Point Cloud Sequences Research Articles

Exploring Point-BEV Fusion for 3D Point Cloud Object Tracking With Transformer.

With the prevalent use of LiDAR sensors in autonomous driving, 3D point cloud object tracking has received increasing attention. In a point cloud sequence, 3D object tracking aims to predict the location and orientation of an object in consecutive frames. Motivated by the success of transformers, we propose Point Tracking TRansformer (PTTR), which efficiently predicts high-quality 3D tracking results in a coarse-to-fine manner with the help of transformer operations. PTTR consists of three novel designs. 1) Instead of random sampling, we design Relation-Aware Sampling to preserve relevant points to the given template during subsampling. 2) We propose a Point Relation Transformer for effective feature aggregation and feature matching between the template and search region. 3) Based on the coarse tracking results, we employ a novel Prediction Refinement Module to obtain the final refined prediction through local feature pooling. In addition, motivated by the favorable properties of the Bird's-Eye View (BEV) of point clouds in capturing object motion, we further design a more advanced framework named PTTR++, which incorporates both the point-wise view and BEV representation to exploit their complementary effect in generating high-quality tracking results. PTTR++ substantially boosts the tracking performance on top of PTTR with low computational overhead. Extensive experiments over multiple datasets show that our proposed approaches achieve superior 3D tracking accuracy and efficiency.

Intra-Frame Graph Structure and Inter-Frame Bipartite Graph Matching with ReID-Based Occlusion Resilience for Point Cloud Multi-Object Tracking

Three-dimensional multi-object tracking (MOT) using lidar point cloud data is crucial for applications in autonomous driving, smart cities, and robotic navigation. It involves identifying objects in point cloud sequence data and consistently assigning unique identities to them throughout the sequence. Occlusions can lead to missed detections, resulting in incorrect data associations and ID switches. To address these challenges, we propose a novel point cloud multi-object tracker called GBRTracker. Our method integrates an intra-frame graph structure into the backbone to extract and aggregate spatial neighborhood node features, significantly reducing detection misses. We construct an inter-frame bipartite graph for data association and design a sophisticated cost matrix based on the center, box size, velocity, and heading angle. Using a minimum-cost flow algorithm to achieve globally optimal matching, thereby reducing ID switches. For unmatched detections, we design a motion-based re-identification (ReID) feature embedding module, which uses velocity and the heading angle to calculate similarity and association probability, reconnecting them with their corresponding trajectory IDs or initializing new tracks. Our method maintains high accuracy and reliability, significantly reducing ID switches and trajectory fragmentation, even in challenging scenarios. We validate the effectiveness of GBRTracker through comparative and ablation experiments on the NuScenes and Waymo Open Datasets, demonstrating its superiority over state-of-the-art methods.

A Novel Symmetric Fine-Coarse Neural Network for 3D Human Action Recognition Based on Point Cloud Sequences

Human action recognition has facilitated the development of artificial intelligence devices focusing on human activities and services. This technology has progressed by introducing 3D point clouds derived from depth cameras or radars. However, human behavior is intricate, and the involved point clouds are vast, disordered, and complicated, posing challenges to 3D action recognition. To solve these problems, we propose a Symmetric Fine-coarse Neural Network (SFCNet) that simultaneously analyzes human actions’ appearance and details. Firstly, the point cloud sequences are transformed and voxelized into structured 3D voxel sets. These sets are then augmented with an interval-frequency descriptor to generate 6D features capturing spatiotemporal dynamic information. By evaluating voxel space occupancy using thresholding, we can effectively identify the essential parts. After that, all the voxels with the 6D feature are directed to the global coarse stream, while the voxels within the key parts are routed to the local fine stream. These two streams extract global appearance features and critical body parts by utilizing symmetric PointNet++. Subsequently, attention feature fusion is employed to capture more discriminative motion patterns adaptively. Experiments conducted on public benchmark datasets NTU RGB+D 60 and NTU RGB+D 120 validate SFCNet’s effectiveness and superiority for 3D action recognition.

