Point Cloud Denoising Research Articles

SITF: A Self-Supervised Iterative Training Framework for Point Cloud Denoising

Despite existing supervised point cloud denoising methods having made great progress, they require paired ideal noisy-clean datasets for training which is expensive and impractical in real-world applications. Moreover, they may perform the denoising process multiple times with fixed network parameters for better denoising results at test time. To address above issues, this paper proposes a self-supervised iterative training framework (SITF) for point cloud denoising, which only requires single noisy point clouds and a noise model. Given an off-the-shelf denoising network and original noisy point clouds, firstly, an intermediate noisier-noisy dataset is created by adding additional noises from the known noise model to noisy point clouds (i.e. learning targets). Secondly, after training on the noisier-noisy dataset, the denoising network is employed to denoise the original noisy point clouds to obtain the learning targets for the next iteration. The above two steps are iteratively and alternatively performed to get a better and better trained denoising network. Furthermore, to get better learning targets for the next round, this paper also proposes a novel iterative denoising network (IDN) architecture of stacked source attention denoising modules. The IDN explicitly models the iterative denoising process internally within a single network via reforming the given denoising network. Experimental results show that existing supervised networks trained through the SITF can achieve competitive denoising results and even outperform supervised networks under high noise conditions. The source code can be found at: https://github.com/VCG-NJUST/SITF.

PCDNF: Revisiting Learning-Based Point Cloud Denoising via Joint Normal Filtering.

Point cloud denoising is a fundamental and challenging problem in geometry processing. Existing methods typically involve direct denoising of noisy input or filtering raw normals followed by point position updates. Recognizing the crucial relationship between point cloud denoising and normal filtering, we re-examine this problem from a multitask perspective and propose an end-to-end network called PCDNF for joint normal filtering-based point cloud denoising. We introduce an auxiliary normal filtering task to enhance the network's ability to remove noise while preserving geometric features more accurately. Our network incorporates two novel modules. First, we design a shape-aware selector to improve noise removal performance by constructing latent tangent space representations for specific points, taking into account learned point and normal features as well as geometric priors. Second, we develop a feature refinement module to fuse point and normal features, capitalizing on the strengths of point features in describing geometric details and normal features in representing geometric structures, such as sharp edges and corners. This combination overcomes the limitations of each feature type and better recovers geometric information. Extensive evaluations, comparisons, and ablation studies demonstrate that the proposed method outperforms state-of-the-art approaches in both point cloud denoising and normal filtering.

A vision measurement method for ship hull plates based on multi-view stereo and image segmentation

The background denoising of point clouds is challenging for the measurement of ship hull plates. To address this issue, a vision measurement method is proposed by combining an identity cost volume network (ICV-Net) and a segment anything model (SAM). By applying the ship hull plate masks segmented by SAM to the corresponding depth maps inferred by ICV-Net, the background noise can be directly removed at the depth map level. Additionally, an edge extraction optimization algorithm complementary to SAM is proposed to eliminate the thickness-related noise. Moreover, an efficient depth map refinement algorithm is proposed by imposing the smoothness prior of ship hull plate surfaces and multi-view photometric consistency. Sub-pixel accurate depth maps of ship hull plates can be obtained by the depth map refinement. Experimental results suggest that the proposed method can automatically, accurately and efficiently remove the background noise of ship hull plates, which is crucial toward practical application.

Self-supervised multi-echo point cloud denoising in snowfall

Snowfall can cause noise to light detection and ranging (LiDAR) data. This is a problem since it is used in many outdoor applications, e.g., autonomous driving. We propose the task of multi-echo denoising, where the goal is to pick the echo that represents the objects of interest and discard other echoes. Thus, the idea is to pick points from alternative echoes unavailable in standard strongest echo point clouds. Intuitively, we are trying to see through the snowfall. We propose a novel self-supervised deep learning method and the characteristics similarity regularization to achieve this goal. The characteristics similarity regularization utilizes noise characteristics to increase performance. The experiments with a real-world multi-echo snowfall dataset prove the efficacy of multi-echo denoising and superior performance to the baseline. Moreover, based on extensive experiments on a semi-synthetic dataset, our method achieves superior performance compared to the state-of-the-art in self-supervised snowfall denoising. Our work enables more reliable point cloud acquisition in snowfall. The code is available at https://github.com/alvariseppanen/SMEDen.

