基于改进滴水算法的芯片图像粘连字符分割  

TOUCHING CHARACTER SEGMENTATION FOR CHIP IMAGE BASED ON IMPROVED DROP-FALL ALGORITHM

作　　者：陈慧敏 张合生[1] 朱晓锦[1] Chen Huimin;Zhang Hesheng;Zhu Xiaojin(School of Mechatronic Engineering and Automation,Shanghai University,Shanghai 200444,China)

机构地区：[1]上海大学机电工程与自动化学院,上海200444

出　　处：《计算机应用与软件》2023年第2期246-252,301,共8页Computer Applications and Software

摘　　要：为克服粘连字符对芯片字符识别的影响,提出一种基于改进滴水算法的粘连字符分割方法。在芯片图像进行必要的预处理后,提取单张芯片的标准字符信息,利用水平投影和连通域分析提取粘连区域。通过垂直投影和上下轮廓得到上下轮廓极值点;分别以粘连区域左右两端为起点,利用标准字符信息对极值点进行筛选,选取方差较小的一组作为候选分割点;通过分割点对应关系补齐遗漏的有效分割点;上下有效分割点分别作为水滴起始点和终止点,定义新的滴落路径对粘连字符进行分割。实验结果表明,该算法可有效解决芯片图像的字符粘连问题,并且在分割精准度上拥有良好的表现。In order to overcome the influence of touching character on chip character recognition, an improved drop-fall algorithm is proposed to segment the touching character. After the chip image was preprocessed effectively, standard character information of a single chip was extracted, and the touching area was extracted by horizontal projection and analysis of connected domain. Upper and lower contours extreme points were obtained by vertical projection, upper and lower contours. Using the left and right ends of the touching area as the starting point, the extreme points were filtered using standard character information, and a group with small variance was selected as the candidate segmentation points. Missing valid segmentation points were completed by their corresponding relationships. The upper and lower segmentation points were taken as the starting and ending points respectively, and a new drop path was defined to segment touching characters. The experimental result shows that the algorithm is enable to segment touching characters in chip images. In addition, the improved algorithm shows good performance in segmentation accuracy.

关 键 词：字符分割 滴水算法 粘连字符 芯片图像 垂直投影 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]