摘要
设计变电站远程智能巡检图像目标快速识别算法,通过高效的目标识别性能,提升变电站巡检水平。利用无人机搭载相机,采集变电站远程智能巡检图像,通过小波变换方法滤波处理所采集的智能巡检图像。利用自适应阈值分割方法,分割已滤波处理的变电站远程智能巡检图像,将智能巡检图像划分为前景图像与背景图像。以智能巡检图像的前景图像为基础,构建高斯差分金字塔,提取图像的SIFT特征。依据所提取智能巡检图像的SIFT特征构建样本集,作为特征加权支持向量机的输入,利用特征加权支持向量机输出目标快速识别结果。实验结果表明,该算法可以快速识别变电站远程智能巡检图像中的目标,目标识别时间低于500 ms。
This paper designs substation remote intelligent inspection image target fast recognition algorithm,through efficient target recog-nition performance,improves substation inspection level.The remote intelligent inspection image of substation is collected by the UAV equipped with a camera,and the collected intelligent inspection image is filtered by wavelet transform method.The adap-tive threshold segmentation method is used to segment the filtered remote intelligent inspection image of substation,and the intel-ligent inspection image is divided into foreground image and background image.Based on the foreground image of the intelligent inspection image,the Gaussian difference pyramid is constructed and the SIFT features of the image are extracted.The sample set is constructed according to the SIFT features of the extracted intelligent inspection images,which is used as the input of the fea-ture-weighted support vector machine,and the fast target recognition results are output by the feature-weighted support vector ma-chine.The experimental results show that the algorithm can quickly identify the target in the substation remote intelligent inspec-tion image,and the target recognition time is less than 500 ms.
作者
卢铭翔
惠小东
金鑫
张家兴
陈煜敏
LU Ming-xiang;HUI Xiao-dong;JIN Xin;ZHANG Jia-xing;CHEN Yu-min(China Southern Power Grid Digital Grid Group Co.,Ltd.,Guangzhou 510700 China)
出处
《自动化技术与应用》
2025年第2期145-149,共5页
Techniques of Automation and Applications
基金
广东省科技攻关项目(202000450021)
南方电网数字电网集团有限公司(JY-OO-01-ZC-21-005-TQ)。
关键词
远程智能巡检
目标识别算法
SIFT特征
特征加权
remote intelligence inspection
object recognition algorithm
SIFT feature
feature weighting
