秋 (广东工业大学信息工程学院 ,广州 　510006) 摘 　要 　提出了一种新的识别户外不同交通标志形状的算法 。为了减少数字噪声的影响和分离提取独立的交通 标志形状 ,基于颜色分割输出的外边缘可通过离散曲线演变进行简化和分解 。正切空间中弧线的相似程度决定离 散曲线演变的程度 。形状的识别是通过模板匹配来实现的 ,待识别形状与模板之间的最小几何差异决定形状的类 别 。实验结果表明本算法是平移 、旋转和尺度恒定的 ,能够在复杂的交通场景中进行可靠的形状识别 。 关键词 　颜色分割 　模板匹配 　交通标志形状辨别法 中图法分类号 : TP391　　文献标识码 : A 　　文章编号 : 1006896 1 (2009) Traff ic S ign D etection and Shape Recogn ition in O utdoor Env ironm en ts XU Shaoqiu ( F acu lty of Inf orm a tion Eng ineering, Guangdong Un iversity of Technology, Guangzhou 510006) A b stract　A novel app roach for detecting and recogn ising variou s traffic sign shap e s in ou tdoor environm en ts is p ropo sed In order to reduce the influence of digital no ise, sep arate and extract the shap e of each individual traffic sign, the external boundarie s of traffic sign s segm ented ba sed on color inform ation are simp lified and decompo sed through discrete curve evolutionThe evolution level is determ ined by the arc sim ilarity m ea su re in tangen t sp aceThe recogn ition of a candidate shap e is ach ieved through m atch ing again st temp late sThe m in im um geom etric difference between a candidate shap e and temp late s is u tilized to determ ine the cla ssification of the candidate shap eThe exp erim en tal re su lts ju stify that the p ropo sed algorithm is tran slation, rotation, and scaling invariant, and gives reliab le shap e reco