Search

Due to the complexity of underwater environments, underwater target recognition based on radiated noise has always been challenging. This paper proposes a multi-scale frequency-adaptive network for underwater target recognition. Based on the different distribution densities of Mel filters in the low-frequency band, a three-channel improved Mel energy spectrum feature is designed first. Second, by combining a frequency-adaptive module, an attention mechanism, and a multi-scale fusion module, a multi-scale frequency-adaptive network is proposed to enhance the model’s learning ability. Then, the model training is optimized by introducing a time–frequency mask, a data augmentation strategy involving data confounding, and a focal loss function. Finally, systematic experiments were conducted based on the ShipsEar dataset. The results showed that the recognition accuracy for five categories reached 98.4%, and the accuracy for nine categories in fine-grained recognition was 88.6%. Compared with existing methods, the proposed multi-scale frequency-adaptive network for underwater target recognition has achieved significant performance improvement.

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.

Based on the interference spectrogram in the element–frequency domain using the data measured by the horizontal linear array, the source range can be estimated through the striation-based beamforming (SBF) method and its variants. Estimation of the striation slope is the basis for these ranging methods. But in practical scenarios, the tonal interferences and other noise make it difficult to estimate the slope. In this paper, we proposed a two-dimensional low-pass filtering method after the two-dimensional discrete Fourier transform (2D-DFT) of the element–frequency domain spectrogram. The signals can be separated from the interference and noise using this filter. For the linear frequency-modulated signal without a known waveform, we also proposed an extraction and phase compensation method based on the time–frequency spectrogram, and the acoustic data obtained can be used for source ranging. The experimental results indicate that the methods proposed are feasible.

The element–frequency acoustic intensity of a horizontal line array with a sufficient aperture exhibits interference striation patterns, which can be used for source range estimation without prior environmental information. Under multisource scenarios, the interference striations of the sources overlap with each other, leading to great difficulty in utilizing the information of striations. In this paper, the wavenumber filtering method is applied to each sensor of the horizontal line array to extract the surface-reflected–bottom-reflected modes and reconstruct the recognizable interference spectrogram for each source. Then, via beamforming along the striations, the source ranges can be estimated individually with little prior environmental information and without the long-time observation of moving sources. The required sensor spacing is analyzed, and the spatial filtering capabilities for a single source from different bearings and two sources for which azimuth angles are close to each other have also been investigated. The simulation results indicate that the proposed algorithm can estimate ranges of multiple sources within 25 km, with relative errors of less than 4%.

Extensive research on deep neural networks for LiDAR point clouds has contributed inexhaustible momentum to the development of computer 3D vision applications. However, storage and energy consumption have always been a challenge for deploying these deep models on mobile devices. Quantization provides a feasible route, of which current primary research is focused on uniform bit-width quantization without considering different layers or filters’ sensitivity to different bit-widths. This article proposes a novel hybrid compression method based on relaxed mixed-precision quantization, relaxed weights pruning, and knowledge distillation to overcome the limitations of uniform quantization illustrated above, while further improving model accuracy and reducing model memory consumption. It employs a differentiable searching method to search for the optimal bit allocation and weight sparsity, while conducting feature distillation, accordingly considering the feature degradation by pooling operation in point cloud deep models. The proposed method combines three compression techniques to balance the trade-off between compression accuracy and model size. Pruning alleviates the increasing memory consumption problem caused by mixed-precision quantization, while distillation improves compression accuracy without increasing model size. The experiments validate that the proposed method outperforms state-of-the-art typical uniform quantization methods in terms of accuracy with an acceptable and relatively competitive compression performance.

With the advent of the aging population, fall detection has gradually become a research hotspot. Aiming at the detection of human fall using millimeter-wave radar, a Range-Doppler heat map Sequence detection Network (RDSNet) model that combines the convolutional neural network and long short-term memory network is proposed in this study. First, feature extraction is performed using the convolutional neural network. After obtaining the feature vector, the feature vector corresponding to the dynamic sequence is inputted to the long short-term memory network. Subsequently, the time correlation information of the heat map sequence is learned. Finally, the detection results are obtained using the classifier. Moreover, diverse human movement information of different objects is collected using millimeter-wave radar, and a range-Doppler heat map dataset is built in this work. Comparative experiments show that the proposed RDSNet model can reach an accuracy of 96.67% and the calculation delay is not higher than 50 ms. The proposed RDSNet model has good generalization capabilities and provides new technical ideas for human fall detection and human posture recognition.

Monocular depth estimation techniques are used to recover the distance from the target to the camera plane in an image scene. However, there are still several problems, such as insufficient estimation accuracy, the inaccurate localization of details, and depth discontinuity in planes parallel to the camera plane. To solve these problems, we propose the Global Feature Interaction Network (GFI-Net), which aims to utilize geometric features, such as object locations and vanishing points, on a global scale. In order to capture the interactive information of the width, height, and channel of the feature graph and expand the global information in the network, we designed a global interactive attention mechanism. The global interactive attention mechanism reduces the loss of pixel information and improves the performance of depth estimation. Furthermore, the encoder uses the Transformer to reduce coding losses and improve the accuracy of depth estimation. Finally, a local–global feature fusion module is designed to improve the depth map’s representation of detailed areas. The experimental results on the NYU-Depth-v2 dataset and the KITTI dataset showed that our model achieved state-of-the-art performance with full detail recovery and depth continuation on the same plane.

