[1]聂子博,曹建军,翁年凤,等.基于直线生成的卷积霍夫线段检测[J].计算机技术与发展,2024,34(05):30-36.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0037]
　NIE Zi-bo,CAO Jian-jun,WENG Nian-feng,et al.Convolutional Hough Line Segment Detection Based on Line Generation[J].,2024,34(05):30-36.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0037]
点击复制

基于直线生成的卷积霍夫线段检测()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
34
期数:
2024年05期
页码:
30-36
栏目:
媒体计算
出版日期:
2024-05-10

文章信息/Info

Title:
Convolutional Hough Line Segment Detection Based on Line Generation
文章编号:
1673-629X(2024)05-0030-07
作者:
聂子博1; 曹建军1; 翁年凤1; 余旭1; 王孟大1; 2
1.国防科技大学 第六十三研究所 数据工程研究中心,江苏 南京 210007;2.南京信息工程大学 计算机学院、软件学院、网络空间安全学院,江苏 南京 210044
Author(s):
NIE Zi-bo1; CAO Jian-jun1; WENG Nian-feng1; YU Xu1; WANG Meng-da1; 2
1.The Sixty-third Research Institute,National University of Defense Technology,Nanjing 210007,China;2.School of Computer Science,Nanjing University of Information Science and Technology,Nanjing 210044,China
关键词:
线段检测; 霍夫变换; 卷积核; 异或; 直线生成; 布雷森汉姆算法
Keywords:
line segment detection; Hough transform; convolution kernel; exclusive or; line generation; Bresenham algorithm
分类号:
TP391
DOI:
10.20165/j.cnki.ISSN1673-629X.2024.0037
摘要:
直线检测作为计算机视觉的上游任务,为下游包括工业视觉、遥感图像分析等任务提供支撑。直线检测的一大方向是霍夫直线检测,但现有霍夫检测基于近似原理设计计票器,直线检测准确度不高。为提高霍夫变换线段检测的准确度,利用卷积改进霍夫直线检测的计票器并提出基于直线生成的卷积霍夫线段检测方法。利用中值滤波对原始图像中的复杂纹理平滑处理后检测图像中的边界;通过按位异或卷积去除边界检测结果中的噪点并保留候选的线段端点;将候选的线段端点两两组合并使用布雷森汉姆算法进行线段生成,由利用卷积改进的计票器判断生成的线段是否存在于边界上;确认端点所构成线段位于边界后求取端点对之间的参数并合并参数相似的加检测结果,得到最终线段检测结果。对比实验中该方法的F1指标为0.7626,优于对比方法中最高的0.6523,证明该方法保留了霍夫变换较高鲁棒性的同时提高了检测结果的准确性。
Abstract:
As an upstream task in computer vision,line detection provides support for downstream tasks including industrial vision and remote sensing image analysis.One major direction in line detection is Hough line detection.However,the accumulators in existing Hough detection methods are generally designed based on approximate principles,leading to lower accuracy in line detection.To enhance the precision of Hough line detection,convolution is employed to improve the accumulator in Hough line detection and Convolutional Hough Line Segment Detection based on line generation (CHLSD) is proposed.Firstly,median filtering is utilized to smooth the complex textures in original images,and extract edges from smoothed images.Next,the extraction results are denoised using a bitwise Exclusive OR operation and candidate end points of line segments are retained.Then,Bresenham algorithm generates line segments with paired candidate end points,accumulator improved with convolution is used to determine if generated line segments match extracted edges.Finally,parameters for determined end points are calculated and lines with similar parameters are merged to obtain final detection result.Experiments show that CHLSD improves the detection precision while retaining the robustness of Hough line detection,achieving an F1 score of 0.7626,which is superior to the comparison methods’ score of 0.6523.

