As a fundamental and critical task in various visual applications, image matching can identify then correspond the same or similar structure/content from two or more images. Over the past decades, growing amount and diversity of methods have been proposed for image matching, particularly with the development of deep learning techniques over the recent years. However, it may leave several open questions about which method would be a suitable choice for specific applications with respect to different scenarios and task requirements and how to design better image matching methods with superior performance in accuracy, robustness and efficiency. This encourages us to conduct a comprehensive and systematic review and analysis for those classical and latest techniques. Following the feature-based image matching pipeline, we first introduce feature detection, description, and matching techniques from handcrafted methods to trainable ones and provide an analysis of the development of these methods in theory and practice. Secondly, we briefly introduce several typical image matching-based applications for a comprehensive understanding of the significance of image matching. In addition, we also provide a comprehensive and objective comparison of these classical and latest techniques through extensive experiments on representative datasets. Finally, we conclude with the current status of image matching technologies and deliver insightful discussions and prospects for future works. This survey can serve as a reference for (but not limited to) researchers and engineers in image matching and related fields.

/pdf/image-matching-from-handcrafted-to-deep-features-a-survey-27kn6w026w.pdf

Image Matching from Handcrafted to Deep Features: A Survey

We propose 3DSmoothNet, a full workflow to match 3D point clouds with a siamese deep learning architecture and fully convolutional layers using a voxelized smoothed density value (SDV) representation. The latter is computed per interest point and aligned to the local reference frame (LRF) to achieve rotation invariance. Our compact, learned, rotation invariant 3D point cloud descriptor achieves 94.9% average recall on the 3DMatch benchmark data set, outperforming the state-of-the-art by more than 20 percent points with only 32 output dimensions. This very low output dimension allows for near realtime correspondence search with 0.1 ms per feature point on a standard PC. Our approach is sensor- and scene-agnostic because of SDV, LRF and learning highly descriptive features with fully convolutional layers. We show that 3DSmoothNet trained only on RGB-D indoor scenes of buildings achieves 79.0% average recall on laser scans of outdoor vegetation, more than double the performance of our closest, learning-based competitors. Code, data and pre-trained models are available online at https://github.com/zgojcic/3DSmoothNet.

/pdf/the-perfect-match-3d-point-cloud-matching-with-smoothed-3k7v16l3h9.pdf

The Perfect Match: 3D Point Cloud Matching With Smoothed Densities

We present Deep Global Registration, a differentiable framework for pairwise registration of real-world 3D scans. Deep global registration is based on three modules: a 6-dimensional convolutional network for correspondence confidence prediction, a differentiable Weighted Procrustes algorithm for closed-form pose estimation, and a robust gradient-based SE(3) optimizer for pose refinement. Experiments demonstrate that our approach outperforms state-of-the-art methods, both learning-based and classical, on real-world data.

/pdf/deep-global-registration-3nht5po8s3.pdf

Deep Global Registration

Deep Global Registration.

The integration of multi-platform, multi-angle, and multi-temporal LiDAR data has become important for geospatial data applications. This paper presents a comprehensive review of LiDAR data registration in the fields of photogrammetry and remote sensing. At present, a coarse-to-fine registration strategy is commonly used for LiDAR point clouds registration. The coarse registration method is first used to achieve a good initial position, based on which registration is then refined utilizing the fine registration method. According to the coarse-to-fine framework, this paper reviews current registration methods and their methodologies, and identifies important differences between them. The lack of standard data and unified evaluation systems is identified as a factor limiting objective comparison of different methods. The paper also describes the most commonly-used point cloud registration error analysis methods. Finally, avenues for future work on LiDAR data registration in terms of applications, data, and technology are discussed. In particular, there is a need to address registration of multi-angle and multi-scale data from various newly available types of LiDAR hardware, which will play an important role in diverse applications such as forest resource surveys, urban energy use, cultural heritage protection, and unmanned vehicles.

Registration of Laser Scanning Point Clouds: A Review

A fast and robust local descriptor for 3D point cloud registration

TOLDI: An effective and robust approach for 3D local shape description

Feature correspondence selection is pivotal to many feature-matching based tasks in computer vision. Searching spatially k-nearest neighbors is a common strategy for extracting local information in many previous works. However, there is no guarantee that the spatially k-nearest neighbors of correspondences are consistent because the spatial distribution of false correspondences is often irregular. To address this issue, we present a compatibility-specific mining method to search for consistent neighbors. Moreover, in order to extract and aggregate more reliable features from neighbors, we propose a hierarchical network named NM-Net with a series of graph convolutions that is insensitive to the order of correspondences. Our experimental results have shown the proposed method achieves the state-of-the-art performance on four datasets with various inlier ratios and varying numbers of feature consistencies.

/pdf/nm-net-mining-reliable-neighbors-for-robust-feature-2fn96cw3br.pdf

NM-Net: Mining Reliable Neighbors for Robust Feature Correspondences

The local reference frame (LRF), as an independent coordinate system constructed on the local 3D surface, is broadly employed in 3D local feature descriptors. The benefits of the LRF include rotational invariance and full 3D spatial information, thereby greatly boosting the distinctiveness of a 3D feature descriptor. There are numerous LRF methods in the literature; however, no comprehensive study comparing their repeatability and robustness performance under different application scenarios and nuisances has been conducted. This paper evaluates eight state-of-the-art LRF proposals on six benchmarks with different data modalities (e.g., LiDAR, Kinect, and Space Time) and application contexts (e.g., shape retrieval, 3D registration, and 3D object recognition). In addition, the robustness of each LRF to a variety of nuisances, including varying support radii, Gaussian noise, outliers (shot noise), mesh resolution variation, distance to boundary, keypoint localization error, clutter, occlusion, and partial overlap, is assessed. The experimental study also measures the performance under different keypoint detectors, descriptor matching performance when using different LRFs and feature representation combinations, as well as computational efficiency. Considering the evaluation outcomes, we summarize the traits, advantages, and current limitations of the tested LRF methods.

Toward the Repeatability and Robustness of the Local Reference Frame for 3D Shape Matching: An Evaluation.

This article presents an efficient and robust estimator called compatibility-guided sampling consensus (CG-SAC) to achieve accurate 3-D point cloud registration. For correspondence-based registration methods, the random sample consensus (RANSAC) is served as a de facto solution for rigid transformation estimation from a number of feature correspondences. Unfortunately, RANSAC still suffers from two major limitations. First, it generates a hypothesis with at least three samples and desires a very large number of iterations to attain reasonable results, making it relatively time consuming. Second, the randomness during sampling can result in inaccurate results as it is highly potential to miss the optimal hypothesis. To solve these problems, we propose a compatibility-guided sampling strategy to eliminate randomness during sampling. In particular, only two correspondences are required by our method for hypothesis generation. We then rank correspondence pairs according to their compatibility scores because compatible correspondences are more likely to be correct and can yield more reasonable hypotheses. In addition, we propose a new geometric constraint named the distance between salient points (DSP) to measure the compatibility of two correspondences. Experiments on a set of real-world point cloud data with different application contexts and data modalities confirm the effectiveness of the proposed method. Comparison with several state-of-the-art estimators demonstrates the overall superiority of our CG-SAC estimator with regards to precision and time efficiency.

Jiaqi Yang

Papers

A fast and robust local descriptor for 3D point cloud registration

TOLDI: An effective and robust approach for 3D local shape description

NM-Net: Mining Reliable Neighbors for Robust Feature Correspondences

Toward the Repeatability and Robustness of the Local Reference Frame for 3D Shape Matching: An Evaluation.

Compatibility-Guided Sampling Consensus for 3-D Point Cloud Registration