采用分段分形维数识别车牌字符

肖旺新; 张雪; 黄卫

采用分段分形维数识别车牌字符

肖旺新^{1, 3,},
张雪²,
黄卫³

1.
嘉应学院计算机系, 广东梅州 514015
2.
东南大学计算机系, 江苏南京 210096
3.
东南大学智能运输系统研究中心, 江苏南京 210096

详细信息

作者简介:
肖旺新(1975-), 男, 湖南邵阳人, 东南大学博士研究生, 从事交通图像处理与识别研究

中图分类号: U491.51
计量
- 文章访问数: 844
- HTML全文浏览量: 494
- PDF下载量: 374
- 被引次数: 0
出版历程
- 收稿日期: 2003-07-28
- 刊出日期: 2004-04-25

Recognition of vehicle license character based on multi-fractal dimension

1.
Department of Computer Science and Engineering, Jiaying University, Meizhou 514015, China
2.
Department of Computer Science and Engineering, Southeast University, Nanjing 210096, China
3.
ITS Research Center, Southeast University, Nanjing 210096, China

More Information

Author Bio:
XIAO Wang-xin(1975-), male, doctoral student, 86-25-83793131, xzhangxue@seu.edu.cn

Article Text (Baidu Translation)

摘要

摘要: 研究了在小波多尺度分解情况下, 采用分形维数描述车牌字符特征, 并在分区情况下计算出典型车牌字符的分段分形维数。发现小波变换能获取丰富的车牌字符特征, 采用分段分形维数描述车牌解决了单一维数描述车牌字符时, 有些字符维数相差太小, 导致阈值难以选取的问题。结果表明, 小波分形方法是研究车牌字符识别的有效的新方法。
- 交通信息工程 /
- 车牌 /
- 字符识别 /
- 二次B样条小波 /
- 分段分形维数
Abstract: Fractal dimension of some typical vehicle license characters were calculated in the condition of three zoning to vehicle license. It is found that prolific property of vehicle license can be obtained through wavelet transform, multi-fractal dimension is a good tool to solve the problem that threshold is hard to decide when too close between some characters using single fractal dimension. Simulation results show that the method is feasible to recognize vehicle license characters.
- traffic information engineering /
- vehicle license /
- character recognition /
- quadric B-spline wavelet /
- multi-fractal dimension

HTML全文

图 1 二维离散小波变换图

Figure 1. 2-D discrete wavelet transform

下载: 全尺寸图片幻灯片

图 2 图像分解过程

Figure 2. Image decomposed processes

下载: 全尺寸图片幻灯片

表 1 离散滤波器的离散系数

Table 1. Discrete coefficients of discrete filters

n	-3	-2	-1
H			0.125
G
K	0.007 813	0.054 685	0.171 875
n	0	1	2	3
H	0.375	0.375	0.125
G	-2.0	2.0
K	-0.171 875	-0.054 685	-0.007 813	0

下载: 导出CSV

表 2 汉字模板分段分形维数

Table 2. Multiscale fractal dimension of Chinese characters template

字符	H₁	H₂	H₃	字符	H₁	H₂	H₃
苏	0.5156	0.5187	0.7461	宁	0.3043	0.5384	0.7196
鄂	0.5667	0.4203	0.9904	津	0.1220	0.7705	0.5984
沪	0.3803	0.6125	0.8465	吉	0.3760	0.6429	0.7718
琼	0.5802	0.4214	0.8012	蒙	0.3808	0.8509	0.5033
辽	0.2916	0.4513	0.7693	黑	0.3420	0.5986	0.6743
桂	0.6349	0.4803	0.7081	海	0.3768	0.5946	0.8951
晋	0.3971	0.7559	0.7399	甘	0.3710	0.8260	0.3833
川	0.3497	0.5321	0.7240	台	0.5480	0.4129	0.7131
浙	0.4647	0.5795	0.7060	皖	0.4483	0.6142	0.7938
滇	0.5225	0.5805	0.6750	疆	0.5287	0.7784	0.5139
赣	0.6531	0.6518	0.5989	粤	0.6221	0.5701	0.7784

