PLA Information Engineering University

Point cloud learning has lately attracted increasing attention due to its wide applications in many areas, such as computer vision, autonomous driving, and robotics. As a dominating technique in AI, deep learning has been successfully used to solve various 2D vision problems. However, deep learning on point clouds is still in its infancy due to the unique challenges faced by the processing of point clouds with deep neural networks. Recently, deep learning on point clouds has become even thriving, with numerous methods being proposed to address different problems in this area. To stimulate future research, this paper presents a comprehensive review of recent progress in deep learning methods for point clouds. It covers three major tasks, including 3D shape classification, 3D object detection and tracking, and 3D point cloud segmentation. It also presents comparative results on several publicly available datasets, together with insightful observations and inspiring future research directions.

Abstract. Zoning which is to divide the study area into different zones according to their geographical differences at the global, national or regional level, includes natural division, economic division, geographical zoning of departments, comprehensive zoning and so on. Zoning is of important practical significance, for example, knowing regional differences and characteristics, regional research and regional development planning, understanding the favorable and unfavorable conditions of the regional development etc. Geographical environment is arising from the geographical position linkages. Geographical environment unit division is also a type of zoning. The geographical environment indicators are deeply studied and summed up in the article, including the background, the associated and the potential. The background indicators are divided into four categories, such as the socio-economic, the political and military, the strategic resources and the ecological environment, which can be divided into more sub-indexes. While the sub-indexes can be integrated to comprehensive index system by weighted stacking method. The Jenks natural breaks classification method, also called the Jenks optimization method, is a data classification method designed to determine the best arrangement of values into different classes. This is done by seeking to minimize each class’s average deviation from the class mean, while maximizing each class’s deviation from the means of the other groups. In this paper, the experiment of Chinese surrounding geographical environment unit division has been done based on the natural breaks (jenks) method, the geographical environment index system and the weighted stacking method, taking South Asia as an example. The result indicates that natural breaks (jenks) method is of good adaptability and high accuracy on the geographical environment unit division. The geographical environment research was originated in the geopolitics and flourished in the geo-economics. The main representatives of the geopolitics are German geographer Friedrich Ratzel, British geographer Mackinder and American geographical politician Nicholas John Spykman etc. The main representative of the geo-economics is American geographical economist Edward Luttwak. China has the most neighboring countries in the world, and its geographical environment is extremely complex. With the continuous development of globalization, China's relations with neighboring countries have become more complex and more closely. So it is very meaningful to have depth research on geographical environment unit division of China.

Deep learning methods have recently been successfully explored for hyperspectral image (HSI) classification. However, training a deep-learning classifier notoriously requires hundreds or thousands of labeled samples. In this paper, a deep few-shot learning method is proposed to address the small sample size problem of HSI classification. There are three novel strategies in the proposed algorithm. First, spectral–spatial features are extracted to reduce the labeling uncertainty via a deep residual 3-D convolutional neural network. Second, the network is trained by episodes to learn a metric space where samples from the same class are close and those from different classes are far. Finally, the testing samples are classified by a nearest neighbor classifier in the learned metric space. The key idea is that the designed network learns a metric space from the training data set. Furthermore, such metric space could generalize to the classes of the testing data set. Note that the classes of the testing data set are not seen in the training data set. Four widely used HSI data sets were used to assess the performance of the proposed algorithm. The experimental results indicate that the proposed method can achieve better classification accuracy than the conventional semisupervised methods with only a few labeled samples.

China's BeiDou navigation system (BDS) has evolved from the demonstration navigation satellite system (BDS-1) to the regional navigation satellite system (BDS-2). Now, the global BeiDou navigation system (BDS-3) is in construction and is proceeding well. The design and functions of BDS-3 are quite different from those of both BDS-1 and BDS-2. In this paper, the general design, the coordinate reference system, and the system time basis of BDS-3 are introduced. Several new payloads designed to accomplish different objectives are described as well as the platforms on which they are hosted. Since BDS-3 consists of several different constellations, the general service capabilities and special service functions provided by these different constellations are described. The performances of the initial BDS-3 platforms are evaluated based on the available eight-medium Earth orbit (MEO) satellite configuration. The results of satellite orbit determination and prediction with and without the BDS-3 inter-satellite links (ISL) are compared and analyzed.

