Normal Estimation for 3D Point Clouds via Local Plane Constraint and Multi-scale Selection

Zhou, Jun; Huang, Hua; Liu, Bin; Liu, Xiuping

doi:10.1016/j.cad.2020.102916

Cited by 42 publications

(15 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Existing methods for computing normal vectors for LiDAR point clouds could be divided into learning-based and traditional deterministic approaches. For instance, [24,25,26] use a neural network to estimate the normal of point cloud directly. Using learning-based approaches would require extensive training data in order to achieve good performance and might need additional fine-tuning when using different types of LiDAR setups.…”

Section: Normal Surface Calculationmentioning

confidence: 99%

CPSeg: Cluster-free Panoptic Segmentation of 3D LiDAR Point Clouds

Li¹,

Razani²,

Xu³

et al. 2021

Preprint

View full text Add to dashboard Cite

A fast and accurate panoptic segmentation system for LiDAR point clouds is crucial for autonomous driving vehicles to understand the surrounding objects and scenes. Existing approaches usually rely on proposals or clustering to segment foreground instances. As a result, they struggle to achieve real-time performance. In this paper, we propose a novel real-time end-to-end panoptic segmentation network for LiDAR point clouds, called CPSeg. In particular, CPSeg comprises a shared encoder, a dual decoder, a task-aware attention module (TAM) and a cluster-free instance segmentation head. TAM is designed to enforce these two decoders to learn rich task-aware features for semantic and instance embedding. Moreover, CPSeg incorporates a new cluster-free instance segmentation head to dynamically pillarize foreground points according to the learned embedding. Then, it acquires instance labels by finding connected pillars with a pairwise embedding comparison. Thus, the conventional proposal-based or clustering-based instance segmentation is transformed into a binary segmentation problem on the pairwise embedding comparison matrix. To help the network regress instance embedding, a fast and deterministic depth completion algorithm is proposed to calculate surface normal of each point cloud in real-time. The proposed method is benchmarked on two large-scale autonomous driving datasets, namely, SemanticKITTI and nuScenes. Notably, extensive experimental results show that CPSeg achieves the state-of-the-art results among real-time approaches on both datasets.

show abstract

Section: Normal Surface Calculationmentioning

confidence: 99%

CPSeg: Cluster-free Panoptic Segmentation of 3D LiDAR Point Clouds

Li¹,

Razani²,

Xu³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…For example, Boulch et al [34] associate a 2D grid representation to the local neighborhood of a 3D point via a Hough transform, and formulate the normal estimation as a discrete classification problem in the Hough space. The second group of methods [7,8,9,10,11] directly estimate surface normals from unstructured point clouds. For instance, PCPNet [7] estimates surface normal by a deep multi-scale PointNet [28] architecture, which processes the multiple neighborhood scales jointly, thus leading to the phenomenon of oversmoothing.…”

Section: Learning-based Normal Estimationmentioning

confidence: 99%

“…To overcome the oversmoothing phenomenon, Nesti-Net [8] applied a MoE [36] structure to predict the optimal scale rather than direct concatenation of multiple scales, which yields an improved performance. Similarly, Zhou et al [9] improve surface normal estimation by using an extra feature constraint mechanism and a novel multiscale neighborhood selection strategy. Hashimoto et al [10] propose a joint model that exploits a PointNet for local feature extraction and a 3DCNN for the spatial feature encoding to efficiently incorporate local and spatial structures.…”

Section: Learning-based Normal Estimationmentioning

confidence: 99%

“…Surface fitting based methods. The former learningbased normal estimation studies [7,8,9,10,11,12] directly regress the surface normal with fully connected layers, which leads to weak generalization ability and unstable prediction results. To solve the these limitations, Lenssen et al [15] and DeepFit [13] first utilize the network to predict point-wise weights acting as a soft selection of the neighboring points, then estimate the surface normal by the differentiable and weighted least squares plane or polynomial surface fitting.…”

Section: Learning-based Normal Estimationmentioning

confidence: 99%

“…The most direct way for normal estimation is to regress the normal vector from the feature extracted on neighbor- ing points [7,8,9,10,11,12]. However, such brute-force regression only forces the network to memorize normal vectors, which leads to limited generalization ability.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

AdaFit: Rethinking Learning-based Normal Estimation on Point Clouds

Zhu¹,

Liu²,

Dong³

et al. 2021

Preprint

View full text Add to dashboard Cite

This paper presents a neural network for robust normal estimation on point clouds, named AdaFit, that can deal with point clouds with noise and density variations. Existing works use a network to learn point-wise weights for weighted least squares surface fitting to estimate the normals, which has difficulty in finding accurate normals in complex regions or containing noisy points. By analyzing the step of weighted least squares surface fitting, we find that it is hard to determine the polynomial order of the fitting surface and the fitting surface is sensitive to outliers. To address these problems, we propose a simple yet effective solution that adds an additional offset prediction to improve the quality of normal estimation. Furthermore, in order to take advantage of points from different neighborhood sizes, a novel Cascaded Scale Aggregation layer is proposed to help the network predict more accurate point-wise offsets and weights. Extensive experiments demonstrate that AdaFit achieves state-of-the-art performance on both the synthetic PCPNet dataset and the realword SceneNN dataset. The code is publicly available at https://github.com/Runsong123/AdaFit.

show abstract