图像模糊及抖动偏移的机器视觉位移监测方法  

A visual displacement monitoring method considering image blur and offset

作　　者：陶庭叶[1] 胡尚 时梦杰 谢广阔 马敏 TAO Tingye;HU Shang;SHI Mengjie;XIE Guangkuo;MA Min(School of Civil Engineering,Hefei University of Technology,Hefei 230009,China)

机构地区：[1]合肥工业大学土木与水利工程学院,合肥230009

出　　处：《测绘科学》2022年第12期83-89,222,共8页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(42104019);煤炭资源清洁利用与矿山环境保护湖南省重点实验室开放研究基金项目(E22015)

摘　　要：针对在利用计算机视觉方法进行变形监测时,因相机抖动导致监测结果出现较大偏差的问题,该文提出了一种抗抖动的视觉监测方法。首先,提出了一种通用图像质量评估方法,用以评估图像质量。利用高斯滤波提取出图像均值广义高斯模型提取均值减损对比度归一化系数矩阵,并计算其方差均值作为图像模糊评价指标,进而识别剔除模糊图像。进一步,在清晰图像中选取稳定参考目标,利用尺度不变特征转换特征匹配估计全局抖动参数,修正图像偏移。提出的方法可有效筛选出模糊图像,对图像抖动误差的计算精度较高,可以有效解决视觉监测过程中的相机抖动问题,满足工程变形监测需求。Aiming at the problem of large deviation of monitoring results caused by inevitably camera jitter when using computer vision method for deformation monitoring,this paper proposed an anti-jitter visual monitoring method.Firstly,an image quality assessment method was put forward to extract their mean subtracted contrast normalized coefficients and evaluated images’qualities.After that,blurred images were recognized and eliminated exactly.Further,for the remaining clear images,the stable reference targets were selected and the scale invariant feature transform method were employed to estimate the global movement parameters and correct the image offset.Test results on different datasets showed that the proposed method could effectively screen out blurred images with high accuracy in the calculation of image jitters error.It could effectively solve the problem of camera shake in the process of visual monitoring,which met the needs of engineering deformation monitoring exactly.

关 键 词：图像质量评估 均值减损对比度归一化 抖动估计 位移监测 视觉测量 尺度不变特征转换 

分 类 号：P237[天文地球—摄影测量与遥感]