摘要
针对机器人自动整理图书研究中的关键问题——利用机器视觉识别书脊问题展开研究。文章提出一种基于轮廓的书脊识别方法。利用Canny算子检测书脊边缘,根据书脊两侧轮廓为平行的线段特点,利用概率霍夫变换提取书脊两侧的轮廓,通过线段位置关系得到书脊ROI区域;根据书脊上文字与书脊颜色,亮度相差较大的特点,利用K-means对色调和亮度进行聚类,并通过判断形态特征得到文字区域,通过文字排列方式判断是否为书脊。将文章方法应用于实际检测中,并与利用LSD识别书脊的方法进行对比,实验结果表明文章方法识别正确率更高。
Research on the key problem in the study of automatic book sorting by robots-the use of machine vision to identify the spine of the book.This paper proposes a method of spine recognition based on outlineand uses Canny operator to detect the edge of the spine.According to the characteristics of parallel line segments on both sides of the spine,the Probabilistic Hough Transform is used to extract the contours on both sides of the spine.The ROI area of the spine is obtained through the position relationship of the line segments;according to the text on the spine and the color of the spine,the brightness is different.The larger feature uses K-means to cluster the hue and brightness,and obtains the text area by judging the morphological characteristics,thus judging whether it is the spine by the text arrangement.The method in this paper is applied to actual detection and compared with the method of using LSD to identify the spine.The experimental results show that the method in this paper has a higher recognition accuracy.
作者
王钰深
黄悦恒
黄谦
刘冲
WANG Yushen;HUANG Yueheng;HUANG Qian;LIU Chong
出处
《科技创新与应用》
2020年第36期7-11,共5页
Technology Innovation and Application
基金
国家自然科学基金(编号:51805074)
大学生创新创业训练计划项目(编号:201910145145)。
关键词
机器视觉
书脊识别
概率霍夫变换
CANNY算子
machine vision
spine recognition
Probabilistic Hough Transform
Canny operator
