影像间的匹配点通常受基本矩阵或单应矩阵模型约束,利用不同的描述模型引导匹配会得到不同的匹配结果,并且将直接影响后续的三维重建结果。引入GRIC测度来检测匹配数据的拟合模型,推导了GRIC测度拟合单应矩阵模型和基本矩阵模型的误差方程。模拟数据和真实影像数据实验表明,GRIC测度比利用误差大小来检测拟合模型更加可靠,可以更有效地剔除误匹配点。
Matching points between two images usually conform to the constraint described by fundamental matrix or homography matrix model. Using different models to guide matching may lead to different matching results and directly affect the results of the follow-up of three-dimensional reconstruction. GRIC Measure is introduced to detect the fitting model of matching data, and the error equations of homography matrix model and fundamental matrix model in GRIC Measure are derived. Experimental results of both simulation data and real images data show that using GRIC Measure is more reliable than the magnitude of error to detect the fitting model, and can eliminate the false matching points more effectively.