针对双目视觉测量中的立体匹配问题，提出了一种基于主成分分析(PCA)算法和尺度不变特征变换(SIFT)匹配算法相结合的双目视觉立体匹配新方法。该系统以GSI 编码点为特征点，利用双目相机对以散斑为背景的编码标识板进行成像，利用PCA 算法与SIFT 算法结合的新算法对采集的图像进行特征提取与匹配解算。可实现对GSI 编码点特征的准确提取和立体匹配,并测量出不同位姿下特征点之间的精确位移。实验验证部分引入GSI 编码技术并在行程为1000 mm×1000 mm 二维高精度平移台上进行，实际测得位移的绝对误差在5×10^-2 mm 之内，验证了该系统的高精确性。

Aim at stereo matching problem in binocular vision measurement, a binocular vision stereo matching method based on principal component analysis(PCA) algorithm and scale invariant feature transform(SIFT) algorithm is put forward. The system uses GSI code as its feature points, and the binocular camera imaging on code identification plate using speckle as its background. Then a new algorithm combining the PCA with the SIFT is used to extract feature and solve matching problem on the acquisition images. We can achieve accurate extraction and stereo matching on these GSI code points and measure the precise position and orientation under different displacement between the feature points. The experimental verification part introduced the GSI coding technique and it is carried at 1000 mm×1000 mm stroke dimensional precision translation stage. Actual measurement of displacement got the absolute error within 5×10^-2 mm, which can verify the high accuracy of the proposed system.