为提升平行摄影隧洞序列影像的同名点匹配率及精度,对成像规律及匹配方法展开研究.提出了一种窗口尺度自动改正的LK光流法,核心是利用物像空间关系构建尺度差模型,并采用可变窗口进行特征点光流跟踪.结果表明:隧洞序列影像像点尺度差呈对称性径向分布,差异值符合幂函数模型增长趋势;将该方法应用在多组尺度差立体像对的匹配实验中,均取得了优于0.3 个像素精度的实验结果,较基本光流法至少提升34.3%,最大提升45.5%.研究成果可为顾及尺度差的匹配方法提供参考,也可为隧洞平行摄影立体影像匹配提供基础.
In pursuit of heightened accuracy in matching homologous points within parallel tunnel sequence images,this paper delved into the realms of imaging perspective and matching methodologies.A novel automated scale correction approach,rooted in the Lucas-Kanade(LK)optical flow method and implemented within a predefined window,was introduced.The core concept in-volved constructing a scale difference model based on the spatial relationship between the object and its image,coupled with the u-tilization of a variable window for tracking feature points'optical flow.The obtained results revealed a symmetric radial distribution of scale differences in tunnel sequence images,with difference values adhering to a power-law growth trend.When applied in matching experiments involving stereo image pairs with multiple scale differences,the proposed method consistently obtained ex-periment precision exceeding 0.3 pixels.This method was superior to basic optical flow method,with improvements ranging from 34.3%to 45.5%.These research findings not only contribute references for matching methods that account for scale differences,but also establish a foundation for stereo image matching in parallel tunnel photography.