Optimized quantization parameter selection for video-based point cloud compression

Point clouds are sets of points used to visualize three-dimensional (3D) objects. Point clouds can be static or dynamic. Each point is characterized by its 3D geometry coordinates and attributes such as color. High-quality visualizations often require millions of points, resulting in large storage and transmission costs, especially for dynamic point clouds. To address this problem, the moving picture experts group has recently developed a compression standard for dynamic point clouds called video-based point cloud compression (V-PCC). The standard generates two-dimensional videos from the geometry and color information of the point cloud sequence. Each video is then compressed with a video coder, which converts each frame into frequency coefficients and quantizes them using a quantization parameter (QP). Traditionally, the QPs are severely constrained. For example, in the low-delay configuration of the V-PCC reference software, the quantization parameter values of all the frames in a group of pictures are set to be equal. We show that the rate-distortion performance can be improved by relaxing this constraint and treating the QP selection problem as a multi-variable constrained combinatorial optimization problem, where the variables are the QPs. To solve the optimization problem, we propose a variant of the differential evolution (DE) algorithm. Differential evolution is an evolutionary algorithm that has been successfully applied to various optimization problems. In DE, an initial population of randomly generated candidate solutions is iteratively improved. At each iteration, mutants are generated from the population. Crossover between a mutant and a parent produces offspring. If the performance of the offspring is better than that of the parent, the offspring replaces the parent. While DE was initially introduced for continuous unconstrained optimization problems, we adapt it for our constrained combinatorial optimization problem. Also, unlike standard DE, we apply individual mutation to each variable. Furthermore, we use a variable crossover rate to balance exploration and exploitation. Experimental results for the low-delay configuration of the V-PCC reference software show that our method can reduce the average bitrate by up to 43% compared to a method that uses the same QP values for all frames and selects them according to an interior point method.

AGAR - Attention Graph-RNN for Adaptative Motion Prediction of Point Clouds of Deformable Objects

This article focuses on motion prediction for point cloud sequences in the challenging case of deformable 3D objects, such as human body motion. First, we investigate the challenges caused by deformable shapes and complex motions present in this type of representation, with the ultimate goal of understanding the technical limitations of state-of-the-art models. From this understanding, we propose an improved architecture for point cloud prediction of deformable 3D objects. Specifically, to handle deformable shapes, we propose a graph-based approach that learns and exploits the spatial structure of point clouds to extract more representative features. Then, we propose a module able to combine the learned features in a adaptative manner according to the point cloud movements. The proposed adaptative module controls the composition of local and global motions for each point, enabling the network to model complex motions in deformable 3D objects more effectively. We tested the proposed method on the following datasets: MNIST moving digits, the Mixamo human bodies motions [ 15 ], JPEG [ 5 ] and CWIPC-SXR [ 32 ] real-world dynamic bodies. Simulation results demonstrate that our method outperforms the current baseline methods given its improved ability to model complex movements as well as preserve point cloud shape. Furthermore, we demonstrate the generalizability of the proposed framework for dynamic feature learning by testing the framework for action recognition on the MSRAction3D dataset [ 19 ] and achieving results on par with state-of-the-art methods.

A 4D spontaneous micro‐expression database: Establishment and evaluation

AbstractMicro‐expressions are spontaneous and unconscious facial movements that reveal individuals’ genuine inner emotions. They hold significant potential in various psychological testing fields. As the face is a 3D deformation object, the emergence of facial expression leads to spatial deformation of the face. However, existing databases primarily offer 2D video sequences, limiting descriptions of 3D spatial information related to micro‐expressions. Here, a new micro‐expression database is proposed, which contains 2D image sequences and corresponding 3D point cloud sequences. These samples were classified using both an objective method based on the facial action coding system and a non‐objective emotion classification method that considers video contents and participants’ self‐reports. A variety of feature extraction techniques are applied to 2D data, including traditional algorithms and deep learning methods. Additionally, a novel local curvature‐based algorithm is developed to extract 3D spatio‐temporal deformation features from the 3D data. The authors evaluated the classification accuracies of these two features individually and their fusion results under leave‐one‐subject‐out (LOSO) and tenfold cross‐validation. The results demonstrate that fusing 3D features with 2D features results in improved recognition performance compared to using 2D features alone.

