Abstract: A dense matching algorithm of close-range images constrained by iterative triangulation network was proposed to slove that traditional sparse matching could not meet the requirements of high precision 3-D modeling. Different from traditional statical ‘region by region’ propagation based on region-growing method, our algorithm utilized dynamic updating triangle matching propagation.Using SIFT and RANSAC methods, reliable sparse matching points would be obtained. Then corresponding Delaunay TIN were established. Meanwhile matching primitive were extracted from centeoid of triangle which area was larger than threshold in the left image. Our algorithm combined epipolar constraint, intensity correlation and corresponding triangles as constraint conditions for matching. After travering every triangle in turn, new corresponding points obtained by matching and initial points were used to establish Delaunay TIN. Iterate with above-mentioned matching strategy until that there was no new corresponding point. This paper adopts three groups typical close-range images to experiment. The results demonstrate that the proposed algorithm in this paper can obtain reliable matching results and apply different types of close-range images.