Point Cloud Classification Using Content-Based Transformer via Clustering in Feature Space

Recently, there have been some attempts of Transformer in 3D point cloud classification. In order to reduce computations, most existing methods focus on local spatial attention,but ignore their content and fail to establish relationships between distant but relevant points. To overcome the limitation of local spatial attention, we propose a point content-based Transformer architecture, called PointConT for short. It exploits the locality of points in the feature space(content-based), which clusters the sampled points with similar features into the same class and computes the self-attention within each class, thus enabling an effective trade-off between capturing long-range dependencies and computational complexity. We further introduce an inception feature aggregator for point cloud classification, which uses parallel structures to aggregate high-frequency and low-frequency information in each branch separately. Extensive experiments show that our PointConT model achieves a remarkable performance on point cloud shape classification. Especially, our method exhibits 90.3% Top-1 accuracy on the hardest setting of ScanObjectN N. Source code of this paper is available at https://github.com/yahuiliu99/PointC onT.

Automatic Extraction Method of 3D Feature Guidelines for Complex Cultural Relic Surfaces Based on Point Cloud

Cultural relics line graphic serves as a crucial form of traditional artifact information documentation,which is a simple and intuitive product with low cost of displaying compared with 3D models.Dimensionality reduction is undoubtedly necessary for line drawings.However,most existing methods for artifact drawing rely on the principles of orthographic projection that always cannot avoid angle occlusion and data overlapping while the surface of cultural relics is complex.Therefore,conformal mapping was introduced as a dimensionality reduction way to compensate for the limitation of orthographic projection.Based on the given criteria for assessing surface complexity,this paper proposed a three-dimensional feature guideline extraction method for complex cultural relic surfaces.A 2D and 3D combined factor that measured the importance of points on describing surface features,vertex weight,was designed.Then the selection threshold for feature guideline extraction was determined based on the differences between vertex weight and shape index distributions.The feasibility and stability were verified through experiments conducted on real cultural relic surface data.Results demonstrated the ability of the method to address the challenges associated with the automatic generation of line drawings for complex surfaces.The extraction method and the obtained results will be useful for line graphic drawing,displaying and propaganda of cultural relics.

Building Facade Point Clouds Segmentation Based on Optimal Dual-Scale Feature Descriptors

To address the current issues of inaccurate segmentation and the limited applicability of segmentation methods for building facades in point clouds, we propose a facade segmentation algorithm based on optimal dual-scale feature descriptors. First, we select the optimal dual-scale descriptors from a range of feature descriptors. Next, we segment the facade according to the threshold value of the chosen optimal dual-scale descriptors. Finally, we use RANSAC (Random Sample Consensus) to fit the segmented surface and optimize the fitting result. Experimental results show that, compared to commonly used facade segmentation algorithms, the proposed method yields more accurate segmentation results, providing a robust data foundation for subsequent 3D model reconstruction of buildings.

Research on color image matching method based on feature point compensation in dark light environment

Image matching refers to the process of matching two or more images obtained at different time,different sensors or different conditions through a large number of feature points in the image.At present,image matching is widely used in target recognition and tracking,indoor positioning and navigation.Local features missing,however,often occurs in color images taken in dark light,making the extracted feature points greatly reduced in number,so as to affect image matching and even fail the target recognition.An unsharp masking(USM)based denoising model is established and a local adaptive enhancement algorithm is proposed to achieve feature point compensation by strengthening local features of the dark image in order to increase amount of image information effectively.Fast library for approximate nearest neighbors(FLANN)and random sample consensus(RANSAC)are image matching algorithms.Experimental results show that the number of effective feature points obtained by the proposed algorithm from images in dark light environment is increased,and the accuracy of image matching can be improved obviously.