Recognition and optimisation method of impact deformation patterns based on point cloud and deep clustering: Applied to thin-walled tubes

The recognition and clustering of deformation modes are key to constructing impact deformation constraints for thin-walled structures. This paper transforms the clustering and recognition problem of structural impact deformation modes into a problem of clustering and recognition of point cloud sequences based on pseudo-labels. The effectiveness of the method is assessed, and the experimental results show that the accuracy of the proposed method can reach up to 92.17 % when using a pre-training deep neural network feature extractor, which is not only close to the 98.50 % accuracy of supervised learning classification models but also has a 16.84 % improvement in accuracy compared to the deep clustering method based on K-Means. Under different clustering conditions, the proposed method can effectively classify and recognise samples with similar deformation modes and has the ability to summarise and induce new deformation modes when the number of clusters exceeds the number of manual labels. Furthermore, this paper presents a multi-objective optimisation method for structural crashworthiness under impact deformation constraints based on the NSGA-II algorithm. This method constructs impact deformation constraints using surrogate models and deformation clustering and recognition models. The experimental results show that the proposed method can effectively constrain the generation of the population. It is found that there are a large number of Pareto solutions that do not belong to the expected impact deformation mode under the condition of no deformation mode constraint. In contrast, almost all the obtained Pareto solutions conform to the expected impact deformation mode under the condition of deformation mode constraint. In summary, under the condition of impact deformation constraint, the obtained Pareto solutions can satisfy the crashworthiness requirements while conforming to the expected impact deformation mode.

Semantic Complete Scene Forecasting from a 4D Dynamic Point Cloud Sequence

We study a new problem of semantic complete scene forecasting (SCSF) in this work. Given a 4D dynamic point cloud sequence, our goal is to forecast the complete scene corresponding to the future next frame along with its semantic labels. To tackle this challenging problem, we properly model the synergetic relationship between future forecasting and semantic scene completion through a novel network named SCSFNet. SCSFNet leverages a hybrid geometric representation for high-resolution complete scene forecasting. To leverage multi-frame observation as well as the understanding of scene dynamics to ease the completion task, SCSFNet introduces an attention-based skip connection scheme. To ease the need to model occlusion variations and to better focus on the occluded part, SCSFNet utilizes auxiliary visibility grids to guide the forecasting task. To evaluate the effectiveness of SCSFNet, we conduct experiments on various benchmarks including two large-scale indoor benchmarks we contributed and the outdoor SemanticKITTI benchmark. Extensive experiments show SCSFNet outperforms baseline methods on multiple metrics by a large margin, and also prove the synergy between future forecasting and semantic scene completion.The project page with code is available at scsfnet.github.io.

X4D-SceneFormer: Enhanced Scene Understanding on 4D Point Cloud Videos through Cross-Modal Knowledge Transfer

The field of 4D point cloud understanding is rapidly developing with the goal of analyzing dynamic 3D point cloud sequences. However, it remains a challenging task due to the sparsity and lack of texture in point clouds. Moreover, the irregularity of point cloud poses a difficulty in aligning temporal information within video sequences. To address these issues, we propose a novel cross-modal knowledge transfer framework, called X4D-SceneFormer. This framework enhances 4D-Scene understanding by transferring texture priors from RGB sequences using a Transformer architecture with temporal relationship mining. Specifically, the framework is designed with a dual-branch architecture, consisting of an 4D point cloud transformer and a Gradient-aware Image Transformer (GIT). The GIT combines visual texture and temporal correlation features to offer rich semantics and dynamics for better point cloud representation. During training, we employ multiple knowledge transfer techniques, including temporal consistency losses and masked self-attention, to strengthen the knowledge transfer between modalities. This leads to enhanced performance during inference using single-modal 4D point cloud inputs. Extensive experiments demonstrate the superior performance of our framework on various 4D point cloud video understanding tasks, including action recognition, action segmentation and semantic segmentation. The results achieve 1st places, i.e., 85.3% (+7.9%) accuracy and 47.3% (+5.0%) mIoU for 4D action segmentation and semantic segmentation, on the HOI4D challenge, outperforming previous state-of-the-art by a large margin. We release the code at https://github.com/jinglinglingling/X4D.

