摘要
针对客运列车车号特征,提出一种基于小波分解的车号定位及阈值分割方法。首先对灰度图像进行中值滤波和拉普拉斯锐化处理,然后采用小波分解和改进的Canny边缘检测算法,并结合形态学处理和车号先验知识对车号精确定位,最后,结合Niblack算法,提出一种连通域个数指导阈值分割的方法。实验结果表明,该算法能适应不同光照条件下多型车厢的车号定位,二值化效果好,具有较好的鲁棒性。
Aiming at the features of passenger train number,a new method,which is based on wavelet decomposition,is proposed for the localization and threshold segmentation of train number. The gray-level image is firstly smoothed by using median fil-ter,and followed by Laplace sharpening processing. Then the wavelet decomposition and improved Canny edge detection algo-rithm are used,and the morphological processing prior knowledge of the rehicle number are used to locate the region accurately. Fi-nally,a method of threshold segmentation combined Niblack with the number of connected domain is proposed. As experimental re-sults show,this method is more insensitive to different types of train number and illumination variance,and performs better in binarization and robustness.
引文
[1]杨绍华,李建勇,王恒.基于BP神经网络的货车车号识别方法研究[J].铁路计算机应用,2007,16(12):4-7.YANG Shaohua,LI Jianyong,WANG Heng. Research onmethod of vehicle number recognition based on BP NeuralNetwork[J]. Railway Computer Application,2007,16(12):4-7.
[2]陈春雷,王阳萍,祝咏升.基于视频列车车号识别系统的研究[J].兰州交通大学学报,2007,26(6):55-61.CHEN Chunlei,WANG Yangping,ZHU Yongsheng. Re-search on Recognition System of Train Character Based onVideo[J]. Journal of Lanzhou Jiaotong University,2007,26(6):55-61.
[3]杨大志.基于图像处理的铁路货车车号识别及其应用研究[D].成都:西南交通大学,2010:29-35.YANG Dazhi. Train Wagon Number Recognition and Ap-plication Based on Image Processing[D]. Chengdu:South-west Jiaotong University,2010:29-35.
[4]王欣蔚.基于深度学习的TEDS系统车号识别算法研究[D].北京:北方工业大学,2016:17-22.WANG Xinwei. Research on TEDS Character RecognitionAlgorithm Based on Deep Learning[D]. Beijing:NorthChina University of Technology,2016:17-22.
[5] NIBLACK W. An introduction to image processing[J].Applied Computing,1986,15(4):219-229.
[6]GUO J C,FEI Y L,HUA X P. Research on Vein ImagePreprocessing Based on NiBlack Algorithm[C]//Informa-tion and Management Sciences--Processings of the NinthInternational Conference on Information and ManagementSciences,2010:190-197.
[7] KILLINGBECK J P,JOLICARD G. The Niblack-Nigamiterative method revisited[J]. Journal of MathematicalChemistry,2010,47(3):929-936.
[8]曹怀信,郭志华.小波分析基础[M].北京:科学出版社,2016:152-164.CAO Huaixin,GUO Zhihua. Fundamental Wavelet Analy-sis[M]. Beijing:Science Press,2016:152-164.
[9]崔男,李婧,孙艳蕊.基于数学形态学的渐变色字符定位算法[J].计算机与数字工程,2016(10):2011-2014.CUI Nan,LI Jing,SUN Yanrui. Character Location Algo-rithm for the Image of Gradient Color Character Based onMathematical Morphology[J]. Computer and Digital Engi-neering,2016(10):2011-2014.
[10]徐天芝,张贵仓,贾园.基于形态学梯度的分水岭彩色图像分割[J].计算机工程与应用,2016(11):200-203.XU Tianzhi,ZHANG Guicang,JIA Yuan. Color imagesegmentation based on morphology gradients and wa-ter-shed algorithm[J]. Computer Engineering and Appli-cations,2016(11):200-203.
[11]许宏科,秦严严,陈会茹.一种基于改进Canny的边缘检测算法[J].红外技术,2014(3):210-214.XU Hongke,QIN Yanyan,CHEN Huiru. An ImprovedAlgorithm for Edge Detection Based on Canny[J]. Infra-red Technology,2014(3):210-214.
[12]张大坤,罗三明.形态学中闭运算功能的扩展及其应用[J].计算机工程与应用,2010(27):185-187.ZHANG Dakun,LUO Sanming. Function extension ofmorphological closing operation and its application[J].Computer Engineering and Applications,2010(27):185-187.
[13]杨叶梅.基于改进光流法的运动目标检测[J].计算机与数字工程,2011(9):108-110.YANG Yemei. Moving Objects Tracking Based on Im-proved Optical Flow Method[J]. Computer and DigitalEngineering,2011(9):108-110.
[14]OTSU N. A threshold selection method from gray-levelhistograms[J]. IEEE Trans. SMC,1979,9(1):62-66.
[15]李了了,邓善熙,丁兴号.基于大津法的图像分块二值化算法[J].微计算机信息杂志,2005(24):76-77.LI Liaoliao,DENG Shanxi,DING Xinghao. Binarization Algorithm Based on Image Partition Derived from Da-Jing Method[J]. Control and Automation,2005(24):76-77.