Image Relaxation Matching Based on Feature Points for DSM Generation

In photogrammetry and remote sensing, image matching is a basic and crucial process for automatic DEM generation. In this paper we presented a image relaxation matching method based on feature points. This method can be considered as an extention of regular grid point based matching. It avoids the shortcome of grid point based matching. For example, with this method, we can avoid low or even no texture area where errors frequently appear in cross correlaton matching. In the mean while, it makes full use of some mature techniques such as probability relaxation, image pyramid and the like which have already been successfully used in grid point matching process. Application of the technique to DEM generaton in different regions proved that it is more reasonable and reliable.

Effective Self-calibration for Camera Parameters and Hand-eye Geometry Based on Two Feature Points Motions

A novel and effective self-calibration approach for robot vision is presented, which can effectively estimate both the camera intrinsic parameters and the hand-eye transformation at the same time. The proposed calibration procedure is based on two arbitrary feature points of the environment, and three pure translational motions and two rotational motions of robot endeffector are needed. New linear solution equations are deduced, and the calibration parameters are finally solved accurately and effectively. The proposed algorithm has been verified by simulated data with different noise and disturbance. Because of the need of fewer feature points and robot motions, the proposed method greatly improves the efficiency and practicality of the calibration procedure. © 2017 Chinese Association of Automation.

STATISTICAL PROBABILITY BASED ALGORITHM FOR EXTRACTING FEATURE POINTS IN 2-DIMENSIONAL IMAGE

An algorithm for automatically extracting feature points is developed after the area of feature points in 2-dimensional (2D) imagebeing located by probability theory, correlated methods and criterion for abnormity. Feature points in 2D image can be extracted only by calculating standard deviation of gray within sampled pixels area in our approach statically. While extracting feature points, the limitation to confirm threshold by tentative method according to some a priori information on processing image can be avoided. It is proved that the proposed algorithm is valid and reliable by extracting feature points on actual natural images with abundant and weak texture, including multi-object with complex background, respectively. It can meet the demand of extracting feature points of 2D image automatically in machine vision system.

Arm-Root Curve Fitting Based on Body Surface Feature Points for Young Male

This paper is aiming to obtain an arm-root curve function performing the human arm-root size and shape realistically. A gypsum replica of upper arm for young male was made and scanned for extracting the 3D coordinates of 4 feature points of shoulder point, the anterior/posterior armpit point and the axillary point describing the real arm-root shape under the normalized definitions, and the 5 landmarks were confirmed additionally for improving the fitting precision. Then, the wholly and piecewise fitting of arm-root curve with 9 feature points and mark points in total were generated respectively based on least square polynomial fitting method. Comparing to the wholly fitting, the piecewise fitted function segmented by the line between anterior and posterior axillary points showed a high fitting degree of arm-root morphology with R-square of 1, the length difference between fitted curve and gypsum curve is 0.003 cm within error range. And it provided a basic curve model with standard feature points to simulate arm-root morphology realistically by curve fitting for accurate body measurement extraction.

基于PointNet++的邻域特征增强点云语义分割方法

随着智能驾驶、机器人导航等以点云为基础的应用蓬勃发展,点云语义分割逐渐成为研究热点。然而,现有的点云语义分割方法存在局部特征提取不充分、特征融合不完整的缺陷。针对这些不足,提出了对应的解决方案。对于局部特征提取不充分的现象,通过嵌入邻域点的坐标、方向、距离等相关信息去关联邻域点的显式特征。对于特征融合不完整的现象,提出了一种最大池化与自注意力池化相结合的混合池化方法。网络架构基于PointNet++,并结合提出的局部特征提取和融合方法,在S3DIS数据集上的实验结果表明,与基线方法PointNet++相比,各评价指标都有不同程度的提高,证实了新方法的有效性和优越性。

Observation points classifier ensemble for high-dimensional imbalanced classification

