-
摘要: 为了对道路标志图像进行自动分类, 通过图像颜色空间变换, 将图像的RGB量值转换为H (色度) S (饱和度) I (亮度) 量值, 采用Sobel算子进行道路标志图像的边缘检测, 利用行扫描法进行区域填充, 以获取二值化的道路标志图像区域, 提取道路标志二值化图像的不变矩与形状参数作为图像特征值, 设计BP神经网络道路标志图像几何形状分类器, 以道路标志图像的H、I为特征值, 设计了欧式距离分类器, 实现道路标志背景颜色的识别。融合道路标志图像几何形状和背景颜色的识别算法, 并利用道路标志的分类知识和自动分类方法, 能有效实现道路标志图像的自动识别。Abstract: To achieve automatic classifying traffic sign, a sign classifying method was put forward, sign HSI color model was obtained from sign RGB model, the edge of traffic sign image was detected by Sobel operator, sign region was filled by line-scanning method to gain the binary image of traffic sign, the invariant-moments and form factors of traffic sign image were extracted as image feathers, the shape of traffic sign was recognized by BP nerve net classifier, a euclid-distance classifier was designed to recognize the background color of traffic sign by H and I values.Automatic classifying result shows that the shape and color of traffic sign can be recognised accurately by the method.
-
Key words:
- traffic safety /
- traffic signs /
- driving safety assistance /
- image recognition
-
表 1 道路标志的形状与背景颜色
Table 1. Forms and background colors of traffic signs
种类 要素 形状 背景颜色 警告标志 三角形 黄色 禁令标志 圆形 白色 指示标志 圆形或矩形 蓝色 指路标志 矩形 高速公路为绿色或黄色, 一般公路为蓝色 表 2 道路标志图像形状特征值
Table 2. Feature values of traffic sign forms
图像序号 图像大小 ϕ1 ϕ2 ϕ3 ϕ4 ϕ5 ϕ6 ϕ7 F 1 280×280 0.235 9 5.483 1×10-4 4.000 0×10-3 7.400 0×10-3 -1.028 9×10-5 1.740 0×10-4 3.934 7×10-5 1.466 9 2 125×125 0.228 1 2.528 7×10-4 4.000 0×10-3 5.800 0×10-3 -6.728 2×10-6 9.275 9×10-4 2.741 6×10-5 1.493 5 3 125×125 0.228 1 2.531 8×10-4 4.000 0×10-3 5.800 0×10-3 -6.736 1×10-6 9.288 0×10-5 2.745 5×10-5 1.493 7 4 85×73 0.193 5 1.082 9×10-4 4.100 0×10-3 4.225 6×10-4 1.309 8×10-8 4.319 2×10-6 5.530 8×10-7 2.647 2 5 100×84 0.190 6 4.768 6×10-7 4.000 0×10-3 1.700 5×10-4 2.360 1×10-8 -1.173 1×10-7 1.387 0×10-7 1.454 5 6 100×84 0.191 3 4.348 3×10-7 4.100 0×10-3 2.073 1×10-4 2.292 7×10-8 -1.337 7×10-7 1.898 4×10-7 1.458 5 7 300×256 0.193 3 3.395 2×10-5 4.100 0×10-3 5.178 4×10-4 1.036 1×10-7 3.016 3×10-6 7.495 7×10-7 1.461 5 8 204×175 0.203 3 1.100 0×10-3 4.200 0×10-3 2.300 0×10-3 2.648 2×10-6 7.584 2×10-6 6.753 2×10-6 4.660 8 9 120×57 0.379 6 8.340 0×10-2 4.690 0×10-2 8.540 0×10-2 5.400 0×10-3 2.470 0×10-2 8.863 0×10-5 2.944 9 10 124×80 0.273 7 2.840 0×10-2 1.040 0×10-2 2.400 0×10-2 3.778 2×10-4 4.000 0×10-3 1.111 8×10-5 14.942 2 11 119×53 0.512 9 2.041 0×10-1 1.820 0×10-1 2.387 0×10-1 4.980 0×10-2 1.078 0×10-1 1.655 5×10-6 1.557 9 12 133×90 0.249 8 2.120 0×10-2 6.000 0×10-3 1.720 0×10-2 1.754 2×10-4 2.500 0×10-3 -1.193 0×10-6 1.309 5 13 109×100 0.168 2 6.626 1×10-6 4.723 8×10-7 1.854 0×10-4 6.653 3×10-10 3.744 0×10-7 -1.602 4×10-9 1.241 7 14 134×133 0.167 1 1.744 9×10-4 2.561 4×10-7 7.448 4×10-6 -9.333 1×10-12 -9.812 2×10-8 -4.328 6×10-12 1.241 4 15 158×158 0.175 3 2.500 0×10-3 3.357 8×10-6 3.512 3×10-6 1.169 7×10-11 6.990 1×10-9 -2.946 3×10-12 1.420 3 16 125×125 0.167 7 8.975 2×10-4 3.402 5×10-4 4.866 7×10-5 -5.635 1×10-9 -4.472 4×10-7 2.732 1×10-9 1.363 7 17 300×300 0.159 4 3.088 2×10-5 7.027 1×10-8 1.613 1×10-5 1.515 4×10-11 8.759 4×10-8 8.079 4×10-12 0.987 3 18 300×258 0.173 5 3.833 5×10-4 1.316 7×10-5 1.600 0×10-3 2.376 1×10-7 3.181 3×10-5 7.167 4×10-9 0.986 4 19 300×284 0.159 8 8.319 9×10-6 5.645 0×10-8 6.388 1×10-5 1.191 9×10-10 1.812 3×10-7 2.257 9×10-10 0.988 6 20 300×279 0.163 6 4.816 9×10-5 6.650 8×10-7 4.725 8×10-4 8.377 4×10-9 3.278 9×10-6 -1.079 0×10-10 0.988 1 21 100×100 0.159 2 7.022 6×10-7 2.252 1×10-8 5.786 7×10-6 -1.764 3×10-12 4.045 0×10-9 1.118 6×10-12 0.982 5 22 100×100 0.159 3 2.221 1×10-6 2.536 5×10-8 1.282 9×10-5 -5.651 2×10-12 4.311 5×10-9 -4.649 4×10-12 0.983 9 23 125×125 0.169 4 6.864 0×10-5 3.964 4×10-7 1.100 0×10-3 2.011 4×10-8 9.031 5×10-6 1.046 9×10-8 0.982 6 24 125×125 0.160 7 1.915 5×10-7 3.807 7×10-7 1.524 2×10-4 -6.435 7×10-11 5.887 1×10-8 1.159 4×10-9 0.984 8 表 3 道路标志图像形状识别结果