Point Cloud Denoising in Outdoor Real-World Scenes Based on Measurable Segmentation

Point Cloud Denoising in Outdoor Real-World Scenes Based on Measurable Segmentation

PathNet: Path-Selective Point Cloud Denoising.

Current point cloud denoising (PCD) models optimize single networks, trying to make their parameters adaptive to each point in a large pool of point clouds. Such a denoising network paradigm neglects that different points are often corrupted by different levels of noise and they may convey different geometric structures. Thus, the intricacy of both noise and geometry poses side effects including remnant noise, wrongly-smoothed edges, and distorted shape after denoising. We propose PathNet, a path-selective PCD paradigm based on reinforcement learning (RL). Unlike existing efforts, PathNet enables dynamic selection of the most appropriate denoising path for each point, best moving it onto its underlying surface. We have two more contributions besides the proposed framework of path-selective PCD for the first time. First, to leverage geometry expertise and benefit from training data, we propose a noise- and geometry-aware reward function to train the routing agent in RL. Second, the routing agent and the denoising network are trained jointly to avoid under- and over-smoothing. Extensive experiments show promising improvements of PathNet over its competitors, in terms of the effectiveness for removing different levels of noise and preserving multi-scale surface geometries. Furthermore, PathNet generalizes itself more smoothly to real scans than cutting-edge models.

Building Extraction from LiDAR Point Clouds Based on Revised RandLA-Net

Abstract. 3D building models is crucial for applications in smart cities. Automatic reconstruction of 3D buildings has been investigated based on various data sources. Point clouds from airborne LiDAR scanners can be used to extract buildings data due to its high accuracy and point density. In this paper, we present a methodology to segment buildings and corresponding rooftop structure from point clouds. First, RandLA-Net, which is an efficient and lightweight neural network for semantic segmentation of large-scale point clouds, is revised and adopted for building segmentation. By implementing local feature aggregation of each point, RandLA-Net can effectively preserve geometric details in point clouds. Besides 3D coordinates of point clouds, we incorporated point attributes including pulse intensity and return numbers into the network as additional features. Feature normalizations are applied to the input features. To achieve a better result of the local feature aggregation, hyperparameters of the network are fine-tuned according to the density of points and building size. Based on the classified building point clouds, DBSCAN clustering algorithm is implemented for segmenting individual buildings. Elevation histogram analysis is conducted to determine optimal threshold values for delineating candidate rooftop point clouds of individual buildings. For the buildings with multiple rooftops, multiple elevation threshold values are necessary to extract corresponding rooftops or walls. Then DBSCAN is employed again for segmentation of individual rooftops and denoising of point clouds of each building. Finally, Alpha-shape analysis is applied based on adaptive threshold values to build the envelope of each rooftop. Experiments show that our implementation of building segmentation using RandLA-net achieves higher mean IoU (Intersection over Union) and better classification performance in building segmentation. ISPRS benchmark data was used in our experiment and our methodology produce results with accuracy of 90.79%.