In this paper,an Observation Points Classifier Ensemble(OPCE)algorithm is proposed to deal with High-Dimensional Imbalanced Classification(HDIC)problems based on data processed using the Multi-Dimensional Scaling(MDS)feature extraction technique.First,dimensionality of the original imbalanced data is reduced using MDS so that distances between any two different samples are preserved as well as possible.Second,a novel OPCE algorithm is applied to classify imbalanced samples by placing optimised observation points in a low-dimensional data space.Third,optimization of the observation point mappings is carried out to obtain a reliable assessment of the unknown samples.Exhaustive experiments have been conducted to evaluate the feasibility,rationality,and effectiveness of the proposed OPCE algorithm using seven benchmark HDIC data sets.Experimental results show that(1)the OPCE algorithm can be trained faster on low-dimensional imbalanced data than on high-dimensional data;(2)the OPCE algorithm can correctly identify samples as the number of optimised observation points is increased;and(3)statistical analysis reveals that OPCE yields better HDIC performances on the selected data sets in comparison with eight other HDIC algorithms.This demonstrates that OPCE is a viable algorithm to deal with HDIC problems.

A Hybrid Features Based Detection Method for Inshore Ship Targets in SAR Imagery

Convolutional Neural Networks(CNNs)have recently attracted much attention in the ship detection from Synthetic Aperture Radar(SAR)images.However,compared with optical images,SAR ones are hard to understand.Moreover,due to the high similarity between the man-made targets near shore and inshore ships,the classical methods are unable to achieve effective detection of inshore ships.To mitigate the influence of onshore ship-like objects,this paper proposes an inshore ship detection method in SAR images by using hybrid features.Firstly,the sea-land segmentation is applied in the pre-processing to exclude obvious land regions from SAR images.Then,a CNN model is designed to extract deep features for identifying potential ship targets in both inshore and offshore water.On this basis,the high-energy point number of amplitude spectrum is further introduced as an important and delicate feature to suppress false alarms left.Finally,to verify the effectiveness of the proposed method,numerical and comparative studies are carried out in experiments on Sentinel-1 SAR images.

Robust Radiometric Normalization of the near Equatorial Satellite Images Using Feature Extraction and Remote Sensing Analysis

Relative radiometric normalization (RRN) minimizes radiometric differences among images caused by inconsistencies of acquisition conditions rather than changes in surface. Scale invariant feature transform (SIFT) has the ability to automatically extract control points (CPs) and is commonly used for remote sensing images. However, its results are mostly inaccurate and sometimes contain incorrect matching caused by generating a small number of false CP pairs. These CP pairs have high false alarm matching. This paper presents a modified method to improve the performance of SIFT CPs matching by applying sum of absolute difference (SAD) in a different manner for the new optical satellite generation called near-equatorial orbit satellite and multi-sensor images. The proposed method, which has a significantly high rate of correct matches, improves CP matching. The data in this study were obtained from the RazakSAT satellite a new near equatorial satellite system. The proposed method involves six steps: 1) data reduction, 2) applying the SIFT to automatically extract CPs, 3) refining CPs matching by using SAD algorithm with empirical threshold, and 4) calculation of true CPs intensity values over all image’ bands, 5) preforming a linear regression model between the intensity values of CPs locate in reverence and sensed image’ bands, 6) Relative radiometric normalization conducting using regression transformation functions. Different thresholds have experimentally tested and used in conducting this study (50 and 70), by followed the proposed method, and it removed the false extracted SIFT CPs to be from 775, 1125, 883, 804, 883 and 681 false pairs to 342, 424, 547, 706, 547, and 469 corrected and matched pairs, respectively.

Automatic Feature Point Detection and Tracking of Human Actions in Time-of-flight Videos

Detecting feature points on the human body in video frames is a key step for tracking human movements. There have been methods developed that leverage models of human pose and classification of pixels of the body image. Yet, occlusion and robustness are still open challenges. In this paper, we present an automatic, model-free feature point detection and action tracking method using a time-of-flight camera. Our method automatically detects feature points for movement abstraction. To overcome errors caused by miss-detection and occlusion, a refinement method is devised that uses the trajectory of the feature points to correct the erroneous detections. Experiments were conducted using videos acquired with a Microsoft Kinect camera and a publicly available video set and comparisons were conducted with the state-of-the-art methods. The results demonstrated that our proposed method delivered improved and reliable performance with an average accuracy in the range of 90 %.The trajectorybased refinement also demonstrated satisfactory effectiveness that recovers the detection with a success rate of 93.7 %. Our method processed a frame in an average time of 71.1 ms.