As one of the four global satellite navigation systems, Compass not only enhances satellite visibility and availability for positioning, navigation and timing (PNT) for users in China and the surrounding areas, but also improves PNT precision for global users. The improvements in satellite visibility and the dilution of precision are analyzed under GNSS compatibility and interoperation conditions. The contribution of the Compass satellite navigation system to global users, especially the benefits that users can acquire from the combination of Compass, GPS, GLONASS, and Galileo navigation systems, is analyzed using simulation data.

Semantic change detection (SCD) extends the multiclass change detection (MCD) task to provide not only the change locations but also the detailed land-cover/land-use (LCLU) categories before and after the observation intervals. This fine-grained semantic change information is very useful in many applications. Recent studies indicate that the SCD can be modeled through a triple-branch convolutional neural network (CNN), which contains two temporal branches and a change branch. However, in this architecture, the communications between the temporal branches and the change branch are insufficient. To overcome the limitations in existing methods, we propose a novel CNN architecture for the SCD, where the semantic temporal features are merged in a deep CD unit. Furthermore, we elaborate on this architecture to reason the bi-temporal semantic correlations. The resulting bi-temporal semantic reasoning network (Bi-SRNet) contains two types of semantic reasoning blocks to reason both single-temporal and cross-temporal semantic correlations, as well as a novel loss function to improve the semantic consistency of change detection results. Experimental results on a benchmark dataset show that the proposed architecture obtains significant accuracy improvements over the existing approaches, while the added designs in the Bi-SRNet further improve the segmentation of both semantic categories and the changed areas. The codes in this article are accessible at <uri>https://github.com/ggsDing/Bi-SRNet</uri>.

In this study, we develop a novel deep hierarchical vision transformer (DHViT) architecture for hyperspectral and light detection and ranging (LiDAR) data joint classification. Current classification methods have limitations in heterogeneous feature representation and information fusion of multi-modality remote sensing data (e.g., hyperspectral and LiDAR data), these shortcomings restrict the collaborative classification accuracy of remote sensing data. The proposed deep hierarchical vision transformer architecture utilizes both the powerful modeling capability of long-range dependencies and strong generalization ability across different domains of the transformer network, which is based exclusively on the self-attention mechanism. Specifically, the spectral sequence transformer is exploited to handle the long-range dependencies along the spectral dimension from hyperspectral images, because all diagnostic spectral bands contribute to the land cover classification. Thereafter, we utilize the spatial hierarchical transformer structure to extract hierarchical spatial features from hyperspectral and LiDAR data, which are also crucial for classification. Furthermore, the cross attention (CA) feature fusion pattern could adaptively and dynamically fuse heterogeneous features from multi-modality data, and this contextual aware fusion mode further improves the collaborative classification performance. Comparative experiments and ablation studies are conducted on three benchmark hyperspectral and LiDAR datasets, and the DHViT model could yield an average overall classification accuracy of 99.58%, 99.55%, and 96.40% on three datasets, respectively, which sufficiently certify the effectiveness and superior performance of the proposed method.

In the modern metropolis, the coverage of community hospitals is getting larger and larger. But high-level hospitals treating serious and sudden illness are still kind of scarce social resources. Shenzhen is the forefront of China's reform and opening up, its solution to the problem of medical allocation can be used as a reference for other cities. In this paper, we assume that primary, secondary and tertiary hospitals are main places for treating major diseases in Shenzhen, and analyze each sub-districts' spatial accessibility to them based on travel time and spatial difference within the city. A kernel density two-step floating catchment area method (KD2SFCA) is used to calculate how many medical resources each sub-district could share, in which the travel time from residential district to hospitals is used as an important parameter in evaluating the accessibility between suppliers and demanders. According to statistics of travel time by both driving and public transportation, the impedance function in KD2SFCA is modified and actual data of Shenzhen is used to fit its parameters, which could better simulate the attenuation trend of hospital service capabilities over the travel time. The spatial accessibilities are calculated under different travel modes and multiple time thresholds, and then spatial autocorrelation analysis method is used to analyze the spatial correlation of residents' accessibility to these medical resources. From both analysis of spatial accessibility and spatial autocorrelation, distinct variations in spatial distribution of high-level hospitals could be observed: of which the south and central parts of Shenzhen have significantly higher accessibilities, while the eastern and western regions are relatively lower. In particular, 12 sub-districts gain quite lower scores than others, showing that constructions on high-level hospitals should be strengthened further. In conclusion, the spatial configuration of high level hospitals in Shenzhen is not well balanced, Further optimization is urgently needed.

