Sparse-to-Dense Multi-Encoder Shape Completion of Unstructured Point Cloud

Point clouds obtained from 3D sensors are usually sparse. Existing methods mainly focus on upsampling sparse point clouds in a supervised manner by using dense ground truth point clouds. In this paper, we propose a self-supervised point cloud upsampling network (SSPU-Net) to generate dense point clouds without using ground truth. To achieve this, we exploit the consistency between the input sparse point cloud and generated dense point cloud for the shapes and rendered images. Specifically, we first propose a neighbor expansion unit (NEU) to upsample the sparse point clouds, where the local geometric structures of the sparse point clouds are exploited to learn weights for point interpolation. Then, we develop a differentiable point cloud rendering unit (DRU) as an end-to-end module in our network to render the point cloud into multi-view images. Finally, we formulate a shape-consistent loss and an image-consistent loss to train the network so that the shapes of the sparse and dense point clouds are as consistent as possible. Extensive results on the CAD and scanned datasets demonstrate that our method can achieve impressive results in a self-supervised manner.

Abstract. Photogrammetric processing algorithms can suffer problems due to either the initial image quality (noise, low radiometric quality, shadows and so on) or to certain surface materials (shiny or textureless objects). This can result in noisy point clouds and/or difficulties in feature extraction. Specifically, dense point clouds which are generated with photogrammetric method using a lightweight thermal camera, are more noisy and sparse than the point clouds of high-resolution digital camera images. In this paper, new method which produces more reliable and dense thermal point cloud using the sparse thermal point cloud and high resolution digital point cloud was considered. Both thermal and digital images were obtained with UAS (Unmanned Aerial System) based lightweight Optris PI 450 and Canon EOS 605D camera images. Thermal and digital point clouds, and orthophotos were produced using photogrammetric methods. Problematic thermal point cloud was transformed to a high density thermal point cloud using image processing methods such as rasterizing, registering, interpolation and filling. The results showed that the obtained thermal point cloud - up to chosen processing parameters - was 87% more densify than the original point cloud. The second improvement was gained at the height accuracy of the thermal point cloud. New densified point cloud has more consistent elevation model while the original thermal point cloud shows serious deviations from the expected surface model.

Photogrammetric processing algorithms can suffer problems due to either the initial image quality (noise, low radiometric quality, shadows and so on) or to certain surface materials (shiny or textureless objects). This can result in noisy point clouds and/or difficulties in feature extraction. Specifically, dense point clouds which are generated with photogrammetric method using a lightweight thermal camera, are more noisy and sparse than the point clouds of high-resolution digital camera images. In this paper, new method which produces more reliable and dense thermal point cloud using the sparse thermal point cloud and high resolution digital point cloud was considered. Both thermal and digital images were obtained with UAS (Unmanned Aerial System) based lightweight Optris PI 450 and Canon EOS 605D camera images. Thermal and digital point clouds, and orthophotos were produced using photogrammetric methods. Problematic thermal point cloud was transformed to a high density thermal point cloud using image processing methods such as rasterizing, registering, interpolation and filling. The results showed that the obtained thermal point cloud - up to chosen processing parameters - was 87% more densify than the original point cloud. The second improvement was gained at the height accuracy of the thermal point cloud. New densified point cloud has more consistent elevation model while the original thermal point cloud shows serious deviations from the expected surface model.

With the widespread use of 3D acquisition devices, there is an increasing need of consolidating captured noisy and sparse point cloud data for accurate representation of the underlying structures. There are numerous algorithms that rely on a variety of assumptions such as local smoothness to tackle this ill‐posed problem. However, such priors lead to loss of important features and geometric detail. Instead, we propose a novel data‐driven approach for point cloud consolidation via a convolutional neural network based technique. Our method takes a sparse and noisy point cloud as input, and produces a dense point cloud accurately representing the underlying surface by resolving ambiguities in geometry. The resulting point set can then be used to reconstruct accurate manifold surfaces and estimate surface properties. To achieve this, we propose a generative neural network architecture that can input and output point clouds, unlocking a powerful set of tools from the deep learning literature. We use this architecture to apply convolutional neural networks to local patches of geometry for high quality and efficient point cloud consolidation. This results in significantly more accurate surfaces, as we illustrate with a diversity of examples and comparisons to the state‐of‐the‐art.

The fusion of laser point cloud and visual image depends on the point cloud density and the target framing effect, the traditional laser point cloud processing for sparse point cloud clustering effect is poor, it is difficult to frame small objects as well as medium and long distance objects. Then the subsequent sensor fusion is easy to miss the recognition of obstacles. In this paper, we improve the frame selection method for sparse point clouds, firstly build a deep learning framework pointpillar, use pointpillar to frame the sparse laser point clouds, then spatially calibrate the lidar coordinate system and camera coordinate system, project the lidar point clouds to the camera image, improve the late fusion method, effectively use the detection results of single sensor, and finally The late-fusion is performed with the target detection results of the camera image to output the exact distance as well as the category of the target. Experiments show that compared with the recognition effect of the traditional fusion algorithm, the number of frames is increased by 6 and the missed recognition rate is reduced from 31.41% to 12.31%.

