扫描图像中时序曲线的识别及数据提取.pdfVIP

计算机研究与发展 Journal of Computer Research and Development ISSN 1000-1239/ CN 11-1777/ 丁P 44(Suppl.): 251-256. 2007 扫描图像中时序曲线的识别及数据提取 间光伟 胡海涛 (华北电力大学计算机科学与技术系 北京 102206) (guangweLyan@263.net) A Recognition and Data Extraction Algorithm for Scanned Sequential Curve Images Yan Guangwei and Hu Haitao (De阳rtment of Co阴阳ter Science and Technology , North China Electric Power University , Beijing 102206) Abstract An algorithm for curve recognition and data extraction in scanned images is presented. In a scanned image. there are several pen-drawn sequential curves under the background of grid , the a?m ?s to get the sampling data of the curves. At first , denoising is implemented. The background noise is cleared by the line-by-line self-adapted threshold algorithm and the blocked self-adapted threshold algorithm. A selectivity eroding algorithm is presented to clear the remaining vertical and horizontal grid noise. A template matching algorithm is presented to clear the local lonely noise. Then the curves are thinned and extra points are c1eared and only single幡pixel幡width skeleton is left. Then a 8-neighborhood tracing algorithm is used to trace and sample the skeleton curv邸. To recognize and sample the curve peaks accurately , a curve contour recognition algorithm is presented. Dashed is recognized by means of clustering. Experiments show that the curve data can be extracted effectively. The algorithm can be used in the digitization of various sequential curve drawings. Key words pen-draw curvej gridj thinníngj denoisingj curve tracingj data extraction; digitization 摘要研究扫描阁像中时序曲线的识别与数据提取.数据源为扫描后栅格背景下的多条笔绘曲线圈 像，目的是提取每条笔给曲线的时序数据.提出了该类扫描朗的处玻框架，并改进了多种算法.采用远 行自适应阅值及分块白适应阅值算法进行背景噪声的粗滤除;采用选择性腐蚀算法进行垂直和水乎栅 格噪声的进一步滤除;采用模板匹配技术完成孤立民域噪声的清除;对去嗓后的笔给曲线进行单像素细 化，并采用 8 邻域技术进行曲线跟踪;对曲线波峰区域采用轮廓提取法进行识别;采用虚线来类的方式 识别虚曲线.实验结果表明，该方法较好地完成了对时序曲线的识别及数据提取，可应用于各类具有栅 格背景的时序曲线朋纸的数字化. 关键词笔绘曲线;栅格;细化;噪声滤除;曲线跟踪;数据提取;数字化 中图法分类号 TP391 许多笔给(子给或机给)的时序曲线图纸需要数 字化.如石袖录井曲线图纸和 ECG 图纸等.这些图 纸往往是多年前由手工或机