Frozen CLIP Transformer Is an Efficient Point Cloud Encoder

The pretrain-finetune paradigm has achieved great success in NLP and 2D image fields because of the high-quality representation ability and transferability of their pretrained models. However, pretraining such a strong model is difficult in the 3D point cloud field due to the limited amount of point cloud sequences. This paper introduces Efficient Point Cloud Learning (EPCL), an effective and efficient point cloud learner for directly training high-quality point cloud models with a frozen CLIP transformer. Our EPCL connects the 2D and 3D modalities by semantically aligning the image features and point cloud features without paired 2D-3D data. Specifically, the input point cloud is divided into a series of local patches, which are converted to token embeddings by the designed point cloud tokenizer. These token embeddings are concatenated with a task token and fed into the frozen CLIP transformer to learn point cloud representation. The intuition is that the proposed point cloud tokenizer projects the input point cloud into a unified token space that is similar to the 2D images. Comprehensive experiments on 3D detection, semantic segmentation, classification and few-shot learning demonstrate that the CLIP transformer can serve as an efficient point cloud encoder and our method achieves promising performance on both indoor and outdoor benchmarks. In particular, performance gains brought by our EPCL are 19.7 AP50 on ScanNet V2 detection, 4.4 mIoU on S3DIS segmentation and 1.2 mIoU on SemanticKITTI segmentation compared to contemporary pretrained models. Code is available at \url{https://github.com/XiaoshuiHuang/EPCL}.

MDS-PCGR: A Bimodal Gait Recognition Framework With the Fusion of 4-D Radar Point Cloud Sequences and Micro-Doppler Signatures

mDS-PCGR: A Bimodal Gait Recognition Framework With the Fusion of 4-D Radar Point Cloud Sequences and Micro-Doppler Signatures

A 4D strong spatio-temporal feature learning network for behavior recognition of point cloud sequences

A 4D strong spatio-temporal feature learning network for behavior recognition of point cloud sequences

Language-Guided 3-D Action Feature Learning Without Ground-Truth Sample Class Label.

This work pays the first research effort to leverage point cloud sequence-based Self-supervised 3-D Action Feature Learning (S3AFL), under text's cross-modality weak supervision. We intend to fill the huge performance gap between point cloud sequence and 3-D skeleton-based manners. The key intuition derives from the observation that skeleton-based manners actually hold the human pose's high-level knowledge that leads to attention on the body's joint-aware local parts. Inspired by this, we propose to introduce the text's weak supervision of high-level semantics into a point cloud sequence-based paradigm. With RGB-point cloud pair sequence acquired via RGB-D camera, text sequence is first generated from RGB component using pretrained image captioning model, as auxiliary weak supervision. Then, S3AFL runs in a cross and intra-modality contrastive learning (CL) way. To resist text's missing and redundant semantics, feature learning is conducted in a multistage way with semantic refinement. Essentially, text is only required for training. To facilitate the feature's representation power on fine-grained actions, a multirank max-pooling (MR-MP) way is also proposed for the point set network to better maintain discriminative clues. Experiments verify that the text's weak supervision can facilitate performance by 10.8%, 10.4%, and 8.0% on NTU RGB + D 60, 120, and N-UCLA at most. The performance gap between point cloud sequence and skeleton-based manners has been remarkably narrowed down. The idea of transferring text's weak supervision to S3AFL can also be applied to a skeleton manner, with strong generality. The source code is available at https://github.com/tangent-T/W3AMT.

Dynamic 3D Point Cloud Sequences as 2D Videos.

