基于结构特征与LLE算法的数字仪表校验系统
|
摘要
针对数字仪表校验人工读取数据时存在的误差问题,提出一种基于多特征与局部线性嵌入算法(LLE)结合的数字仪表校验方法。首先将字符图像从RGB转换到HSI色彩空间,在I空间对图像进行分割,并利用投影法分割出单个字符;然后对字符图像提取结构特征,由于单纯的结构特征对字符的描述不足,导致识别精度较低,故文中利用LLE算法对字符二值图像降维,将降维后的像素特征与结构特征相结合;最后,利用支持向量机(SVM)对字符特征进行识别。实验结果显示,所提方法提高了校验系统的字符识别率。
Since there exist errors in manual reading data during digital meter calibration,a digital meter calibration method based on the combination of multi-features and local linear embedding(LLE)algorithm is proposed. The character images are converted from RGB to HSI color space. Image segmentation is conducted in the I space. The single character is segmented by using the projection method. The structural features of character images are extracted. As the simple structural features cannot sufficiently describe the characters,the recognition accuracy is low. Therefore,the LLE algorithm is used in this paper to reduce the dimensionality of character binary images. The pixel features after dimensionality reduction and structural features are combined. The support vector machine(SVM)is used to recognize the character features. The experimental results show that the proposed method can improve the character recognition rate of the calibration system.
Since there exist errors in manual reading data during digital meter calibration,a digital meter calibration method based on the combination of multi-features and local linear embedding(LLE)algorithm is proposed. The character images are converted from RGB to HSI color space. Image segmentation is conducted in the I space. The single character is segmented by using the projection method. The structural features of character images are extracted. As the simple structural features cannot sufficiently describe the characters,the recognition accuracy is low. Therefore,the LLE algorithm is used in this paper to reduce the dimensionality of character binary images. The pixel features after dimensionality reduction and structural features are combined. The support vector machine(SVM)is used to recognize the character features. The experimental results show that the proposed method can improve the character recognition rate of the calibration system.
引文
[1]臧景茹,杨晓琰,刘晶.数字仪表自动校验系统开发[J].华北电力技术,2006(2):34-37.ZANG Jingru,YANG Xiaoyan,LIU Jing. Development of automatic calibration system for digital meter[J]. North China electric power,2006(2):34-37.
[2]郭爽.数码管数字仪表自动识别方法的研究[J].通信技术,2012,45(8):91-93.GUO Shuang. Study on automatic identification method of digital tube[J]. Communications technology,2012,45(8):91-93.
[3]张海波,段会川,张曙光,等.一种数字仪表显示值快速识别方法[J].计算机工程与应用,2005,41(4):223-226.ZHANG Haibo,DUAN Huichuan,ZHANG Shuguang,et al.A rapid recognition approach for the display value of digital meters[J]. Computer engineering and applications,2005,41(4):223-226.
[4]刘丹,穆林丽,余晓锷.一种七段数码管式医用仪表的自动识别方法[J].科学技术与工程,2010,10(16):4037-4039.LIU Dan,MU Linli,YU Xiao’e. A new method of medical digital meter′s computer automatic reading[J]. Science technology and engineering,2010,10(16):4037-4039.
[5]洪涛,梁伟建,卢玉凤.标牌粘连字符自适应定位分割重建与识别[J].中国图象图形学报,2014,19(6):886-895.HONG Tao,LIANG Weijian,LU Yufeng. Merged characters segmentation reconstruction and recognition of labels based on adaptive location[J]. Journal of image and graphics,2014,19(6):886-895.
[6]申小阳,唐轶峻,姜柏军,等.仪表的数字字符识别系统[J].仪表技术与传感器,2005(6):55-57.SHEN Xiaoyang,TANG Yijun,JIANG Bojun,et al. Numeral characters recognition system of instrument[J]. Instrument technique and sensor,2005(6):55-57.
[7]刘自立,周洪,余力,等.基于投影矩阵法的仪表字符实时识别系统[J].仪表技术与传感器,2008(8):98-100.LIU Zili,ZHOU Hong,YU Li,et al. Real-time instrument character recognition system based on projective matrix method[J]. Instrument technique and sensor,2008(8):98-100.
[8] GUNASEKARAN M,GANESHMOORTHY S. OCR recognition system using feed forward and back propagation neural network[J]. SSRN electronic journal,2013(10):52-60.
[9]李智成,李文婷,梅术正,等.基于机器视觉的高压计量数显表自动校对系统电测与仪表,,():LI Zhicheng,LI Wenting,MEI Shuzheng,et al. Automatic calibration system of high voltage measuring digital display instrument based on machine vision[J]. Electrical measurement&instrumentation,2017,54(20):105-109.
