摘要
视觉导航算法的设计是高压线路巡检机器人领域的一项关键技术,为了解决当前目标检测算法准确率低、抗干扰能力弱、小目标检测效果差等问题,提出一种最近邻加权灰度均值算法,用于提取目标的感兴趣区域(ROI)。采用相线直径匹配的方式对ROI边框进行精细调整,依据特征提取与机器学习理论实现ROI分类。实验结果表明,该算法具有很高的ROI提取精度,分类器对于正样本的召回率最高可达99.83%,可有效地指导高压巡检机器人进行线路巡检工作。
The designing of the visual navigation algorithm is a key technology in the field of high-voltage inspection robot.In order to solve the remaining problems such as the low detection accuracy,the weak ability of interference resistance and the poor effect for small object detection,a nearest weighted gray mean algorithm is proposed,which can be used to extract the object ROI(region of interest).We adopted a new method named phase diameter matching to fine tune the ROI boundary,and we completed classifying the extracted ROI according to feature extraction and machine learning.The experimental results show that this algorithm has a very high ROI extraction precision and the recall rate for positive samples can reach 99.83%,which can be applied effectively on guiding the inspection robot to work better.
作者
唐法帅
高琦
杜宗展
Tang Fashuai;Gao Qi;Du Zongzhan(School of Mechanical Engineering,Shandong University,Jinan 250061,Shandong,China;Key Laboratory of High Efficiency and Clean Mechanical Manufacture,Ministry of Education,Jinan 250061,Shandong,China;School of Electrical Engineering,Shandong University,Jinan 250061,Shandong,China)
出处
《计算机应用与软件》
北大核心
2021年第10期201-208,296,共9页
Computer Applications and Software
基金
深圳市科技计划项目(JCYJ20170818103220315)。