The widely deployed ways to capture a set of unorganized points, e.g., merged laser scans, fusion of depth images, and structure-from- , usually yield a 3-D noisy point cloud. Accurate normal estimation for the noisy point cloud makes a crucial contribution to the success of various applications. However, the existing normal estimation wisdoms strive to meet a conflicting goal of simultaneously performing normal filtering and preserving surface features, which inevitably leads to inaccurate estimation results. We propose a normal estimation neural network (Norest-Net), which regards normal filtering and feature preservation as two separate tasks, so that each one is specialized rather than traded off. For full noise removal, we present a normal filtering network (NF-Net) branch by learning from the noisy height map descriptor (HMD) of each point to the ground-truth (GT) point normal; for surface feature recovery, we construct a normal refinement network (NR-Net) branch by learning from the bilaterally defiltered point normal descriptor (B-DPND) to the GT point normal. Moreover, NR-Net is detachable to be incorporated into the existing normal estimation methods to boost their performances. Norest-Net shows clear improvements over the state of the arts in both feature preservation and noise robustness on synthetic and real-world captured point clouds.

OctoMap is a popular 3D mapping framework which can model the data consistently and keep the 3D models compact with the octree. However, the occupancy map derived by OctoMap can be incorrect when the input point clouds are with noisy measurements. Point cloud filters can reduce the noisy data, but it is unreasonable to apply filters in a sparse point cloud. In this paper, we present a k-nearest neighbours (k-NN) based inverse sensor model for occupancy mapping. This method represents the occupancy information of one point with the average distance from the point to its k-NN in the point cloud. The average distances derived by all the points and their corresponding k-NN are assumed to be normally distributed. Our inverse sensor model is presented based on this normal distribution. The proposed approach is able to deal with sparse and noisy point clouds. We implement the model in the OctoMap to carry out experiments in the real environment. The experimental results show that the 3D occupancy map generated by our approach is more reliable than that generated by the inverse sensor model in OctoMap.

Three-dimensional (3D) point clouds have a wide range of applications in the field of 3D vision. The quality of the acquired point cloud data considerably impacts the subsequent work of point cloud processing. Due to the sparsity and irregularity of point cloud data, processing point cloud data has always been challenging. However, existing deep learning-based point cloud dense reconstruction methods suffer from excessive smoothing of reconstruction results and too many outliers. The reason for this is that it is not possible to extract features for local and global features at different scales and provide different levels of attention to different regions in order to obtain long-distance dependence for dense reconstruction. In this paper, we use a parallel multi-scale feature extraction module based on graph convolution and an upsampling method with an added multi-head attention mechanism to process sparse and irregular point cloud data to obtain extended point clouds. Specifically, a point cloud training patch with 256 points is inputted. The PMS module uses three residual connections in the multi-scale feature extraction stage. Each PMS module consists of three parallel DenseGCN modules with different size convolution kernels and different averaging pooling sizes. The local and global feature information of the augmented receptive field is extracted efficiently. The scale information is obtained by averaging the different pooled augmented receptive fields. The scale information was obtained using the different average pooled augmented receptive fields. The upsampling stage uses an upsampling rate of r=4, The self-attentive features with a different focus on different point cloud data regions obtained by fusing different weights make the feature representation more diverse. This operation avoids the bias of one attention, and each focuses on extracting valuable fine-grained feature information. Finally, the coordinate reconstruction module obtains 1024 dense point cloud data. Experiments show that the proposed method demonstrates good evaluation metrics and performance and is able to obtain better visual quality. The problems of over-smoothing and excessive outliers are effectively mitigated, and the obtained sparse point cloud is more dense.

High-fidelity point cloud completion with low-resolution recovery and noise-aware upsampling

Three-dimensional (3D) object recognition is crucial for intelligent autonomous agents such as autonomous vehicles and robots alike to operate effectively in unstructured environments. Most state-of-art approaches rely on relatively dense point clouds and performance drops significantly for sparse point clouds. Unsupervised domain adaption allows to minimise the discrepancy between dense and sparse point clouds with minimal unlabelled sparse point clouds, thereby saving additional sparse data collection, annotation and retraining costs. In this work, we propose a novel method for point cloud based object recognition with competitive performance with state-of-art methods on dense and sparse point clouds while being trained only with dense point clouds.