[10] KO D G,SONG S H,KANG K M,et al. Convolutional neural networks for character-level classification[J]. IEEE transactions on smart processing&computing,2017,6(1):53-59.
[11]杨芳,李红睿,田学东.基于RBF神经网络的汉字粗分类方法[J].计算机工程与应用,2009,45(6):170-172.YANG Fang,LI Hongrui,TIAN Xuedong. Coarse classification scheme for Chinese character based on RBF neural network[J]. Computer engineering and applications,2009,45(6):170-172.
[12]张好好,吴游,于睿,等.一种基于压缩感知的数字图像水印算法[J].现代电子技术,2014,37(22):10-13.ZHANG Haohao,WU You,YU Rui,et al. A digital image watermark algorithm based on compressive perception[J].Modern electronics technique,2014,37(22):10-13.
[13]丁娇,梁栋,阎庆.基于D-LLE算法的多特征植物叶片图像识别方法[J].计算机工程与应用,2015,51(9):158-163.DING Jiao,LIANG Dong,YAN Qing. Recognition method of multi-feature plant leaves based on D-LLE algorithm[J]. Computer engineering and applications,2015,51(9):158-163.
[2]郭爽.数码管数字仪表自动识别方法的研究[J].通信技术,2012,45(8):91-93.GUO Shuang. Study on automatic identification method of digital tube[J]. Communications technology,2012,45(8):91-93.
[3]张海波,段会川,张曙光,等.一种数字仪表显示值快速识别方法[J].计算机工程与应用,2005,41(4):223-226.ZHANG Haibo,DUAN Huichuan,ZHANG Shuguang,et al.A rapid recognition approach for the display value of digital meters[J]. Computer engineering and applications,2005,41(4):223-226.
[4]刘丹,穆林丽,余晓锷.一种七段数码管式医用仪表的自动识别方法[J].科学技术与工程,2010,10(16):4037-4039.LIU Dan,MU Linli,YU Xiao’e. A new method of medical digital meter′s computer automatic reading[J]. Science technology and engineering,2010,10(16):4037-4039.
[5]洪涛,梁伟建,卢玉凤.标牌粘连字符自适应定位分割重建与识别[J].中国图象图形学报,2014,19(6):886-895.HONG Tao,LIANG Weijian,LU Yufeng. Merged characters segmentation reconstruction and recognition of labels based on adaptive location[J]. Journal of image and graphics,2014,19(6):886-895.
[6]申小阳,唐轶峻,姜柏军,等.仪表的数字字符识别系统[J].仪表技术与传感器,2005(6):55-57.SHEN Xiaoyang,TANG Yijun,JIANG Bojun,et al. Numeral characters recognition system of instrument[J]. Instrument technique and sensor,2005(6):55-57.
[7]刘自立,周洪,余力,等.基于投影矩阵法的仪表字符实时识别系统[J].仪表技术与传感器,2008(8):98-100.LIU Zili,ZHOU Hong,YU Li,et al. Real-time instrument character recognition system based on projective matrix method[J]. Instrument technique and sensor,2008(8):98-100.
[8] GUNASEKARAN M,GANESHMOORTHY S. OCR recognition system using feed forward and back propagation neural network[J]. SSRN electronic journal,2013(10):52-60.
[9]李智成,李文婷,梅术正,等.基于机器视觉的高压计量数显表自动校对系统电测与仪表,,():LI Zhicheng,LI Wenting,MEI Shuzheng,et al. Automatic calibration system of high voltage measuring digital display instrument based on machine vision[J]. Electrical measurement&instrumentation,2017,54(20):105-109.
[10] KO D G,SONG S H,KANG K M,et al. Convolutional neural networks for character-level classification[J]. IEEE transactions on smart processing&computing,2017,6(1):53-59.
[11]杨芳,李红睿,田学东.基于RBF神经网络的汉字粗分类方法[J].计算机工程与应用,2009,45(6):170-172.YANG Fang,LI Hongrui,TIAN Xuedong. Coarse classification scheme for Chinese character based on RBF neural network[J]. Computer engineering and applications,2009,45(6):170-172.
[12]张好好,吴游,于睿,等.一种基于压缩感知的数字图像水印算法[J].现代电子技术,2014,37(22):10-13.ZHANG Haohao,WU You,YU Rui,et al. A digital image watermark algorithm based on compressive perception[J].Modern electronics technique,2014,37(22):10-13.
[13]丁娇,梁栋,阎庆.基于D-LLE算法的多特征植物叶片图像识别方法[J].计算机工程与应用,2015,51(9):158-163.DING Jiao,LIANG Dong,YAN Qing. Recognition method of multi-feature plant leaves based on D-LLE algorithm[J]. Computer engineering and applications,2015,51(9):158-163.