基于轮廓曲线的快速高精度可靠图像匹配

Fast and reliablehigh-precision contourcurvesimage matching based on profile curve

为了提高一键式检测中图像匹配精度和速度,提出一种基于轮廓曲线的快速高精度图像配准算法:根据定义的图像匹配差异度度量,采用图像金字塔搜索匹配策略,利用相应的相似性度量进行图像匹配。具体流程为:对采集图像和模板分别建立图像金字塔,对每层图像和模板使用sobel算子提取边缘。对顶层图像使用归一化角点距离矩阵与模板进行粗匹配,然后使用同心圆划分法进行细匹配,获取精确位置后映射至金字塔下一层,再次使用同心圆方法细匹配获取精确位置,目标坐标逐层向下映射,最终获取原图中目标的真实位置。实验表明该方法提升了一键式测量仪的匹配速度和匹配精度,而且不受遮挡、非线性光照变化、对比度低、全局对比度反转、局部对比度反转等情响。减少测量准备时间,提升了一键式测量仪的性能。

In order to improvethe detection of one-clickimage matchingaccuracy andspeed, isproposed based on the rapidprecision imageregistration algorithmcontourcurves： differenceimage matching metricde finedby imagepyramidmatchingsearchstrategy, using the correspondingsimilarity measureimage matching. Specificprocedures：collecting imagesand template stocreate an imagepyramid, respectively, foreachimage and the template usingthe sobel operatoredge extraction. Top-levelimages usingnor malizedrou ghmatching angleand distance matrix template,use themat chaftera fineconcentric division method, after obtaining precise locationmapping to the nextlayer ofthe pyramid, once againwith a fineconcentricmat chingmethodto obtain accurateposition, targetcoordinatemappinglayer eventuallyo btainingthe true positionof theoriginaltarget. Experiments show that by layerdown, the methodim provesmatching speedandaccuracy ofmatchingone-buttonmeter, and unobstructed, nonlinear changes inillumination, low contrast, globalcontrast change, local contrastreversaland otherloudemotion. Reduce the measurementpreparation time,improvedone-buttonmeterperformance.