面向无人驾驶的高速公路指路标志字符检测
|
摘要
面向无人驾驶需要实时获得高速公路道路视觉信息的要求,提出一种高速公路指路标志字符检测方法。通过颜色和几何形状特征,得到高速公路指路标志牌,并以标志牌为对象检测字符。检测字符的步骤为:通过二值化方法得到二值图,经过矫正方法将倾斜的标志牌进行倾斜矫正,并对二值图进行过滤;通过多级投影策略对二值图进行切割,得到字符块,再根据版面排列对相关联的字符块进行合并得到字符区域;提取字符区域的HOG特征,通过支持向量机去除伪目标字符区域。实验结果表明在自然场景下高速公路指路标志字符检测的查全率达到97%以上,准确率达到96%以上。
For the requirement of unmanned driving to obtain the visual information of expressway road in real time,a method for detecting the character of expressway sign was proposed. By using the color and the geometric features,the method obtained the highway guide sign plate and detected the character with the sign board. The steps of detecting characters were as follows. First,the binary image was obtained by the binarization method. The inclined sign was corrected by the correction method and the binary image was filtered. Secondly,the character block was obtained by using the multi-level projection strategy for the binary image,and the associated character blocks were merged according to the layout to get the character area. Finally,the HOG feature of the character area was extracted,and the pseudo target character area was removed by the support vector machine. The results of experiment indicated that recall rate of character detection exceeded 97%,and precision rate was more than 96%.
For the requirement of unmanned driving to obtain the visual information of expressway road in real time,a method for detecting the character of expressway sign was proposed. By using the color and the geometric features,the method obtained the highway guide sign plate and detected the character with the sign board. The steps of detecting characters were as follows. First,the binary image was obtained by the binarization method. The inclined sign was corrected by the correction method and the binary image was filtered. Secondly,the character block was obtained by using the multi-level projection strategy for the binary image,and the associated character blocks were merged according to the layout to get the character area. Finally,the HOG feature of the character area was extracted,and the pseudo target character area was removed by the support vector machine. The results of experiment indicated that recall rate of character detection exceeded 97%,and precision rate was more than 96%.
引文
[1]Ellahyani A,Ansari M E,Jaafari I E.Traffic sign detection and recognition based on random forests[J].Applied Soft Computing,2016,46(C):805-815.
[2]Lillo-Castellano J M,Mora-Jiménez I,Figuera-Pozuelo C,et al.Traffic sign segmentation and classification using statistical learning methods[J].Neurocomputing,2015,153:286-299.
[3]Zaklouta F,Stanciulescu B.Real-time traffic sign recognition in three stages[J].Robotics and Autonomous Systems,2014,62(1):16-24.
[4]Shi C,Wang C,Xiao B,et al.Scene text detection using graph model built upon maximally stable extremal regions[J].Pattern Recognition Letters,2013,34(2):107-126.
[5]Li X,Wang W,Jiang S,et al.Fast and effective text detection[C]//Proceedings of the 15th IEEE International Conference on Image Processing,2008:969-972.
[6]Gonzalez A,Bergasa L M,Yebes J,et al.Text location in complex images[C]//Proceedings of the 21st International Conference on Pattern Recognition,2012:617-620.
[7]谷明琴,蔡自兴,任孝平.自然场景中字符型交通标志的检测方法[J].中南大学学报(自然科学版),2013,44(5):1861-1868.
[8]王振,魏志强.交通标志牌字符提取算法[J].计算机应用,2011,31(1):266-269.
[9]刘居峰.自然场景中交通标志文字检测算法研究[D].北京:北京交通大学,2014.
[10]陈亚杰.自然场景中交通标志版面理解算法研究[D].北京:北京交通大学,2016.
[11]Dalal N,Triggs B.Histograms of Oriented Gradients for Human Detection[J].CVPR,2005,1(12):886-893.
[12]徐渊,许晓亮,李才年,等.结合SVM分类器与HOG特征提取的行人检测[J].计算机工程,2016,42(1):56-60.
[13]纪娟,秦珂,杨若瑜.基于HOG和几何特征的服装细节要素识别与分类[J].图学学报,2016,37(1):84-90.
[14]王见,陈义,邓帅.基于改进SVM分类器的动作识别方法[J].重庆大学学报(自然科学版),2016(1):12-17.
[15]张云龙,齐国红,张晴晴.基于SVM的复杂环境下苹果图像的分割[J].智能计算机与应用,2017,7(1):14-16.
[16]张霞,黄继风.结合LBP直方图和SVM的视频火焰检测[J].计算机应用与软件,2016,33(8):216-220.
[17]Burges C J C.A Tutorial on Support Vector Machines for Pattern Recognition[J].Data Mining&Knowledge Discovery,1998,2(2):121-167.
[2]Lillo-Castellano J M,Mora-Jiménez I,Figuera-Pozuelo C,et al.Traffic sign segmentation and classification using statistical learning methods[J].Neurocomputing,2015,153:286-299.
[3]Zaklouta F,Stanciulescu B.Real-time traffic sign recognition in three stages[J].Robotics and Autonomous Systems,2014,62(1):16-24.
[4]Shi C,Wang C,Xiao B,et al.Scene text detection using graph model built upon maximally stable extremal regions[J].Pattern Recognition Letters,2013,34(2):107-126.
[5]Li X,Wang W,Jiang S,et al.Fast and effective text detection[C]//Proceedings of the 15th IEEE International Conference on Image Processing,2008:969-972.
[6]Gonzalez A,Bergasa L M,Yebes J,et al.Text location in complex images[C]//Proceedings of the 21st International Conference on Pattern Recognition,2012:617-620.
[7]谷明琴,蔡自兴,任孝平.自然场景中字符型交通标志的检测方法[J].中南大学学报(自然科学版),2013,44(5):1861-1868.
[8]王振,魏志强.交通标志牌字符提取算法[J].计算机应用,2011,31(1):266-269.
[9]刘居峰.自然场景中交通标志文字检测算法研究[D].北京:北京交通大学,2014.
[10]陈亚杰.自然场景中交通标志版面理解算法研究[D].北京:北京交通大学,2016.
[11]Dalal N,Triggs B.Histograms of Oriented Gradients for Human Detection[J].CVPR,2005,1(12):886-893.
[12]徐渊,许晓亮,李才年,等.结合SVM分类器与HOG特征提取的行人检测[J].计算机工程,2016,42(1):56-60.
[13]纪娟,秦珂,杨若瑜.基于HOG和几何特征的服装细节要素识别与分类[J].图学学报,2016,37(1):84-90.
[14]王见,陈义,邓帅.基于改进SVM分类器的动作识别方法[J].重庆大学学报(自然科学版),2016(1):12-17.
[15]张云龙,齐国红,张晴晴.基于SVM的复杂环境下苹果图像的分割[J].智能计算机与应用,2017,7(1):14-16.
[16]张霞,黄继风.结合LBP直方图和SVM的视频火焰检测[J].计算机应用与软件,2016,33(8):216-220.
[17]Burges C J C.A Tutorial on Support Vector Machines for Pattern Recognition[J].Data Mining&Knowledge Discovery,1998,2(2):121-167.