In a typical large-scale data center, a set of applications are hosted over virtual machines (VMs) running on a large number of physical machines (PMs). Such a virtualization technique can be used for conserving power consumption by minimizing the number of PMs that should be turned on according to the application requirements to resource. However, the resource demands for VMs is dynamic in nature since the number of user requests the applications should handle is rapidly changing in practice. It is a great challenge to online reconfigure the VMs (i.e., optimize the number and the locations for the VMs) according to the dynamic resource demands. Especially for the emerging applications of large-scale data centers for cloud computing systems, existing approaches either fails to find a best configuration of VMs or cannot produce a result in an acceptable time. In this paper, we propose an online self-reconfiguration approach for reallocating VMs in large-scale data centers. It first accurately predicts the future workloads of the applications with Brown's quadratic exponential smoothing. Based on such a prediction, it adopts a genetic algorithm to efficiently find the optimal reconfiguration policy. The resource utilization of large-scale cloud computing data centers can thus be improved and their energy consumption can be greatly conserved. We conduct extensive experiments and the results verify that our approach can effectively switch off more unnecessary running PMs comparing with current approaches without a performance degradation of the whole system.

Security issues have resulted in severe damage to the cloud computing environment, adversely affecting the healthy and sustainable development of cloud computing. Intrusion detection is one of the technologies for protecting the cloud computing environment from malicious attacks. However, network traffic in the cloud computing environment is characterized by large scale, high dimensionality, and high redundancy, these characteristics pose serious challenges to the development of cloud intrusion detection systems. Deep learning technology has shown considerable potential for intrusion detection. Therefore, this study aims to use deep learning to extract essential feature representations automatically and realize high detection performance efficiently. An effective stacked contractive autoencoder (SCAE) method is presented for unsupervised feature extraction. By using the SCAE method, better and robust low-dimensional features can be automatically learned from raw network traffic. A novel cloud intrusion detection system is designed on the basis of the SCAE and support vector machine (SVM) classification algorithm. The SCAE+SVM approach combines both deep and shallow learning techniques, and it fully exploits their advantages to significantly reduce the analytical overhead. Experiments show that the proposed SCAE+SVM method achieves higher detection performance compared to three other state-of-the-art methods on two well-known intrusion detection evaluation datasets, namely KDD Cup 99 and NSL-KDD.

High-frequency electroencephalography (EEG) signals play an important role in research on human emotions. However, the different network patterns under different emotional states in the high gamma band (50-80 Hz) remain unclear. In this paper, we investigate different emotional states using functional network analysis on various frequency bands. We constructed multiple functional networks on different frequency bands and performed functional network analysis and time-frequency analysis on these frequency bands to determine the significant features that represent different emotional states. Furthermore, we verified the effectiveness of these features by using them in emotion recognition. Our experimental results revealed that the network connections in the high gamma band with significant differences among the positive, neutral, and negative emotional states were much denser than the network connections in the other frequency bands. The connections mainly occurred in the left prefrontal, left temporal, parietal, and occipital regions. Moreover, long-distance connections with significant differences among the emotional states were observed in the high frequency bands, particularly in the high gamma band. Additionally, high gamma band fusion features derived from the global efficiency, network connections, and differential entropies achieved the highest classification accuracies for both our dataset and the public dataset. These results are consistent with literature and provide further evidence that high gamma band EEG signals are more sensitive and effective than the EEG signals in other frequency bands in studying human affective perception.

