SeFM: A Sequential Feature Point Matching Algorithm for Object 3D Reconstruction

TL;DR

SeFM is a novel sequential feature point matching algorithm that significantly increases the number of matched points for 3D object reconstruction, outperforming traditional methods like SIFT and SURF in accuracy and quantity.

Contribution

The paper introduces SeFM, a new sequential matching algorithm utilizing epipolar geometry and dynamic programming to enhance feature point matching in 3D reconstruction.

Findings

Achieves 1,000 to 10,000 times more matching points than conventional algorithms.

Outperforms SIFT and SURF in precision and recall.

Enables semantically visible object reconstruction from two images.

Abstract

3D reconstruction is a fundamental issue in many applications and the feature point matching problem is a key step while reconstructing target objects. Conventional algorithms can only find a small number of feature points from two images which is quite insufficient for reconstruction. To overcome this problem, we propose SeFM a sequential feature point matching algorithm. We first utilize the epipolar geometry to find the epipole of each image. Rotating along the epipole, we generate a set of the epipolar lines and reserve those intersecting with the input image. Next, a rough matching phase, followed by a dense matching phase, is applied to find the matching dot-pairs using dynamic programming. Furthermore, we also remove wrong matching dot-pairs by calculating the validity. Experimental results illustrate that SeFM can achieve around 1,000 to 10,000 times matching dot-pairs, depending on individual image, compared to conventional algorithms and the object reconstruction with only two images is semantically visible. Moreover, it outperforms conventional algorithms, such as SIFT and SURF, regarding precision and recall.