Image Feature Extraction and Matching of Augmented Solar Images in Space Weather

Augmented solar images were used to research the adaptability of four representative image extraction and matching algorithms in space weather domain.These include the scale-invariant feature transform algorithm,speeded-up robust features algorithm,binary robust invariant scalable keypoints algorithm,and oriented fast and rotated brief algorithm.The performance of these algorithms was estimated in terms of matching accuracy,feature point richness,and running time.The experiment result showed that no algorithm achieved high accuracy while keeping low running time,and all algorithms are not suitable for image feature extraction and matching of augmented solar images.To solve this problem,an improved method was proposed by using two-frame matching to utilize the accuracy advantage of the scale-invariant feature transform algorithm and the speed advantage of the oriented fast and rotated brief algorithm.Furthermore,our method and the four representative algorithms were applied to augmented solar images.Our application experiments proved that our method achieved a similar high recognition rate to the scale-invariant feature transform algorithm which is significantly higher than other algorithms.Our method also obtained a similar low running time to the oriented fast and rotated brief algorithm,which is significantly lower than other algorithms.

FeatureMatching Combining Variable Velocity Model with Reverse Optical Flow

The ORB-SLAM2 based on the constant velocity model is difficult to determine the search window of the reprojection of map points when the objects are in variable velocity motion,which leads to a false matching,with an inaccurate pose estimation or failed tracking.To address the challenge above,a new method of feature point matching is proposed in this paper,which combines the variable velocity model with the reverse optical flow method.First,the constant velocity model is extended to a new variable velocity model,and the expanded variable velocity model is used to provide the initial pixel shifting for the reverse optical flow method.Then the search range of feature points is accurately determined according to the results of the reverse optical flow method,thereby improving the accuracy and reliability of feature matching,with strengthened interframe tracking effects.Finally,we tested on TUM data set based on the RGB-D camera.Experimental results show that this method can reduce the probability of tracking failure and improve localization accuracy on SLAM(Simultaneous Localization and Mapping)systems.Compared with the traditional ORB-SLAM2,the test error of this method on each sequence in the TUM data set is significantly reduced,and the root mean square error is only 63.8%of the original system under the optimal condition.

Approach for Predicting the Knowledge Points of Math Questions

In order to provide high-quality learning services,various online systems should possess the fundamental ability to predict the knowledge points and units to which a given test question belongs.The existing methods typically rely on manual labeling or traditional machine learning methods.Manual labeling methods have high time costs and high demands for human resources,while traditional machine learning methods only focus on the shallow features of the topics,ignoring the deep semantic relationship between the topic text and the knowledge point units.These two methods have relatively large limitations in practical applications.This paper proposes a convolutional neural network method combined with multiple features to predict the knowledge point units.We construct a binary classification dataset in the three grades of primary mathematics.Considering the supplementary role of Pinyin to Chinese text and the unique identification characteristics of Unicode encoding for characters,we obtain the Pinyin representation and the Unicode encoding representation of the original Chinese text.Then,we put the three representation methods into the convolutional neural network for training,obtain three kinds of semantic vectors,fuse them,and finally obtain higher-dimensional fusion features.Our experimental results demonstrate that our approach achieves good performance in predicting the knowledge units of test questions.