Detecting people in sprinting motion using HPRDenoise: Point cloud denoising with hidden point removal

<p>LiDARs are utilized in various applications, such as self-driving vehicles and robotics, to aid in sensing the environment. However, LiDARs do not provide instantaneous images and they generate noise, adding to measurement errors. This noise, often referred to as motion blur phenomenon also observed in other imaging sensors results in decreased sensing accuracy for moving objects. This study introduces HPRDenoise, a noise reduction method based on hidden point removal, specifically designed to reduce motion blur during sprinting motion. This method capitalizes on the occlusion produced by a fixed-position LiDAR. We propose a comprehensive denoising approach to filter points from a point cloud without resorting to supervised learning, unlike most existing denoising algorithms. The number of correct frames and accuracy were compared for Raw, ScoreDenoise, which is the state-of-the-art method for random point cloud denoising, and HPRDenoise (Ours). Accuracy is defined as the ratio of the number of correct frames to the total number of frames. Experimental results demonstrate that the detection accuracy of point clouds processed with HPRDenoise is 72.73%, achieving better accuracy than those using conventional methods.</p>

Plant-Denoising-Net (PDN): A plant point cloud denoising network based on density gradient field learning

Effective point cloud denoising is critical in 3D plant phenotyping applications, which reduces interference in subsequent algorithms and improves the accuracy of plant phenotypes measurement. Deep learning-based point cloud denoising algorithms have shown excellent denoising performance on simple CAD models. However, these algorithms suffer from issues including over-smoothing or shrinkage and low efficiency when applied on density uneven, incomplete, various types of noise and complex plant point clouds. We proposed a plant point cloud denoising network (PDN) based on point cloud density gradient field learning, which can effectively address the challenges posed by plant point clouds. PDN consists of three main modules: point density feature (PDF) exception module, umbrella operator feature (UOF) computation module, and point density gradient (DG) estimation module. The performance of PDN was evaluated in experiments using point clouds of multiple plant species with noise of different types. Under different levels of Gaussian noise, our method achieved a relative performance improvement of 7.6%-19.3% compared to the state-of-the-art baseline methods, reaching state-of-the-art denoising performance. For noise of different types, the majority of our denoising results outperformed the baseline methods. In addition, our method was 0.5 and 8.6 times faster than the baseline methods when processing point clouds with low and high noise level, respectively. The good robustness, generalization, and computational efficacy of PDN are expected to facilitate the acquisition of high-precision 3D point clouds for various plant species, enhance the versatility of 3D phenotyping methods, improve the accuracy of the measurement of structural phenotypes, and increase the throughput of data processing, therefore facilitate the development of modern breeding research. The source code and the datasets used in this study is available on GitHub at https://github.com/suetme/PDN-plant-denoising-net.

Denoising point clouds with fewer learnable parameters

Point cloud denoising is a crucial task in the field of geometric processing. Recent years have witnessed significant advancements in deep learning-based point cloud denoising algorithms. These methods, compared to traditional techniques, demonstrate enhanced robustness against noise and produce point cloud data of higher fidelity. Despite their impressive performance, achieving a balance between denoising efficacy and computational efficiency remains a formidable challenge in learning-based methods. To solve this problem, we introduce LPCDNet, a novel lightweight point cloud denoising network. LPCDNet consists of three main components: a lightweight feature extraction module utilizing trigonometric functions for relative position encoding; a non-parametric feature aggregation module to leverage semantic similarities for global context comprehension; and a decoder module designed to realign noise points with the underlying surface. The network is designed to capture both local details and non-local structures, thereby ensuring high-quality denoising outcomes with a minimal parameter count. Extensive experimental evaluations demonstrate that LPCDNet achieves comparable or superior performance to state-of-the-art methods, while significantly reducing the number of learnable parameters and the necessary running time.