下载: 导出CSV

表 3 字母模板分段分形维数

Table 3. Multiscale fractal dimension of letter template

字符	H₁	H₂	H₃	字符	H₁	H₂	H₃
A	0.476 3	0.344 8	0.619 0	N	0.782 8	0.016 3	0.748 4
B	0.076 2	0.474 9	0.874 5	O	0.3350	0.4008	0.6075
C	0.382 6	0.297 1	0.573 1	P	0.4090	0.2858	0.5572
D	0.022 8	0.455 9	0.758 9	Q	0.2701	0.5154	0.4281
E	0.404 6	0.291 9	0.747 1	R	0.5131	0.2644	0.5185
F	-0.294 7	0.550 7	0.727 3	S	0.3458	0.4496	0.5617
G	0.337 9	0.442 2	0.630 5	T	0.3176	0.4816	0.5748
H	0.302 7	0.394 2	1.002 8	U	0.3847	0.3705	0.5585
I	-0.828 5	0.626 9	0.622 3	V	0.4199	0.3976	0.6158
J	0.442 3	0.332 6	0.374 6	W	0.4484	0.4847	0.7105
K	0.040 6	0.499 7	0.676 0	X	0.4233	0.3526	0.5589
L	0.122 3	0.406 7	0.885 2	Y	0.3482	0.4036	0.6003
M	0.314 1	0.291 2	1.110 4	Z	0.3615	0.4625	0.6040

下载: 导出CSV

表 4 数字模板分段分形维数

Table 4. Multiscale fractal dimension of number template

字符	H₁	H₂	H₃	字符	H₁	H₂	H₃
0	0.3810	0.2544	0.8158	5	0.3932	0.2626	0.7231
1	0.1368	0.4235	0.5663	6	0.4295	0.5402	0.4121
2	0.4020	0.4878	0.4324	7	0.4743	0.3124	0.6525
3	0.3790	0.3373	0.7071	8	0.4054	0.4491	0.6426
4	0.3595	0.4390	0.4368	9	0.3430	0.6030	0.4103

下载: 导出CSV

表 5 易混淆的字符分段分形维数

Table 5. Multiscale fractal dimension of undistinguishable characters template

字符	H₁	H₂	H₃	字符	H₁	H₂	H₃
D	0.0228	0.4559	0.7589	8	0.4054	0.4491	0.6426
O	0.3350	0.4008	0.6075	B	0.0762	0.4749	0.8745
黔	0.5406	0.7208	0.7660	7	0.4743	0.3124	0.6525
陕	0.3983	0.6111	0.8561	1	0.1368	0.4235	0.5663
1	0.1368	0.4235	0.5663	A	0.4763	0.3448	0.6190
I	-0.8285	0.6269	0.6223	4	0.3595	0.4390	0.4368

下载: 导出CSV

表 6 汉字12维数据信息

Table 6. 12-D data information of Chinese characters template

字符	h_a1	h_a2	h_a3	h_h1	h_h2	h_h3	h_v1	h_v2	h_v3	h_d1	h_d2	h_d3
鲁	0.8398	0.8442	0.7569	0.4681	0.6423	0.9245	0.3288	0.2924	0.6736	0.3560	0.0856	0.5124
川	0.8070	0.8261	0.7747	0.0581	0.3429	0.2841	0.3097	0.5192	0.7392	0.1926	0.1262	-0.0320
浙	0.8463	0.8534	0.7954	0.2826	0.5571	0.7339	0.3419	0.4818	0.7110	0.3149	0.2243	0.3822
滇	0.8465	0.8528	0.7932	0.5379	0.4264	0.8310	0.2411	0.5537	0.5795	0.4134	0.2143	0.6437
赣	0.8153	0.7902	0.7570	0.5975	0.7312	0.5345	0.5290	0.3998	0.4970	0.3826	0.2882	0.6075
辽	0.7630	0.8076	0.7862	0.3509	0.3919	0.5294	0.0774	0.2978	0.6285	0.2929	0.1869	0.3699
桂	0.7764	0.8131	0.7458	0.5457	0.5470	0.5947	0.6035	0.1711	0.6826	0.3332	0.1059	0.4763
鄂	0.7925	0.8415	0.7412	0.4589	0.5822	0.6590	0.4895	0.1480	0.8091	0.4222	0.1941	0.4353
沪	0.7837	0.7871	0.6930	0.0835	0.5466	0.5730	0.3620	0.4554	0.7767	0.2976	0.3233	0.2615