A Machine Learning-Based Distributed Denial of Service Detection Approach for Early Warning in Internet Exchange Points

The Internet service provider(ISP)is the heart of any country’s Internet infrastructure and plays an important role in connecting to theWorld WideWeb.Internet exchange point(IXP)allows the interconnection of two or more separate network infrastructures.All Internet traffic entering a country should pass through its IXP.Thus,it is an ideal location for performing malicious traffic analysis.Distributed denial of service(DDoS)attacks are becoming a more serious daily threat.Malicious actors in DDoS attacks control numerous infected machines known as botnets.Botnets are used to send numerous fake requests to overwhelm the resources of victims and make them unavailable for some periods.To date,such attacks present a major devastating security threat on the Internet.This paper proposes an effective and efficient machine learning(ML)-based DDoS detection approach for the early warning and protection of the Saudi Arabia Internet exchange point(SAIXP)platform.The effectiveness and efficiency of the proposed approach are verified by selecting an accurate ML method with a small number of input features.A chi-square method is used for feature selection because it is easier to compute than other methods,and it does not require any assumption about feature distribution values.Several ML methods are assessed using holdout and 10-fold tests on a public large-size dataset.The experiments showed that the performance of the decision tree(DT)classifier achieved a high accuracy result(99.98%)with a small number of features(10 features).The experimental results confirmthe applicability of using DT and chi-square for DDoS detection and early warning in SAIXP.

Point Reg Net: Invariant Features for Point Cloud Registration Using in Image-Guided Radiation Therapy

In image-guided radiation therapy, extracting features from medical point cloud is the key technique for multimodality registration. This novel framework, denoted Control Point Net (CPN), provides an alternative to the common applications of manually designed keypoint descriptors for coarse point cloud registration. The CPN directly consumes a point cloud, divides it into equally spaced 3D voxels and transforms the points within each voxel into a unified feature representation through voxel feature encoding (VFE) layer. Then all volumetric representations are aggregated by Weighted Extraction Layer which selectively extracts features and synthesize into global descriptors and coordinates of control points. Utilizing global descriptors instead of local features allows the available geometrical data to be better exploited to improve the robustness and precision. Specifically, CPN unifies feature extraction and clustering into a single network, omitting time-consuming feature matching procedure. The algorithm is tested on point cloud datasets generated from CT images. Experiments and comparisons with the state-of-the-art descriptors demonstrate that CPN is highly discriminative, efficient, and robust to noise and density changes.

A Novel Airborne 3D Laser Point Cloud Hole Repair Algorithm Considering Topographic Features

Hole repair processing is an important part of point cloud data processing in airborne 3-dimensional(3D)laser scanning technology.Due to the fragmentation and irregularity of the surface morphology,when applying the 3D laser scanning technology to mountain mapping,the conventional mathematical cloud-based point cloud hole repair method is not ideal in practical applications.In order to solve this problem,we propose to repair the valley and ridge line first,and then repair the point cloud hole.The main technical steps of the method include the following points:First,the valley and ridge feature lines are extracted by the GIS slope analysis method;Then,the valley and ridge line missing from the hole are repaired by the mathematical interpolation method,and the repaired results are edited and inserted to the original point cloud;Finally,the traditional repair method is used to repair the point cloud hole whose valley line and ridge line have been repaired.Three experiments were designed and implemented in the east bank of the Xiaobaini River to test the performance of the proposed method.The results showed that compared with the direct point cloud hole repair method in Geomagic Studio software,the average repair accuracy of the proposed method,in the 16 m buffer zone of valley line and ridge line,is increased from 56.31 cm to 31.49 cm.The repair performance is significantly improved.

基于改进PointPillars的3D目标检测算法

针对自动驾驶场景下远距离及遮挡目标识别问题,对PointPillars算法进行了改进。引入了并行的空间注意力和通道注意力机制,增强了目标的位置信息及有用特征通道权重,提高了远距离目标的检测精度。在2D CNN骨干网络中引入了自适应空间特征融合模块,解决了特征拼接的信息丢失问题,提高了遮挡目标的检测精度。基于KITTI数据集在3种不同场景难度下分别对SECOND、PointPillars、改进PointPillars这3种算法进行了定量分析验证,并将改进的PointPillars算法进行可视化分析。定量分析表明,改进PointPillars算法在鸟瞰图模式下目标检测精度最大提升2.75%;在三维模式下目标检测精度最大提升2.93%;在AOS模式下目标检测精度最大提升4.05%,可视化结果表明改进的PointPillars算法能有效检测远距离及遮挡目标。