Table 3. Recognition result of traffic sign forms
序号 结果 分类器计算结果 判定结果 正确形状 识别结果 1 0.979 4 0.021 7 -0.000 8 (1, 0, 0) 三角形 正确 2 0.820 0 0.178 0 0.002 3 (1, 0, 0) 三角形 正确 3 0.068 8 0.925 4 0.006 0 (0, 1, 0) 矩形 正确 4 0.001 6 0.992 9 0.005 6 (0, 1, 0) 矩形 正确 5 -0.649 3 1.651 4 -0.002 0 (0, 1, 0) 矩形 正确 6 -0.472 4 1.466 7 0.005 7 (0, 1, 0) 矩形 正确 7 -0.681 0 1.676 5 0.004 6 (0, 1, 0) 矩形 正确 8 -0.813 7 1.814 8 -0.001 0 (0, 1, 0) 矩形 正确 9 -5.282 5 6.299 6 -0.016 9 (0, 1, 0) 矩形 正确 10 -1.173 7 2.170 9 0.002 8 (0, 1, 0) 矩形 正确 11 -0.035 3 1.047 8 -0.012 4 (0, 1, 0) 矩形 正确 12 -0.054 9 0.372 3 0.682 5 (0, 0, 1) 矩形 错误 13 0.006 9 -0.014 2 1.007 1 (0, 0, 1) 圆形 正确 14 0.000 8 -0.005 4 1.004 4 (0, 0, 1) 圆形 正确 -
[1] Gavrila M D M. Traffic sign recognition revisited[A]//Proc. of the 21st DAGM Symposium fur Mustererkennung[C]. Berlin: Springer Verlag, 1999. [2] Hsu S H, Huang C L. Road sign detection and recognition using matching pursuit method[J]. Image and Vision Computing, 2001, 19(3): 119-129. doi: 10.1016/S0262-8856(00)00050-0 [3] 高文, 陈熙霖. 计算机视觉[M]. 北京: 清华大学出版社, 1999. [4] 章毓晋. 图像处理和分析[M]. 北京: 清华大学出版社, 1999. [5] 杨枝灵. Visual C+ +数字图像获取处理及实践应用[M]. 北京: 人民邮电出版社, 2003. [6] 初秀民, 王荣本, 储江伟, 等. 沥青路面图像分割方法研究[J]. 中国公路学报, 2003, 16(3): 11-16. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL200303003.htmChu Xiu-min, Wang Rong-ben, Chu Jiang-wei, et al. Study of asphalt pavement surface distress image segmentation[J]. China Journal of Highway and Transport, 2003, 16(3): 11-16. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL200303003.htm [7] 魏连雨, 马永锋. 城市道路交通系统供需协调发展[J]. 交通运输工程学报, 2004, 4(4): 58-61. http://transport.chd.edu.cn/article/id/200404015Wei Lian-yu, Ma Yong-feng. Supply-demand coordination development of urban road traffic system[J]. Journal of Traffic and Transportation Engineering, 2004, 4(4): 58-61. (in Chinese) http://transport.chd.edu.cn/article/id/200404015 [8] 赵祥模, 南春丽, 施维颖. 交通工程CAE软件系统的设计与实现方法[J]. 交通运输工程学报, 2003, 3(3): 96-100. http://transport.chd.edu.cn/article/id/200303019Zhao Xiang-mo, Nan Chun-li, Shi Wei-ying. Design and realizing method of traffic engineering CAE system[J]. Journal of Traffic and Transportation Engineering, 2003, 3(3): 96-100. (in Chinese) http://transport.chd.edu.cn/article/id/200303019 [9] 魏朗, 高丽敏, 余强, 等. 驾驶员道路安全感受模糊评价模型[J]. 交通运输工程学报, 2004, 4(1): 102-105. http://transport.chd.edu.cn/article/id/200401025Wei Lang, Gao Li-min, Yu Qiang, et al. Fuzzy evaluating model of driver's road safety perception[J]. Journal of Traffic and Transportation Engineering, 2004, 4(1): 102-105. (in Chinese) http://transport.chd.edu.cn/article/id/200401025 [10] 巨永锋, 朱辉, 潘勇. 基于计算机视觉的车流量检测算法[J]. 长安大学学报: 自然科学版, 2004, 24(1): 92-95. https://www.cnki.com.cn/Article/CJFDTOTAL-XAGL200401023.htmJu Yong-feng, Zhu Hui, Pan Yong. Vehicle flow detection algorithm based on computer vision[J]. Journal of Chang'an University: Natural Science Edition, 2004, 24(1): 92-95. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-XAGL200401023.htm