The point cloud data from actual measurements are often sparse and incomplete, making it difficult to apply them directly to visual processing and 3D reconstruction. The point cloud completion task can predict missing parts based on a sparse and incomplete point cloud model. However, the disordered and unstructured characteristics of point clouds make it difficult for neural networks to obtain detailed spatial structures and topological relationships, resulting in a challenging point cloud completion task. Existing point cloud completion methods can only predict the rough geometry of the point cloud, but cannot accurately predict the local details. To address the shortcomings of existing point cloud complementation methods, this paper describes a novel network for adaptive point cloud growth, MAPGNet, which generates a sparse skeletal point cloud using the skeletal features in the composite encoder, and then adaptively grows the local point cloud in the spherical neighborhood of each point using the growth features to complement the details of the point cloud in two steps. In this paper, the Offset Transformer module is added in the process of complementation to enhance the contextual connection between point clouds. As a result, MAPGNet improves the quality of the generated point clouds and recovers more local detail information. Comparing our algorithm with other state-of-the-art algorithms in different datasets, experimental results show that our algorithm has advantages in dense point cloud completion.

Compared with 3D mesh data, unstructured point cloud data lack adjacency relationship between points, which only contain geometric coordinates and little information. This paper focuses on the research of characteristics of unstructured point cloud detection algorithm. We put forward the multiscale 3D Harris feature point detection algorithm, which uses iteration strategy to select the optimal Harris response value in multiple scales. Compared with the classical 3D Harris feature point detection algorithm for mesh data, our algorithm can fully use the local information of point cloud models to detect feature point on point cloud models. It is very robust to rotation transformation of point clouds and noise.

In this study, we propose a high-throughput and low-cost automatic detection method based on deep learning to replace the inefficient manual counting of rapeseed siliques. First, a video is captured with a smartphone around the rapeseed plants in the silique stage. Feature point detection and matching based on SIFT operators are applied to the extracted video frames, and sparse point clouds are recovered using epipolar geometry and triangulation principles. The depth map is obtained by calculating the disparity of the matched images, and the dense point cloud is fused. The plant model of the whole rapeseed plant in the silique stage is reconstructed based on the structure-from-motion (SfM) algorithm, and the background is removed by using the passthrough filter. The downsampled 3D point cloud data is processed by the DGCNN network, and the point cloud is divided into two categories: sparse rapeseed canopy siliques and rapeseed stems. The sparse canopy siliques are then segmented from the original whole rapeseed siliques point cloud using the sparse-dense point cloud mapping method, which can effectively save running time and improve efficiency. Finally, Euclidean clustering segmentation is performed on the rapeseed canopy siliques, and the RANSAC algorithm is used to perform line segmentation on the connected siliques after clustering, obtaining the three-dimensional spatial position of each silique and counting the number of siliques. The proposed method was applied to identify 1457 siliques from 12 rapeseed plants, and the experimental results showed a recognition accuracy greater than 97.80%. The proposed method achieved good results in rapeseed silique recognition and provided a useful example for the application of deep learning networks in dense 3D point cloud segmentation.

Finding high-level semantic information from a point cloud is a challenging task, and it can be used in various applications. For instance, it is useful to compactly represent the scene structure and efficiently understand the scene context. This task is even more challenging when using a hand-held monocular visual SLAM system that outputs a noisy sparse point cloud. In order to tackle this issue, we propose an incremental primitive modeling method using both geometric and statistical analyses for such point cloud. The main idea is to select only reliably-modeled shapes by analyzing the geometric relationship between the point cloud and the estimated shapes. Besides that, a statistical evaluation is incorporated to filter wrongly-detected primitives in a noisy point cloud. As a result of this processing, our approach largely improved precision when compared with state of the art methods. We also show the impact of segmenting and representing a scene using primitives instead of a point cloud.

In the study of Simultaneous Localization and Mapping (SLAM), the existence of dynamic obstacles will have a great impact on it, and when there are many dynamic obstacles, it will lead to great challenges in mapping. Therefore, segmenting dynamic objects in the environment is particularly important. The common data format in the field of autonomous robots is point clouds. How to use point clouds to segment dynamic objects is the focus of this study. The existing point clouds instance segmentation methods are mostly based on dense point clouds. In our application scenario, we use 16-line LiDAR (sparse point clouds) and propose a sparse point clouds instance segmentation method based on spatio-temporal encoding and decoding for autonomous robots in dynamic environments. Compared with other point clouds instance segmentation methods, the proposed algorithm has significantly improved average percision and average recall on instance segmentation of our point clouds dataset. In addition, the annotation of point clouds is time-consuming and laborious, and the existing dataset for point clouds instance segmentation is also very limited. Thus, we propose an autonomous point clouds annotation algorithm that integrates object tracking, segmentation, and point clouds to 2D mapping methods, the resulting data can then be used for training robust model.