Dynamic 3D point cloud sequences serve as one of the most common and practical representation modalities of dynamic real-world environments. However, their unstructured nature in both spatial and temporal domains poses significant challenges to effective and efficient processing. Existing deep point cloud sequence modeling approaches imitate the mature 2D video learning mechanisms by developing complex spatio-temporal point neighbor grouping and feature aggregation schemes, often resulting in methods lacking effectiveness, efficiency, and expressive power. In this paper, we propose a novel generic representation called Structured Point Cloud Videos (SPCVs). Intuitively, by leveraging the fact that 3D geometric shapes are essentially 2D manifolds, SPCV re-organizes a point cloud sequence as a 2D video with spatial smoothness and temporal consistency, where the pixel values correspond to the 3D coordinates of points. The structured nature of our SPCV representation allows for the seamless adaptation of well-established 2D image/video techniques, enabling efficient and effective processing and analysis of 3D point cloud sequences. To achieve such re-organization, we design a self-supervised learning pipeline that is geometrically regularized and driven by self-reconstructive and deformation field learning objectives. Additionally, we construct SPCV-based frameworks for both low-level and high-level 3D point cloud sequence processing and analysis tasks, including action recognition, temporal interpolation, and compression. Extensive experiments demonstrate the versatility and superiority of the proposed SPCV, which has the potential to offer new possibilities for deep learning on unstructured 3D point cloud sequences. Code will be released at https://github.com/ZENGYIMING-EAMON/SPCV.

Beyond Pattern Variance: Unsupervised 3-D Action Representation Learning With Point Cloud Sequence.

This work pays the first research effort to address unsupervised 3-D action representation learning with point cloud sequence, which is different from existing unsupervised methods that rely on 3-D skeleton information. Our proposition is built on the state-of-the-art 3-D action descriptor 3-D dynamic voxel (3DV) with contrastive learning (CL). The 3DV can compress the point cloud sequence into a compact point cloud of 3-D motion information. Spatiotemporal data augmentations are conducted on it to drive CL. However, we find that existing CL methods (e.g., SimCLR or MoCo v2) often suffer from high pattern variance toward the augmented 3DV samples from the same action instance, that is, the augmented 3DV samples are still of high feature complementarity after CL, while the complementary discriminative clues within them have not been well exploited yet. To address this, a feature augmentation adapted CL (FACL) approach is proposed, which facilitates 3-D action representation via concerning the features from all augmented 3DV samples jointly, in spirit of feature augmentation. FACL runs in a global-local way: one branch learns global feature that involves the discriminative clues from the raw and augmented 3DV samples, and the other focuses on enhancing the discriminative power of local feature learned from each augmented 3DV sample. The global and local features are fused to characterize 3-D action jointly via concatenation. To fit FACL, a series of spatiotemporal data augmentation approaches is also studied on 3DV. Wide-range experiments verify the superiority of our unsupervised learning method for 3-D action feature learning. It outperforms the state-of-the-art skeleton-based counterparts by 6.4% and 3.6% with the cross-setup and cross-subject test settings on NTU RGB + D 120, respectively. The source code is available at https://github.com/tangent-T/FACL.

MA-ST3D: Motion Associated Self-Training for Unsupervised Domain Adaptation on 3D Object Detection.

Recently, unsupervised domain adaptation (UDA) for 3D object detectors has increasingly garnered attention as a method to eliminate the prohibitive costs associated with generating extensive 3D annotations, which are crucial for effective model training. Self-training (ST) has emerged as a simple and effective technique for UDA. The major issue involved in ST-UDA for 3D object detection is refining the imprecise predictions caused by domain shift and generating accurate pseudo labels as supervisory signals. This study presents a novel ST-UDA framework to generate high-quality pseudo labels by associating predictions of 3D point cloud sequences during ego-motion according to spatial and temporal consistency, named motion-associated self-training for 3D object detection (MA-ST3D). MA-ST3D maintains a global-local pathway (GLP) architecture to generate high-quality pseudo-labels by leveraging both intra-frame and inter-frame consistencies along the spatial dimension of the LiDAR's ego-motion. It also equips two memory modules for both global and local pathways, called global memory and local memory, to suppress the temporal fluctuation of pseudo-labels during self-training iterations. In addition, a motion-aware loss is introduced to impose discriminated regulations on pseudo labels with different motion statuses, which mitigates the harmful spread of false positive pseudo labels. Finally, our method is evaluated on three representative domain adaptation tasks on authoritative 3D benchmark datasets (i.e. Waymo, Kitti, and nuScenes). MA-ST3D achieved SOTA performance on all evaluated UDA settings and even surpassed the weakly supervised DA methods on the Kitti and NuScenes object detection benchmark.