PointCVaR: Risk-Optimized Outlier Removal for Robust 3D Point Cloud Classification

With the growth of 3D sensing technology, the deep learning system for 3D point clouds has become increasingly important, especially in applications such as autonomous vehicles where safety is a primary concern. However, there are growing concerns about the reliability of these systems when they encounter noisy point clouds, either occurring naturally or introduced with malicious intent. This paper highlights the challenges of point cloud classification posed by various forms of noise, from simple background noise to malicious adversarial/backdoor attacks that can intentionally skew model predictions. While there's an urgent need for optimized point cloud denoising, current point outlier removal approaches, an essential step for denoising, rely heavily on handcrafted strategies and are not adapted for higher-level tasks, such as classification. To address this issue, we introduce an innovative point outlier cleansing method that harnesses the power of downstream classification models. Using gradient-based attribution analysis, we define a novel concept: point risk. Drawing inspiration from tail risk minimization in finance, we recast the outlier removal process as an optimization problem, named PointCVaR. Extensive experiments show that our proposed technique not only robustly filters diverse point cloud outliers but also consistently and significantly enhances existing robust methods for point cloud classification. A notable feature of our approach is its effectiveness in defending against the latest threat of backdoor attacks in point clouds.

Point cloud denoising using a generalized error metric

Effective removal of noises from raw point clouds while preserving geometric features is the key challenge for point cloud denoising. To address this problem, we propose a novel method that jointly optimizes the point positions and normals. To preserve geometric features, our formulation uses a generalized robust error metric to enforce piecewise smoothness of the normal vector field as well as consistency between point positions and normals. By varying the parameter of the error metric, we gradually increase its non-convexity to guide the optimization towards a desirable solution. By combining alternating minimization with a majorization-minimization strategy, we develop a numerical solver for the optimization which guarantees convergence. The effectiveness of our method is demonstrated by extensive comparisons with previous works.

Point Cloud Denoising and Feature Preservation: An Adaptive Kernel Approach Based on Local Density and Global Statistics.

Noise removal is a critical stage in the preprocessing of point clouds, exerting a significant impact on subsequent processes such as point cloud classification, segmentation, feature extraction, and 3D reconstruction. The exploration of methods capable of adapting to and effectively handling the noise in point clouds from real-world outdoor scenes remains an open and practically significant issue. Addressing this issue, this study proposes an adaptive kernel approach based on local density and global statistics (AKA-LDGS). This method constructs the overall framework for point cloud denoising using Bayesian estimation theory. It dynamically sets the prior probabilities of real and noise points according to the spatial function relationship, which varies with the distance from the points to the center of the LiDAR. The probability density function (PDF) for real points is constructed using a multivariate Gaussian distribution, while the PDF for noise points is established using a data-driven, non-parametric adaptive kernel density estimation (KDE) approach. Experimental results demonstrate that this method can effectively remove noise from point clouds in real-world outdoor scenes while maintaining the overall structural features of the point cloud.

Research on the digital protection and three-dimensional modeling technology of ancient buildings

Abstract The starting point of ancient architecture digitization and 3D reconstruction research is data acquisition. Thus, this paper mainly analyzes the processing of 3D laser scanning data and the construction accuracy of 3D reconstruction models of ancient buildings based on point cloud data. This paper discusses the basic theory of point cloud data and the preprocessing of point cloud data, including point cloud splicing, point cloud denoising, point cloud streamlining, point cloud segmentation, and other methods and applicability are discussed in detail. The boundary extraction algorithm based on RANSAC and the graph cut algorithm are utilized to obtain the main frame line of the building, and a proposed method for reconstruction based on multi-view projection is proposed. Classify laser point cloud samples from ancient buildings, calculate each wall color’s R, G, and B values, and obtain preliminary classifications of laser point cloud data categories. The RANSAC algorithm was used to extract the point cloud data plane extraction peak of the ancient tower building, combined with the three-dimensional modeling coordinate difference of the test points, to analyze the accuracy of the three-dimensional modeling technology of ancient buildings in this paper. The number of plane extraction points of the RANSAC algorithm is 9504, and the standard deviation is 0.0505, which is close to 0. The results are more accurate, and the extraction effect has a certain degree of superiority. The plane coordinate deviation of the reconstruction model of the ancient tower is mainly concentrated in the range of 0.10~0.40m, and the individual coordinate difference is more than 0.10m. The errors in the x -direction and y -direction are ±0.18m and ±0.23m, respectively, and it can be seen from the results of the error calculation that the three-dimensional reconstruction plan of the building proposed in this paper can complete the reconstruction of the three-dimensional model of the building with a high degree of accuracy.

