The traditional binocular stereo vision measurement is based on stereo rectification for 3D reconstruction. The measurement accuracy is greatly influenced by the calibration parameter accuracy, interpolation accuracy and binocular sensor structure. When measuring the deformation of surface, especially large curvature ROI, the stereo rectification process could cause the loss or over-fitting of the non-uniform deformation information in the stereoscopic image pairs, which further affects the measurement accuracy. To address this issue, a high-precision stereo matching method is proposed based on reprojection and 3D-DIC, which is suitable for surface deformation measurement without stereo rectification or epi-polar geometric constraint correction. Meanwhile, this method can also be extended to high-precision matching of time series image pairs containing non-uniform deformation. Specifically, this article proposes the estimation method of second-order deformation parameters, the growth matching strategy of second-order deformation parameters, and the global stereo matching strategy and the time series matching strategy. Furthermore, the 3D reconstruction method of feature points independent of stereo rectification and epi-polar correction, and the calculation method of global deformation field and local strain field are given. Experiments show that proposed method can achieve high precision deformation measurement of a curved surface, and the measuring system can achieve an average measurement error of less than 1 μm in a certain depth of field.