Compared with vertical photogrammtry, oblique photogrammetry is radically different for images acquired from sensor with big yaw, pitch, and roll angles. Image matching is a vital step and core problem of oblique low-altitude photogrammetric process. Among the most popular oblique images matching methods are currently SIFT/ASIFT and many affine invariant feature-based approaches, which are mainly used in computer vision, while these methods are unsuitable for requiring evenly distributed corresponding points and high efficiency simultaneously in oblique photogrammetry. In this paper, we present an oblique low-altitude images matching approach using robust perspective invariant features. Firstly, the homography matrix is estimated by a few corresponding points obtained from top pyramid images matching in several projective simulation. Then images matching are implemented by sub-pixel Harris corners and descriptors after shape perspective transforming on the basis of homography matrix. Finally, the error or gross error matched points are excluded by epipolar geometry, RANSAC algorithm and back projection constraint. Experimental results show that the proposed approach can achieve more excellent performances in oblique low-altitude images matching than the common methods, including SIFT and SURF. And the proposed approach can significantly improve the computational efficiency compared with ASIFT and Affine-SURF.