Incomplete Gamma Kernels: Generalizing Locally Optimal Projection Operators.

We present incomplete gamma kernels, a generalization of Locally Optimal Projection (LOP) operators. In particular, we reveal the relation of the classical localized L1 estimator, used in the LOP operator for point cloud denoising, to the common Mean Shift framework via a novel kernel. Furthermore, we generalize this result to a whole family of kernels that are built upon the incomplete gamma function and each represents a localized Lp estimator. By deriving various properties of the kernel family concerning distributional, Mean Shift induced, and other aspects such as strict positive definiteness, we obtain a deeper understanding of the operator's projection behavior. From these theoretical insights, we illustrate several applications ranging from an improved Weighted LOP (WLOP) density weighting scheme and a more accurate Continuous LOP (CLOP) kernel approximation to the definition of a novel set of robust loss functions. These incomplete gamma losses include the Gaussian and LOP loss as special cases and can be applied to various tasks including normal filtering. Furthermore, we show that the novel kernels can be included as priors into neural networks. We demonstrate the effects of each application in a range of quantitative and qualitative experiments that highlight the benefits induced by our modifications.

FCNet: Learning Noise-Free Features for Point Cloud Denoising

The acquisition of point clouds is usually accompanied by noise due to imperfect laser scanning or image-based reconstruction techniques. Deep learning-based methods have achieved impressive performance in point cloud denoising. However, the features captured by a denoising network from noisy point clouds are usually contaminated by noise during training. The feature noise will lead to the oscillation of back-propagated gradients, which interferes with parameter optimization and reduces the denoising performance. In this paper, we propose to explicitly clean up feature noise for point cloud denoising from two aspects: feature noise cleaning and network training. From the first aspect, we propose the feature clean network (FCNet for short) to explicitly clean up the feature noise. From the second aspect, we train FCNet by a teacher-student learning model to learn the noise-free features under the guidance of feature domain losses. Specifically, FCNet is designed with emphasis on two modules: non-local self-similarity (NSS) and weighted average pooling (WAP). NSS module smooths features through a non-local filter based on the inherent non-local self-similarity of point clouds. WAP module applies original weights calculated by the statistical outlier removal algorithm to suppress the feature noise induced by outliers. In the teacher-student learning model, we introduce a clean input using the noisy point and its clean neighbors. The teacher network accepts the clean input to capture noise-free features. The student network is trained to imitate the teacher network to learn noise-free features by minimizing the feature loss. The experiments on synthetic and real scanned point clouds show that FCNet outperforms state-of-the-art point cloud denoising methods.

DLC-SLAM: A Robust LiDAR-SLAM System With Learning-Based Denoising and Loop Closure

The current light detection and ranging simultaneous localization and mapping (LiDAR-SLAM) system suffers greatly from low accuracy and limited robustness when faced with complicated circumstances. From our experiments, we find that current LiDAR-SLAM systems have limited performance when faced with specular surfaces such as glass, certain metals, and building walls that are rich in urban environments. Therefore, in this work, we propose a general LiDAR-SLAM system termed <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">d</u> enoising and <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</u> oop <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">c</u> losure (DLC-SLAM) to tackle the problem of denoising and loop closure in complex environments with many noises and outliers caused by reflective materials. Current approaches for point cloud denoising are mainly designed for small-scale point clouds and cannot be extended to large-scale point cloud scenes. In this work, we first proposed a lightweight network for large-scale point cloud denoising. Subsequently, we have also designed an efficient loop closure network for place recognition in global optimization to improve the localization accuracy of the whole system. Finally, we demonstrated by extensive experiments and benchmark studies that our proposed LiDAR-SLAM system has a significant boost in localization accuracy when faced with noisy point clouds, with a marginal increase in the computational cost. Our systematical implementations are open-sourced to benefit the community.

