复杂场景中机动车行驶证快速检测与识别
|
摘要
为缩短机动车行驶证识别时间并提高识别准确率,本文提出一种基于复杂场景的机动车行驶证快速检测与识别算法.该算法首先进行区域定位,针对行驶证图片存在的背景复杂、角度倾斜的问题,提出背景模板匹配的区域提取算法对行驶证边缘轮廓的位置进行定位以完成倾斜校正,有效解决传统特征匹配方法检测时间长、正确率低的问题.之后使用关键区域的多尺度检测定位算法对校正结果进行正确性评判,避免错误的校正结果对后续识别的影响.接着对校正正确的图像进行模板分割得到字段区域,再进行区域二值化,由于光照不均和背景底纹导致传统的二值化算法效果不佳,提出融合二值化算法,解决了光照和底纹影响以及文字笔划粘连缺失的问题.算法最后通过识别引擎对二值图像进行识别,得到文字识别结果.实验表明本文提出的算法具有快速、多角度、背景与光照鲁棒等优点.
In order to accelerate the recognition procedure of vehicle license and promote accuracy,we proposed a fast detection and recognition scheme in complex scene. This scheme starts with region location. Due to that images are taken in complex scene and different angles,we use a new algorithm based on template matching to complete a correction. Compared to some frequently-used method,this algorithm we proposed can effectively make the correction in shorter time and with higher accuracy. Secondly,we apply a multi-scale detection algorithm to evaluate the correctness of the result from last step. After that we use a template to segment field areas before regional binarization. Some traditional binarization algorithms work poorly on account of uneven illumination and background shading. We present a new binarization algorithm based on integration. Finally,the recognition engine is used to get the text. With experiments on real samples,we find that the algorithms we proposed in this paper are faster,more accurate,and more robust to background complexity and uneven illumination,simultaneously they are theoretical and practical.
In order to accelerate the recognition procedure of vehicle license and promote accuracy,we proposed a fast detection and recognition scheme in complex scene. This scheme starts with region location. Due to that images are taken in complex scene and different angles,we use a new algorithm based on template matching to complete a correction. Compared to some frequently-used method,this algorithm we proposed can effectively make the correction in shorter time and with higher accuracy. Secondly,we apply a multi-scale detection algorithm to evaluate the correctness of the result from last step. After that we use a template to segment field areas before regional binarization. Some traditional binarization algorithms work poorly on account of uneven illumination and background shading. We present a new binarization algorithm based on integration. Finally,the recognition engine is used to get the text. With experiments on real samples,we find that the algorithms we proposed in this paper are faster,more accurate,and more robust to background complexity and uneven illumination,simultaneously they are theoretical and practical.
引文
[1] Wu Yi-quan,Zhang Jin-kuang. Vehicle license plates tilt detection based on the p-th pow er variance of projection coordinates and particle sw arm optimizatio[J]. Journal of Computer-Aided Design&Computer Graphic,2010,22(1):114-120.
[2] Miao Li-gang. License plate detection algorithm based on nearest neighbor chains[J]. Acta Automatica Sinica,2011,37(10):1272-1279.
[3] Tian Po-huang,Peng Tian-qiang,Li Bi-cheng. An approach for video-text extraction based on text traversing line and stroke connectivity[J]. Acta Electronica Sinica,2009,37(1):72-78.
[4] Milyaev S,Barinova O,Novikova T,et al. Image binarization for end-to-end text understanding in natural images[C]. International Conference on Document Analysis and Recognition,IEEE,2013,37(5):128-132.
[5] Escande P,Weiss P,Zhang W. A variational model for multiplicative structured noise removal[J]. Journal of M athematical Imaging&Vision,2017,57(1):1-13.
[6] Duan Lu,Song Yong-hong,Zhang Yuan-lin. Layout analysis algorithm of questionnaire image[J]. Journal of Softw are,2017,28(2):234-245.
[7] Shrawankar U,Kaur A. Adverse conditions and techniques for cross lingual text recognition[C]. International Conference on Innovative M echanisms for Industry Applications,IEEE,2017.
[8] Lee C Y,Osindero S. Recursive recurrent nets with attention modeling for OCR in the w ild[C]. Conference on Computer Vision&Pattern Recognition,IEEE,2016:2231-2239.
[9] Zhao Chun-yang,Zhao Huai-ci. Parallel optimized method and hardw are implementation of SURF algorithm[J]. Journal of Computer-Aided Design&Computer Graphic,2015,27(2):256-263.
[10] Li Bing,Liu Lei,Wei Zhi-qiang. A strong robust real-time image matching algorithm[J]. Journal of Software,2014,25(7):1583-1592.
[11] Liu Wei,Duan Cheng-wei,Yu Bing,et al. Multi-pose pedestrian detection based on posterior HOG feature[J]. Acta Electronica Sinica,2015,43(2):217-224.
[12]mre E,Hilton A. Order statistics of RANSAC and their practical application[J]. International Journal of Computer Vision,2015,111(3):276-297.
[13] Lowanshi V K,Shrivastava S,Richhariya V,et al. An efficient approach for content based image retrieval using SVM,KNN-GA as multilayer classifier[J]. International Journal of Computer Applications,2014,107(21):43-48.
[14] Su B,Lu S,Tan C L. Robust document image binarization technique for degraded document images[J]. IEEE Transactions on Image Processing,2013,22(4):1408-1417.
