ElectricalAutomation《电气自动化》2023年第45卷第1期模式识别PatternIdentifications基于多曝光场景的新型电力图像配准技术潘金月,张维磊(国网安徽省电力有限公司滁州供电公司,安徽滁州239000)摘要:基于现有的图像配准技术,对电力图像拍摄时间差异所造成的曝光度不同而产生的图像配准误差大和配准效率提升难的问题,提出了一种基于电力图像去除亮度差异影响的边缘特征配准技术。通过提取图像的边缘信息,以产生二值化的效果图像代替原始图像进行旋转矩阵的计算,完成原始图像空间一致性调整,实现多曝光场景的新型电力图像配准技术,并通过一系列电力图像试验数据进行了验证。试验结果表明,所提的新型电力图像配准技术在配准效率方面较改进前的基于互相关信息配准方法提升了50%,配准精度达到了亚像素级别,配准方法鲁棒性更高。关键词:图像配准;多曝光图像处理;二值化;边缘检测;电力图像DOI:10.3969/j.issn.1000-3886.2023.01.034[中图分类号]TM391.4[文献标志码]A[文章编号]1000-3886(2023)01-0116-03NewPowerImageRegistrationTechnologyBasedonMulti-exposureScenePanJinyue,ZhangWeilei(ChuzhouElectricPowerSupplyBranch,StateGridAnhuiElectricPowerCo.,Ltd.,ChuzhouAnhui239000,China)Abstract:Basedontheproblemsoflargeimageregistrationerrorduetodifferentexposurecausedbythedifferenceofpowerimageshootingtimeanddifficulttoimprovetheregistrationefficiency,anedgefeatureregistrationtechnologybasedonremovingtheinfluenceofbrightnessdifferenceofpowerimagewasproposed.Byextractingtheedgeinformationoftheimage,itcangenerateabinaryeffectimageinsteadoftheoriginalimagetocalculatetherotationmatrix,completetheoriginalimagespatialconsistencyadjustment,andrealizethenewpowerimageregistrationtechnologyformulti-exposurescenes.Aseriesofelectricpowerimageexperimentaldatawereverified.Thetestresultsshowthattheproposednewpowerimageregistrationtechnologyhasimprovedtheregistrationefficiencyby50%comparedwiththeregistrationmethodbasedoncross-correlationinformationbeforetheimprovement,theregistrationaccuracyhasreachedthesub-pixellevel,andtheregistrationmethodisofhigherrobustness.Keywords:imageregistration;multiexposureimageprocessing;binar...