Random screening-based feature aggregation for point cloud denoising

Raw point clouds captured by sensing devices are often contaminated with noise, which perturbs the fidelity of the original geometric information. Point cloud denoising is therefore an inseparable post-processing step, aiming to remove the noise in the point clouds. Existing point cloud denoising approaches are typically trained on datasets that have uniform point distributions and densities, making them unsuitable for effectively denoising point clouds with severe noise or irregular point distributions. In this paper, we introduce a novel random screening-based feature aggregation method for point cloud denoising. Our key insight is that merging features of dense and sparse points assists with enhancing the quality of point cloud denoising results. In specific, our approach involves randomly screening the features of local point patches and fusing richer geometric information of denser points into sparser point representations. Comprehensive experiments demonstrate that our method achieves state-of-the-art performance in the point cloud denoising task on both synthetic and real-world datasets.

Alternately denoising and reconstructing unoriented point sets

We propose a new strategy to bridge point cloud denoising and surface reconstruction by alternately updating the denoised point clouds and the reconstructed surfaces. In Poisson surface reconstruction, the implicit function is generated by a set of smooth basis functions centered at the octnodes. When the octree depth is properly selected, the reconstructed surface is a good smooth approximation of the noisy point set. Our method projects the noisy points onto the surface and alternately reconstructs and projects the point set. We use the iterative Poisson surface reconstruction (iPSR) to support unoriented surface reconstruction. Our method iteratively performs iPSR and acts as an outer loop of iPSR. Considering that the octree depth significantly affects the reconstruction results, we propose an adaptive depth selection strategy to ensure an appropriate depth choice. To manage the oversmoothing phenomenon near the sharp features, we propose a λ-projection method, which means to project the noisy points onto the surface with an individual control coefficient λi for each point. The coefficients are determined through a Voronoi-based feature detection method. Experimental results show that our method achieves high performance in point cloud denoising and unoriented surface reconstruction within different noise scales, and exhibits well-rounded performance in various types of inputs. The source code is available at https://github.com/Submanifold/AlterUpdate.

Research on the Adaptability of Typical Denoising Algorithms Based on ICESat-2 Data

Photon-counting light detection and ranging (LiDAR) emits and receives weak photon signals, which are easily mixed with background noise caused by the sun, the atmosphere, etc., and is thus difficult to distinguish. Therefore, point-cloud denoising is a key step in point-cloud data processing of photon-counting LiDAR. To explore the adaptability of different denoising algorithms for photon-counting LiDAR data in different times and spaces, in this paper, NASA’s official differential, regressive and Gaussian adaptive nearest neighbor (DRAGANN) algorithm; Herzfeld’s radial basis function (RBF) denoising algorithm; and the density-based spatial clustering of applications with noise (DBSCAN) algorithm based on density clustering are used to denoise the ICESat-2 ATL03 photon point-cloud data. Airborne LiDAR data are used to verify the denoising accuracy, and then the adaptability of the three algorithms is discussed. The results show that the DRAGANN algorithm is suitable for data with moderate Fraction Vegetation Coverage (FVC) (45–75%) at night and is less affected by slope; therefore, it is not limited to terrain slope. The denoising accuracy of the RBF algorithm decreases with increasing FVC and decreases with increasing slope. It is suitable for data with low terrain slope (0~55°) and low FVC (0~220°), which is less affected by observation time; therefore, it is suitable for all-day data. The DBSCAN algorithm is suitable for data with moderate FVC (45~75%) at night, regardless of terrain slope. Unlike the DRAGANN algorithm, the DBSCAN algorithm is greatly affected by solar noise photons, but at night, its denoising accuracy is higher than that of the DRAGANN algorithm. The research results have certain reference significance for the subsequent processing and application of ICESat-2 data.