[15] Yi Jian,Peng Yu-xin,Xiao Jian-guo. Recognition of text in video based on color clustering and multiple frame integration[J]. Journal of Softw are,2011,22(12):2919-2933.
[16] Zemouri E. Enhancement of historical document images by combining global and local binarization technique[J]. International Journal of Information&Electronics Engineering,2014,4(1):1-5.
[1]吴一全,张金矿.基于投影坐标p次方差及粒子群的车牌倾斜检测[J].计算机辅助设计与图形学学报,2010,22(1):114-120.
[2]苗立刚.基于最近邻链的车牌检测算法[J].自动化学报,2011,37(10):1272-1279.
[3]田破荒,彭天强,李弼程.基于文字穿越线和笔画连通性的视频文字提取方法[J].电子学报,2009,37(1):72-78.
[6]段露,宋永红,张元林.一种面向问卷图像的版面分析算法[J].软件学报,2017,28(2):234-245.
[9]赵春阳,赵怀慈. SURF算法并行优化及硬件实现[J].计算机辅助设计与图形学学报,2015,27(2):256-263.
[10]李兵,刘磊,魏志强.一种具有强实时性、强鲁棒性的图像匹配算法[J].软件学报,2014,25(7):1583-1592.
[11]刘威,段成伟,遇冰,等.基于后验HOG特征的多姿态行人检测[J].电子学报,2015,43(2):217-224.
[15]易剑,彭宇新,肖建国.基于颜色聚类和多帧融合的视频文字识别方法[J].软件学报,2011,22(12):2919-2933.
[2] Miao Li-gang. License plate detection algorithm based on nearest neighbor chains[J]. Acta Automatica Sinica,2011,37(10):1272-1279.
[3] Tian Po-huang,Peng Tian-qiang,Li Bi-cheng. An approach for video-text extraction based on text traversing line and stroke connectivity[J]. Acta Electronica Sinica,2009,37(1):72-78.
[4] Milyaev S,Barinova O,Novikova T,et al. Image binarization for end-to-end text understanding in natural images[C]. International Conference on Document Analysis and Recognition,IEEE,2013,37(5):128-132.
[5] Escande P,Weiss P,Zhang W. A variational model for multiplicative structured noise removal[J]. Journal of M athematical Imaging&Vision,2017,57(1):1-13.
[6] Duan Lu,Song Yong-hong,Zhang Yuan-lin. Layout analysis algorithm of questionnaire image[J]. Journal of Softw are,2017,28(2):234-245.
[7] Shrawankar U,Kaur A. Adverse conditions and techniques for cross lingual text recognition[C]. International Conference on Innovative M echanisms for Industry Applications,IEEE,2017.
[8] Lee C Y,Osindero S. Recursive recurrent nets with attention modeling for OCR in the w ild[C]. Conference on Computer Vision&Pattern Recognition,IEEE,2016:2231-2239.
[9] Zhao Chun-yang,Zhao Huai-ci. Parallel optimized method and hardw are implementation of SURF algorithm[J]. Journal of Computer-Aided Design&Computer Graphic,2015,27(2):256-263.
[10] Li Bing,Liu Lei,Wei Zhi-qiang. A strong robust real-time image matching algorithm[J]. Journal of Software,2014,25(7):1583-1592.
[11] Liu Wei,Duan Cheng-wei,Yu Bing,et al. Multi-pose pedestrian detection based on posterior HOG feature[J]. Acta Electronica Sinica,2015,43(2):217-224.
[12]mre E,Hilton A. Order statistics of RANSAC and their practical application[J]. International Journal of Computer Vision,2015,111(3):276-297.
[13] Lowanshi V K,Shrivastava S,Richhariya V,et al. An efficient approach for content based image retrieval using SVM,KNN-GA as multilayer classifier[J]. International Journal of Computer Applications,2014,107(21):43-48.
[14] Su B,Lu S,Tan C L. Robust document image binarization technique for degraded document images[J]. IEEE Transactions on Image Processing,2013,22(4):1408-1417.
[15] Yi Jian,Peng Yu-xin,Xiao Jian-guo. Recognition of text in video based on color clustering and multiple frame integration[J]. Journal of Softw are,2011,22(12):2919-2933.
[16] Zemouri E. Enhancement of historical document images by combining global and local binarization technique[J]. International Journal of Information&Electronics Engineering,2014,4(1):1-5.
[1]吴一全,张金矿.基于投影坐标p次方差及粒子群的车牌倾斜检测[J].计算机辅助设计与图形学学报,2010,22(1):114-120.
[2]苗立刚.基于最近邻链的车牌检测算法[J].自动化学报,2011,37(10):1272-1279.
[3]田破荒,彭天强,李弼程.基于文字穿越线和笔画连通性的视频文字提取方法[J].电子学报,2009,37(1):72-78.
[6]段露,宋永红,张元林.一种面向问卷图像的版面分析算法[J].软件学报,2017,28(2):234-245.
[9]赵春阳,赵怀慈. SURF算法并行优化及硬件实现[J].计算机辅助设计与图形学学报,2015,27(2):256-263.
[10]李兵,刘磊,魏志强.一种具有强实时性、强鲁棒性的图像匹配算法[J].软件学报,2014,25(7):1583-1592.
[11]刘威,段成伟,遇冰,等.基于后验HOG特征的多姿态行人检测[J].电子学报,2015,43(2):217-224.
[15]易剑,彭宇新,肖建国.基于颜色聚类和多帧融合的视频文字识别方法[J].软件学报,2011,22(12):2919-2933.