RAPID AND AUTOMATED BODY MEASUREMENT OF CATTLE BASED ON STATISTICAL SHAPE MODEL

Abstract. The current methods of non-contact livestock body measurement directly deal with the low-quality point cloud data of livestock, which have low robustness and lack practicality. On the one hand, the success rate of keypoint detection for livestock body measurement is low. Due to the severe occlusion and noise in the point cloud data, body measurements of some data cannot be performed. On the other hand, the key frames need to be manually selected from the point cloud sequence during processing. Inspired by the work of 3D reconstruction based on animal statistical shape models, we implement the construction and learning of the statistical shape model of real cattle. Given the establishment of the statistical shape model of cattle, a 3D reconstruction and body measurement approach of real cattle based on low-quality point cloud data is proposed. Nine indicators are calculated and the overall estimation MAPE (Mean Absolute Percentage Error) is 10.27%. The whole process of the body measurement algorithm proposed in our paper can be extended to other quadrupeds.

Global Refinement Algorithm for 3D Scene Reconstruction from a Sequence of Point Clouds

Point cloud registration is a central problem in many computer vision problems. However, ensuring global consistency of the results of pairwise registration of point clouds is still a challenge when there are multiple clouds because different scans should be converted to a common coordinate system. This paper describes a global refinement algorithm that first estimates rotations and then estimates parallel translations. For global refinement of rotations, a closed-form algorithm based on matrices is used. For global refinement of parallel translations, a closed-form algorithm is also used. The proposed algorithm is compared with other global refinement algorithms.

Temporal Feature Matching and Propagation for Semantic Segmentation on 3D Point Cloud Sequences

In real-world LiDAR-based applications, data is generated in the form of 3D point cloud sequences or 4D point clouds. However, the topic of semantic segmentation on 4D point clouds is under-investigated and existing methods are still not able to achieve satisfactory performance to meet the requirement for real-world applications. The temporal information across different point clouds plays an important role in dynamic scene understanding, which is not well explored in existing work. In this paper, we focus on exploring effective temporal information across two consecutive point clouds for semantic segmentation on point cloud sequences. To this end, we design three novel modules to enhance the features of target frames by extracting different temporal information in the local regions and global regions. Experimental results on SemanticKITTI and SemanticPOSS demonstrate that our method achieves superior performance in 4D semantic segmentation by utilizing temporal information.

Prediction method of impact deformation mode based on multimodal fusion with point cloud sequences: Applied to thin-walled structures

One of the challenges in solving the structural crashworthiness optimisation problem under impact deformation constraints is to construct an efficient and concise representation method for structural impact deformation. This representation method is expected to include as much information about the structural impact deformation as possible and be computable through numerical simulation, so that it can be combined with the structural impact process. Based on the comparison among several methods, a representation form is proposed using point cloud sequences of finite element nodes in the structural impact process, and the relevant construction method for the point cloud dataset is provided. The dataset includes point clouds derived from the impact deformation and the initial finite element node, structural design variables, and crashworthiness response values of the thin-walled tube during impact process. Based on the constructed dataset, a multi-modal fusion prediction method for the impact process of thin-walled structures is proposed to ensure the efficiency and reliability of crashworthiness prediction with unstructured data. The surrogate model takes structural design variables and initial deformation representations as inputs while impact deformation representations and crashworthiness responses as outputs. The relative errors of the surrogate model's predictions for the crashworthiness responses and impact deformation point cloud are within 5 % of the numerical results. Moreover, the computational speed is 8169 times faster than finite element simulation, ensuring solution accuracy while significantly improving the efficiency of structural crashworthiness prediction. Additionally, the robustness of the surrogate model and the importance of each module are verified and tested through experimental design, providing reference for further improvements to the surrogate model.