Underwater acoustic signal denoising technology aims to overcome the challenge of recovering valuable ship target signals from noisy audios by suppressing underwater background noise. Traditional statistical-based denoising techniques are difficult to be applied effectively in complex underwater environments, especially in the case of extremely low signal-to-noise ratios (SNRs). To address these problems, we propose a noise-aware deep learning model with fullband-subband attention network (NAFSA-Net) for underwater acoustic signal denoising. NAFSA-Net adopts an encoder to extract the feature representation of the input audio. Subsequently, the noise subnet and the target subnet are designed to estimate the noise component and the target component simultaneously. Specifically, some stacked fullband-subband attention (FSA) blocks are deployed in each subnet to capture both global dependencies and fine-grained local dependencies of features. Furthermore, we introduce an interaction module to transmit auxiliary information from the noise subnet to the target subnet. Finally, we propose an improved weight SI-SNR loss function to optimize the training of our model. Experimental results show that our proposed NAFSA-Net substantially outperforms traditional methods and competitive DNN-based solutions in denoising underwater noisy signals with very low SNRs. More importantly, our proposals achieve equally excellent performance on both unseen datasets, which indicates that NAFSA-Net can be a more robust choice for real-world underwater acoustic denoising systems.

We study the problem of efficient object detection of 3D LiDAR point clouds. To reduce the memory and computational cost, existing point-based pipelines usually adopt task-agnostic random sampling or farthest point sampling to progressively downsample input point clouds, despite the fact that not all points are equally important to the task of object detection. In particular, the foreground points are inherently more important than background points for object detectors. Motivated by this, we propose a highly-efficient single-stage point-based 3D detector in this paper, termed IA-SSD. The key of our approach is to exploit two learnable, task-oriented, instance-aware downsampling strategies to hierarchically select the foreground points belonging to objects of interest. Additionally, we also introduce a contextual centroid perception module to further estimate precise instance centers. Finally, we build our IA-SSD following the encoder-only architecture for efficiency. Extensive experiments conducted on several large-scale detection benchmarks demonstrate the competitive performance of our IA-SSD. Thanks to the low memory footprint and a high degree of parallelism, it achieves a superior speed of 80+ frames-per-second on the KITTI dataset with a single RTX2080Ti GPU. The code is available at https://github.com/yifanzhang713/IA-SSD.

Abstract Point cloud completion aims to reconstruct detailed structures from partial observations. However, previous methods often suffered from inaccurate neighbourhood feature extraction and rough reconstruction of complex structures, which is difficult to complete detailed semantic shapes. To solve this problem, we present a method that applies dynamic transformers with adaptive neighbourhood feature fusion operations to resume complete point clouds. Firstly, we propose an adaptive neighbourhood feature extraction module, which contains a learnable global neighbourhood selection strategy and a traditional local k‐nearest neighbour strategy to dynamically select neighbourhood points according to the shape of different objects. Secondly, we observed that traditional point generation methods based on folding‐series operations limit their capacity of generating complex and faithful shapes. Inspired by cell division, we regard the process of reconstructing as point splitting and propose a genetic hierarchical point generation module, which means current points can inherit shape features from previous points and generate more detailed structures of their own. Extensive experiments of point cloud completion are carried out on PCN and Completion3D datasets to verify the effectiveness of our method. The average category chamfer distances of our method are 7.17(×10 −3 ) in PCN and 7.96(×10 −4 ) in Completion3D, which has better completion performance than other methods.

Completing shapes of point clouds from partial scans is a fundamental problem for 3-D vision and remote sensing. However, recent methods mainly relied on K-nearest neighbors (KNN) operations to extract local features of point clouds, which are susceptible to outliers and have limited ability to capture features from long-range context information. In this letter, we propose a new framework with an encoder–decoder architecture, named adaptive global feature augmentation network (AGFA-Net) for point cloud completion. The network mainly consists of spatial and channel attention blocks. Spatial attention blocks are used to replace KNN operations and aggregate global features adaptively by calculating per-point attention values, and channel attention blocks are used to augment useful features of geometric details. Meanwhile, several skip connections are added between different attention blocks to selectively convey geometric features from local regions of partial point clouds to the completion process. Experimental results and analyses demonstrate that our method can generate finer shapes of point clouds and outperforms other state-of-the-art methods under widely used benchmark point completion network (PCN) dataset and several terrestrial laser scanning (TLS) data.

Deep neural networks have achieved great progress in 3D scene understanding. However, recent methods mainly focused on objects with canonical orientations in contrast with random postures in reality. In this letter, we propose a hierarchical neural network, named Local Frame Network (LFNet), based on the local rotation invariant coordinate frame for robust point cloud analysis. The local point patches in different orientated objects are transformed into an identical distribution based on this coordinate frame, and the transformed coordinates are taken as input features to eliminate the influence of rotations at the input level. Meanwhile, a discrete convolution operator is defined in the constructed coordinate frame to extract rotation invariant features from local patches, which can further remove the influence of rotations at the convolution level. Moreover, a Spatial Feature Encoder (SFE) module is utilized to perceive the spatial structure of the local region. Mathematical analysis and experimental results on two public datasets demonstrate that the proposed method can eliminate the influence of rotations without data augmentation and outperforms other state-of-the-art methods.

Semantic segmentation of 3D scenes is a fundamental problem in 3D computer vision. In this paper, we propose a deep neural network for 3D semantic segmentation of raw point clouds. A multi-scale feature learning block is first introduced to obtain informative contextual features in 3D point clouds. A global and local feature aggregation block is then extended to improve the feature learning ability of the network. Based on these strategies, a powerful architecture named 3DMAX-Net is finally provided for semantic segmentation in raw 3D point clouds. Experiments have been conducted on the Stanford large-scale 3D Indoor Spaces Dataset using only geometry information. Experimental results have clearly shown the superiority of the proposed network.

This paper presents a global localization method for mobile robots based on the geometric information of structured indoor environments. With a global/local point cloud and projection map, lines are extracted from the projection maps using Hough transform. According to the directions of the obtained lines, the orientations of projection maps and point clouds are normalized. Next, the template matching algorithm is applied to the normalized global and local projection maps. Once coarse localization is completed, final accurate localization is achieved using the Iterative Closest Points (ICP) algorithm. Experimental results on several point clouds show that the proposed method can achieve high localization accuracy in real-time. The proposed method can be used for other global localization applications in structured environments.

Fully automatic 3-D point cloud registration is a highly challenging task in light detection and ranging (LiDAR) remote sensing. The coherent point drift (CPD) algorithm provides an appropriate solution for point cloud registration because of its high accuracy. However, real application of the traditional CPD algorithm is limited due to its demanding computational complexity. In this letter, we present a novel accelerated CPD (ACPD) algorithm for fast, accurate, and automatic registration of 3-D point clouds. First, a global squared iterative expectation–maximization (gSQUAREM) technique is integrated to the ACPD algorithm. Then, the dual-tree improved fast Gauss transform method is used to further accelerate the Gaussian summation process during the correspondence probability matrix calculation. Experimental results on two real data sets show that the proposed algorithm can perform fast and accurate registration on LiDAR point clouds.