This paper deals with using feature-based matching, FBM, to automatically accomplish the correspondence for conjugate points in a stereopair then to determine the relative orientation parameters. The approach includes : (1) coarse matching between a reduced image pairs by FBM to determine the affine transformation coefficients for patch prediction, (2) matching critical points in the corresponded patches, and (3) calculating relative orientation parameters and performing consistency check. In the first two steps, FBM is applied. Which includes (1) convolving each image with a Laplacian of Gaussian, LoG, mask, (2) vectorizing the zero crossing image, (3) performing line approximation to extract critical points, and (4) similarity assessment. An aerial stereopair was used in the case study. To improve the matching accuracy, the refinement for the image correspondence by area-based matching, ABM, is also studied. The results indicate that the RMS of Y-parallax is 0.8 pixels by FBM and is 0.25 pixels after the correspondence refinement.