基于残差网络的字符编码标志点识别  

Recognition of Character Coded Marks Based on Residual Network

作　　者：菲劳拉·吐尔洪江 古丽巴哈尔·托乎提[1] 买买提明·艾尼[1,2] 补生来 吴志强 FILORA Turhun;GULBAHAR Tohti;MAMTIMIN Geni;BU Shenglai;WU Zhiqiang(College of Mechanical Engineering,Xinjiang University,Urumqi 830017,China;Baibo Electromechanical Technology Co.,Ltd.,Urumqi 830011,China)

机构地区：[1]新疆大学机械工程学院,乌鲁木齐830017 [2]佰博机电科技有限公司,乌鲁木齐830011

出　　处：《组合机床与自动化加工技术》2024年第8期80-84,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金资助项目(12162031)。

摘　　要：针对大尺寸结构近景摄影测量中如何提高编码标志的容量和识别率等问题,设计了一种由中心圆,字母和数字组成的字符编码标志,提出了一种基于ResNet18模型改进的字符编码点识别方法和基于Zernike矩提出的亚像素定位方法。首先,利用均值滤波、自适应阈值处理和形态学中的闭运算等方法对图像进行预处理,根据连通域分析与筛选对预处理好的图像进行分割得到字符编码点;其次,用字符编码点识别网络进行识别,最终通过亚像素定位方法对编码点进行定位。从实验结果得出字符编码点的识别准确率可达99.89%。该类标志点定位精度高,提出的识别方法准确率高。Aiming at the problem of how to improve the capacity and recognition rate of coded mark in close-range photogrammetry of large structure,this paper designs a character coded marks composed of center circle,letters and numbers,and proposes a character coded point recognition method based on ResNet18 model and a sub-pixel location method based on Zernike moment.Firstly,the image is preprocessed by means filtering,adaptive threshold processing and closing operation in morphology,and then the preprocessed image is segmented according to connected domain analysis and screening to get character coded marks.Secondly,the character coded mark recognition network is used for recognition,and finally the coding points are located by subpixel positioning method.The experimental results show that the recognition accuracy of character coding points can reach 99.89%.This kind of marker points have high positioning accuracy and the proposed recognition method has high accuracy.

关 键 词：近景摄影测量 字符编码标志 字符编码点识别网络 亚像素定位 

分 类 号：TH161[机械工程—机械制造及自动化] TG66[金属学及工艺—金属切削加工及机床]