随机抽样一致性算法是计算机视觉领域应用最广泛的鲁棒性算法。为了进一步提高RANSAC算法的运算速度，首先在介绍RANSAC算法的Tc,d预检验加速模型的基础上，提出了一种两步法用来实现优化的预检验参数选择；然后基于这种优化选择方法提出了自适应Tc,d预检验的新算法，从而实现了不依赖用户选择的RANSAC算法的自适应加速。基于窄基线和宽基线图像对的极线几何计算的实验表明，该新算法相对于标准RANSAC算法的运算速度平均提升超过了400%。

RANSAC is the most widely used robust regression algorithm in computer vision.Starting from the Tc,d pre-evaluation model of RANSAC algorithm，a two-step method is presented for optimal (c,d) selection. Based on this method，the adaptive Tc,d test extension is proposed to achieve user independent RANSAC acceleration. We show experimentally that using both short-baseline and wide-baseline epipolar geometry estimation，the proposed method is up to 400% faster than the standard RANSAC.