相似文献/References:

[1]郑权,刘循,魏海明.基于学习的霍夫变换线段组物体检测算法[J].计算机技术与发展,2014,24(01):26.
　ZHENG Quan,LIU Xun,WEI Hai-ming.An Object Detection Algorithm of Hough Transform Line Segmentation Groups Based on Learning[J].,2014,24(05):26.
[2]陈银平,涂碧海,张毅,等.改进型LOG算子在偏振片绝对角度中的应用[J].计算机技术与发展,2014,24(05):164.
　CHEN Yin-ping[],TU Bi-hai[],ZHANG Yi[],et al.Application of Improved LOG Operator in Absolute Angle of Polarizer[J].,2014,24(05):164.
[3]邵虹,王佳. 发票印刷体数字识别方法的研究[J].计算机技术与发展,2016,26(07):173.
　SHAO Hong,WANG Jia. Research on Recognition Method on Invoice Printing Number[J].,2016,26(05):173.
[4]穆珺[],晏峻峰[],彭清华[],等. 基于霍夫变换的可见光虹膜图像定位[J].计算机技术与发展,2017,27(05):40.
　MU Jun[],YAN Jun-feng[],PENG Qing-hua[],et al. Iris Localization for Visible-light Images Based on Hough Transform[J].,2017,27(05):40.
[5]顾绍通. 直线提取的预处理研究[J].计算机技术与发展,2017,27(09):40.
　GU Shao-tong. Investigation on Preprocessing of Line Extraction[J].,2017,27(05):40.
[6]王 颖,王鹏飞,杨余旺.基于图像分割的目标尺寸特征测量[J].计算机技术与发展,2018,28(02):191.[doi:10．3969/j．issn．1673－629X．2018．02．041]
　WANG Ying,WANG Pengfei,YANG Yuwang.Object Dimension Feature Measurement Based on Image Segmentation[J].,2018,28(05):191.[doi:10．3969/j．issn．1673－629X．2018．02．041]
[7]徐鹏,曾卫明,石玉虎,等.一种指针式油位计读数识别算法[J].计算机技术与发展,2018,28(09):189.[doi:10.3969/ j. issn.1673-629X.2018.09.039]
　XU Peng,ZENG Wei-ming,SHI Yu-hu,et al.A Reading Recognition Algorithm of Pointer Type Oil-level Meter[J].,2018,28(05):189.[doi:10.3969/ j. issn.1673-629X.2018.09.039]
[8]操昊鹏,曾卫明,石玉虎,等.基于 Hough 变换和总体最小二乘法的电力线检测[J].计算机技术与发展,2018,28(10):164.[doi:10.3969/ j. issn.1673-629X.2018.10.034]
　CAO Hao-peng,ZENG Wei-ming,SHI Yu-hu,et al.Power Line Detection Based on Hough Transform and Total Least Squares Method[J].,2018,28(05):164.[doi:10.3969/ j. issn.1673-629X.2018.10.034]
[9]汪 ?涛,成孝刚,李德志,等.基于霍夫变换与角点检测的叶脉特征提取算法[J].计算机技术与发展,2019,29(11):159.[doi:10. 3969 / j. issn. 1673-629X. 2019. 11. 032]
　WANG Tao,CHENG Xiao-gang,LI De-zhi,et al.A Feature Extraction Algorithm for Leaf Vein Based on Hough Transform and Corner Detection[J].,2019,29(05):159.[doi:10. 3969 / j. issn. 1673-629X. 2019. 11. 032]
[10]朱鸿宇,杨 帆,高晓倩,等.基于级联霍夫变换的车道线快速检测算法[J].计算机技术与发展,2021,31(01):88.[doi:10. 3969 / j. issn. 1673-629X. 2021. 01. 016]
　ZHU Hong-yu,YANG Fan,GAO Xiao-qian,et al.A Fast Lane Detection Algorithm Based on Cascade Hough Transform[J].,2021,31(05):88.[doi:10. 3969 / j. issn. 1673-629X. 2021. 01. 016]

更新日期/Last Update: 2024-05-10