下载: 导出CSV

参考文献(13)

[1]	甘龙, 高隽, 梁栋, 等. 基于分形维数的车牌字符识别[J]. 中国公路学报, 2002, 15(4): 75-77. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL200204019.htm GAN Long, GAO Jun, LIANG Dong, et al. Recognition of vehicle licence character based on fractal dimension[J]. China Journal of Highway and Transport, 2002, 15(4): 75-77. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL200204019.htm
[2]	金一粟, 袁宝民, 于万波, 等. 基于分形盒子维数的车牌定位方法[J]. 计算机应用研究, 2002, 9(1): 40-42. https://www.cnki.com.cn/Article/CJFDTOTAL-JSYJ200209011.htm JIN Yi-su, YUAN Bao-min, YU Wan-bo, et al. A location method of vehicle-licence-plate based on fractal box dimension[J]. Research on Computer Application, 2002, 9(1): 40-42. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-JSYJ200209011.htm
[3]	戴青云, 余英林. 一种基于小波与形态学的车牌图像分割方法[J]. 中国图像图形学报, 2000, 5(5): 411-415. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGTB200005008.htm DAI Qing-yun, YU Ying-lin. A king of segmentation method of vehicle-licence-plate image based on wavelet and mathematical morphology[J]. Journal of Image and Graphics, 2000, 5(5): 411-415. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-ZGTB200005008.htm
[4]	沈会良, 李志能. 基于矩和小波变换的数字、字母字符识别研究[J]. 中国图像图形学报, 2000, 5(3): 249-252. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGTB200003015.htm SHEN Hui-liang, LI Zhi-neng. A study of number and letter character recognition based on moments and wavelet transform[J]. Journal of Image and Graphics, 2000, 5(3): 249-252. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGTB200003015.htm
[5]	Parisi R. Car plate recognition by neural network and image processing[A]. In Proc. IEEE ISCAS[C]. USA IEEE, 1998.
[6]	Elliman D G, Lancaster I T. Review of segmentation and contextal analysis techniques for text recognition[J]. Pattern Recognition Society, 1990, 23(4): 337-346.
[7]	辛厚文. 分形理论及其应用[M]. 合肥: 中国科学技术大学出版社, 1993.
[8]	王晓晖, 朱光喜, 朱耀庭. 图像的一种分形表示法及其应用[J]. 电子学报, 1997, 25(10): 28-31. https://www.cnki.com.cn/Article/CJFDTOTAL-DZXU199710007.htm WANG Xiao-hui, ZHU Guang-xi, ZHU Yao-ting. A representation of fractal property of images and its applications[J]. Acta Electronic Sinica, 1997, 25(10): 28-31. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-DZXU199710007.htm
[9]	Chui C K. An Introduction to Wavelet[M]. Academic Press. US, 1992.
[10]	肖旺新, 张雪, 黄卫. 视频交通图像自适应阈值边缘检测[J]. 交通运输工程学报, 2003, 3(4): 104-107. http://transport.chd.edu.cn/article/id/200304020 XIAO Wang-xin, ZHANG Xue, HUANG Wei. Adaptive thresholds edge detection of traffic image[J]. Journal of Traffic and Transportation Engineering, 2003, 3(4): 104-107. (in Chinese) http://transport.chd.edu.cn/article/id/200304020
[11]	Pentland A P. Fractal-based description of natural scenes[J]. IEEE France Pattern Analysis and Machine Intelligence, 1984, 6(6): 661-671.
[12]	程正兴. 小波分析算法与应用[M]. 西安: 西安交通大学出版社, 1998.
[13]	Barrow H G, Tenenhaum JM. Recovering inatrinaic scene characteristics from images[A]. Computer Vision Systems[C]. New York. Acdemic Press, 1978.