基于多尺度多分辨率的直线段检测算法  

LSD algorithm based on multi-scale multi-resolution

作　　者：唐守锋 范梦松 赵仁慈 翟少奇 TANG Shoufeng;FAN Mengsong;ZHAO Renci;ZHAI Shaoqi(School of Information and Control Engineering,China University of Mining and Technology,Xuzhou 221116,China)

机构地区：[1]中国矿业大学信息与控制工程学院,江苏徐州221116

出　　处：《传感器与微系统》2023年第12期134-137,共4页Transducer and Microsystem Technologies

摘　　要：在点线特征结合的同步定位与地图构建(SLAM)系统中,针对直线段检测(LSD)算法易受光照、角度等因素的影响且造成提取直线段过度分割等问题,提出一种基于多尺度、多分辨率的LSD(MMLSD)算法。首先,在对图像处理前,引入图像分辨率阈值判断图像的初步分类;其次,对预处理后的图像进行多尺度多分辨率处理,形成多尺度图像空间或者图像金字塔;然后,在线段合并阶段,对金字塔各层提取的直线段进行由粗至精的线段替代合并;最终,输出提取线段结果。实验结果表明:改进后的算法有效地克服了线段的过度分割现象,同时对低对比度的环境具有良好的鲁棒性。相比LSD算法,所提算法直线检测的平均精度提升13.1%。In simultaneous localization and mapping(SLAM)system combining point-line features,aiming at the problem that line segment detection(LSD)algorithm is easily affected by illumination,angle and other factors and causes excessive segmentation of extracted line segments,a multi-scale and multi-resolution LSD(MMLSD)algorithm is proposed.Firstly,before image processing,the image resolution threshold is introduced to judge the initial classification of the image.Secondly,the pre-processed image is processed by multi-scale and multi-resolution to form multi-scale image space or image pyramid.Then,in the line segment merging stage,the line segment extracted from each layer of the pyramid is replaced by the line segment from coarse to fine.Finally,the extracted line segment result is output.Experimental results show that the improved algorithm effectively overcome the phenomenon of over-segmentation of line segments,and has good robustness to low-contrast environment.Compared with LSD algorithm,the average precision of line detection of the proposed algorithm is improved by 13.1%.

关 键 词：多尺度 多分辨率 直线段检测算法 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]