双目立体视觉测量中的特征点快速匹配算法

Fast image feature point matching algorithm on binocular stereo vision of distance measurement

对于图像特征点的提取与匹配是双目立体视觉测量系统中的重要一步,提出了一种改进的适用于随机噪声、光照变化、局部遮挡等复杂环境下的快速鲁棒性不变特征(SURF)的匹配算法。首先,改进了特征点描述子中的主方向的计算方法,提高了主方向的精确度和计算效率;其次,结合NN/SN算法和RANDSAC算法,提出特征点的反向匹配算法,克服了外部环境光照变化、噪声、局部遮挡等因素对特征点匹配的影响,有效地提高了匹配精度。

For image feature points extraction and matching is an important step of binocular stereo vision measurement system, and this paper proposes an improved algorithm of speeded robustness invariant features （SURF） matching which is applicable to random noise, illumination change, local keep out the complex environment. First of all, it improves the feature points in the descriptor the calculation method of the main direction, the principal direction accuracy and computational efficiency; Secondly, combined with NN/SN algorithm and RANDSAC algorithm, it puts forward the reverse feature point matching algorithm, which overcomes the external environment illumination change, noise, local shelter and ete factors on the influence of the feature point matching, effectively improves the precision of matching.