摘要
配电线路所处环境复杂,传统配电线路无人机巡视目标检测方法无法有效去除外界环境干扰,导致检测结果不可靠。为此,提出一种新的基于视觉感知的配电线路无人机巡视目标检测方法。通过可见光相机与紫外相机对无人机巡线图像进行采集,分析了紫外相机采集原理。通过直方图均衡化处理对采集图像进行增强处理,选用高斯滤波器对采集的配电线路无人机巡视图像进行滤波处理。对配电线路密集度区域和偏心度区域进行提取,获取配电线路无人机巡视目标特征向量。依据得到的特征向量确定无人机巡视目标,求出目标质心坐标,从而实现配电线路无人机巡视目标检测。实验结果表明,所提方法能够在干扰环境下有效实现配电线路无人机巡视目标的检测,检测结果可靠性高。
The environment in which distribution lines are located is complex. The inspection methods of unmanned aerial vehicles in traditional distribution lines can not effectively remove external environmental interference and lead to unreliable detection results. Therefore,a new detection method based on visual perception of unmanned aerial vehicles is proposed. Through the visible light camera and the UV camera,the unmanned aerial vehicle line images were collected and the principle of the ultraviolet camera acquisition was analyzed. Through the histogram equalization process to enhance the image acquisition processing,the selection of Gaussian filter on the distribution line unmanned aircraft inspection image filtering. Extracting the density of the distribution lines and the area of eccentricity to obtain the target vector of the unmanned aerial vehicles in the distribution line. Based on the eigenvector obtained,the target of UAV is determined,and the target centroid coordinate is obtained,so as to realize the detection of the unmanned aerial vehicle（ UAV） inspection target. The experimental results show that the proposed method can effectively detect the unmanned aerial vehicles＇ inspection targets in the interference environment,and the detection results are highly reliable.
作者
李军锋
王鹏
刘晓
曾钰
赵永强
LI Jun-feng;WANG Peng;LIU Xiao;ZENG Yu;ZHAO Yong-qiang(Education Training Evaluation Center Network Education Training Department,Guangdong Power Grid co.,LTD1,Guangzhou 510000,China;Zhanjiang Power Supply Bureau,Guangdong Power Grid Co.LTD2,Zhanjiang 524200,China)
出处
《科学技术与工程》
北大核心
2018年第16期231-236,共6页
Science Technology and Engineering
关键词
视觉感知
配电线路
无人机
巡视
目标检测
visual perception
distribution line unmanned
aerial vehicle inspection
target detection
