蓝智敏, 梁礼明, 盛校棋,等.基于多特征融合的铁路护栏网格提取[J].河南理工大学学报（自然科学版）,2021,40(3):134-141.

LAN Z M, LIANG L M, SHENG X Q, et al.Grid extraction of railway guardrail based on multi-feature fusion[J].Journal of Henan Polytechnic University(Natural Science) ,2021,40(3):134-141.

基于多特征融合的铁路护栏网格提取

蓝智敏, 梁礼明, 盛校棋, 黎富泉, 吴健

江西理工大学 电气工程与自动化学院，江西 赣州341000

摘要:铁路护栏是列车高速安全运行免受异物入侵的必要保障，其图像精确提取是铁路基础设施自动化检测的关键步骤。针对铁路护栏因背景干扰因素多而难以准确提取其图像的问题，提出一种融合多特征和二维最大熵的铁路护栏网格图像提取算法。首先采用双边滤波降噪和伽马变换增强铁路护栏网格图像，其次在提取护栏网格的线性特征、方差特征和矩特征后进行二维最大熵初分割，最后根据三类特征图的初分割进行加权融合与连通域去噪。结果表明，本文算法对铁路护栏网格图像的提取具有较强的鲁棒性。

关键词:铁路护栏;网格提取;特征提取;二维最大熵;连通域去噪

doi:10.16186/j.cnki.1673-9787.2019110004

基金项目:国家自然科学基金资助项目（51365017 ，61463018 ）；江西省自然科学基金资助项目（20192BAB205084 ）；江西省科学技术 研究重点项目（GJJ170491 ）；国家级大学生创新创业训练项目（201810407017）

收稿日期:2019/11/01

修回日期:2020/05/21

出版日期:2021/05/15

Grid extraction of railway guardrail based on multi-feature fusion

LAN Zhimin, LIANG Liming, SHENG Xiaoqi, LI Fuquan, WU jian

School of Electrical Engineering and Automation， Jiangxi University of Science and Technology， Ganzhou 341000 ， Jiangxi， China

Abstract:Railway guardrail is the necessary guarantee for the safe operation of trains at high speed from foreign body intrusion， the accurate extraction of its image is the key step of railway infrastructure automatic detection. Aiming at the problem that railway guardrail image is difficult to accurately extract because of background interference factor， a grid extraction algorithm of railway guardrail fusing multiple features and two-dimensional maximum entropy was proposed. Firstly， the railway guardrail image was denoised by bilateral filtering and enhanced by gamma transform. Then， the linear feature， variance feature and moment feature of the guardrail were extracted separately. The initial segmentation of the feature maps were obtained respectively by employing two-dimensional maximum entropy. Finally， the weighted fusion and connected domain denoising were applied according to the initial segmentation results. The experimental results indicated that the proposed algorithm was robust to the extraction of railway guardrail grid.

Key words:railway guardrail;grid extraction;feature extraction;two-dimensional maximum entropy;connected domain denoising

 基于多特征融合的铁路护栏网格提取_蓝智敏.pdf