A balanced dataset is generally beneficial to underwater acoustic target recognition. However, the imbalanced class distribution is always meted out in a real scene. To address this, a weighted cross entropy loss function based on trigonometric function is proposed. Then, the proposed loss function is applied in a multi-scale residual convolutional neural network (named MR-CNN-A network) embedded with an attention mechanism for the recognition task. Firstly, a multi-scale convolution kernel is used to obtain multi-scale features. Then, an attention mechanism is used to fuse these multi-scale feature maps. Furthermore, a cosx-function-weighted cross-entropy loss function is used to deal with the class imbalance in underwater acoustic data. This function adjusts the loss ratio of each sample by adjusting the loss interval of every mini-batch based on cosx term to achieve a balanced total loss for each class. Two imbalanced underwater acoustic data sets, ShipsEar and autonomous underwater vehicle (self-collected data) are used to evaluate the proposed network. The experimental results show that the proposed network outperforms the support vector machine and a simple convolutional neural network. Compared with the other three loss functions, the proposed loss function achieves better stability and adaptability. The results strongly demonstrate the validity of the proposed loss function and the network.

Completing point clouds from partial terrestrial laser scannings (TLS) is a fundamental step for many 3D visual applications, such as remote sensing, digital city and autonomous driving. However, existing methods mainly followed an ordinary auto-encoder architecture with only partial point clouds as inputs, and adopted K-Nearest Neighbors (KNN) operations to extract local geometric features, which takes insufficient advantage of input point clouds and has limited ability to extract features from long-range geometric relationships, respectively. In this paper, we propose a keypoints-aligned siamese (KASiam) network for the completion of partial TLS point clouds. The network follows a novel siamese auto-encoder architecture, to learn prior geometric information of complete shapes by aligning keypoints of complete-partial pairs during the stage of training. Moreover, we propose two essential blocks cross-attention perception (CAP) and self-attention augment (SAA), which replace KNN operations with attention mechanisms and are able to establish long-range geometric relationships among points by selecting neighborhoods adaptively at the global level. Experiments are conducted on widely used benchmarks and several TLS data, which demonstrate that our method outperforms other state-of-the-art methods by a 4.72% reduction of the average Chamfer Distance of categories in PCN dataset at least, and can generate finer shapes of point clouds on partial TLS data.

Golay complementary waveforms are, by definition, able to generate narrow pulses with low sidelobes via coherent signal processing. However, while an ideal impulse can be obtained for target returns at zero-Doppler, significant sidelobes are observed at nonzero Doppler shifts. In this paper, a Generalized Binominal Design (GBD) procedure is proposed for the waveforms consisting of complementary sets in an attempt to reduce the Doppler-induced sidelobes. Our theoretical analysis as well as simulation results show that the proposed approach performs as good as existing Binominal Design method used for sidelobe suppression with Golay complementary waveforms and can also achieve around 28% enhancement on the Doppler resolution, with an acceptable loss in peak to peak-sidelobe ratio (PPSR).

Loop closure detection is an important component of visual simultaneous localization and mapping (SLAM). However, most existing loop closure detection methods are vulnerable to complex environments and use limited information from images. As higher-level image information and multi-information fusion can improve the robustness of place recognition, a semantic–visual–geometric information-based loop closure detection algorithm (SVG-Loop) is proposed in this paper. In detail, to reduce the interference of dynamic features, a semantic bag-of-words model was firstly constructed by connecting visual features with semantic labels. Secondly, in order to improve detection robustness in different scenes, a semantic landmark vector model was designed by encoding the geometric relationship of the semantic graph. Finally, semantic, visual, and geometric information was integrated by fuse calculation of the two modules. Compared with art-of-the-state methods, experiments on the TUM RBG-D dataset, KITTI odometry dataset, and practical environment show that SVG-Loop has advantages in complex environments with varying light, changeable weather, and dynamic interference.

Recognizing 3D objects from point clouds in the presence of significant clutter and occlusion is a highly challenging task. In this paper, we present a coarse-to-fine 3D object recognition algorithm. During the phase of offline training, each model is represented with a set of multi-scale local surface features. During the phase of online recognition, a set of keypoints are first detected from each scene. The local surfaces around these keypoints are further encoded with multi-scale feature descriptors. These scene features are then matched against all model features to generate recognition hypotheses, which include model hypotheses and pose hypotheses. Finally, these hypotheses are verified to produce recognition results. The proposed algorithm was tested on two standard datasets, with rigorous comparisons to the state-of-the-art algorithms. Experimental results show that our algorithm was fully automatic and highly effective. It was also very robust to occlusion and clutter. It achieved the best recognition performance on all of these datasets, showing its superiority compared to existing algorithms.

Three-dimensional (3D) object detection is an important research in 3D computer vision with significant applications in many fields, such as automatic driving, robotics, and human−computer interaction. However, the low precision is an urgent problem in the field of 3D object detection. To solve it, we present a framework for 3D object detection in point cloud. To be specific, a designed Backbone Network is used to make fusion of low-level features and high-level features, which makes full use of various information advantages. Moreover, the two-dimensional (2D) Generalized Intersection over Union is extended to 3D use as part of the loss function in our framework. Empirical experiments of Car, Cyclist, and Pedestrian detection have been conducted respectively on the KITTI benchmark. Experimental results with average precision (AP) have shown the effectiveness of the proposed network.