In recent years, deep-learning-based hyperspectral image (HSI) classification methods have achieved significant development. The superior capability of feature extraction from these data-driven methods dramatically improves the classification performance. However, the previous methods usually require to retrain the network from scratch to obtain the capability of feature extraction adaptive for the target image when facing a new HSI to be classified, which is a time-consuming and redundant process. In this paper, we consider putting this process ahead and making the network have a robust capability of feature extraction with generalization through pre-training. Therefore, the network enables to directly extract features of the target HSI without re-training. For this purpose, we rethink the three-dimension (3D) HSI data from a perspective of spectral sequence, and we attempt to extract the spectral variation information as the spectrum motion feature. Then, we construct an unsupervised spectrum motion feature learning framework (SMF-UL), which can be pre-trained on mass unlabeled HSI data to learn the knowledge about perceiving spectral variation. Furthermore, to achieve the expansion of source data for pre-training, we develop an extendable training dataset construction method, which can integrate HSIs of different sizes, number of bands and sensors into a unified training set to utilize the rapidly growing mass unlabeled HSI data effectively. Finally, we use the trained network to directly extract the spectrum motion feature of the target HSI for classification, so the laborious re-training of the network can be avoided. Extensive experiments show that the proposed SMF-UL acquires the robust capability of feature extraction with generalization through unsupervised learning on mass unlabeled HSI data, and the classification performance of extracted spectrum motion feature is competitive to advanced in-domain and cross-domain methods, which shows its flexibility and superiority. The code of SMF-UL will be open at: https://github.com/sssssyf/SMF-UL.

Deep learning has achieved great success in hyperspectral image classification. However, when processing new hyperspectral images, the existing deep learning models must be retrained from scratch with sufficient samples, which is inefficient and undesirable in practical tasks. This paper aims to explore how to accurately classify new hyperspectral images with only a few labeled samples, i.e., the hyperspectral images few-shot classification. Specifically, we design a new deep classification model based on relational network and train it with the idea of meta-learning. Firstly, the feature learning module and the relation learning module of the model can make full use of the spatial–spectral information in hyperspectral images and carry out relation learning by comparing the similarity between samples. Secondly, the task-based learning strategy can enable the model to continuously enhance its ability to learn how to learn with a large number of tasks randomly generated from different data sets. Benefitting from the above two points, the proposed method has excellent generalization ability and can obtain satisfactory classification results with only a few labeled samples. In order to verify the performance of the proposed method, experiments were carried out on three public data sets. The results indicate that the proposed method can achieve better classification results than the traditional semisupervised support vector machine and semisupervised deep learning models.

After years of in-depth development of blockchain, various blockchains with different characteristics and suitable for different application scenarios coexist in large numbers. Due to the isolation of blockchains and the high degree of heterogeneity between chains, value transfer and data communication between existing blockchains are facing unprecedented challenges, and the phenomenon of value isolated island is gradually emerging. The cross-chain technology of blockchain is an important technical means to realize the interconnection of blockchains and improve the interoperability and scalability of blockchains. In this paper, the development and application of blockchain cross-chain technology are studied, the background and significance of cross-chain technology are described, the research status of cross-chain technology is expounded, the current mainstream cross-chain technologies and cross-chain projects are introduced, the mentioned cross-chain technologies and cross-chain projects are analyzed and compared. In addition, this paper also summarizes the difficulties existing in the current cross-chain technology and provides solutions for reference, so as to lead to the discussion of the development trend of cross-chain technology, and finally complete the summary of the research content of the full text and the prospect of cross-chain technology. It is hoped that the relevant summary results can help relevant researchers and practitioners quickly grasp the research progress in the field of blockchain interoperability, and obtain relevant knowledge and application methods in this field.

With the development of data mining, machine learning offers opportunities to improve discrimination by analyzing complex interactions among massive variables. To test the ability of machine learning algorithms for predicting risk of type 2 diabetes mellitus (T2DM) in a rural Chinese population, we focus on a total of 36,652 eligible participants from the Henan Rural Cohort Study. Risk assessment models for T2DM were developed using six machine learning algorithms, including logistic regression (LR), classification and regression tree (CART), artificial neural networks (ANN), support vector machine (SVM), random forest (RF) and gradient boosting machine (GBM). The model performance was measured in an area under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, negative predictive value and area under precision recall curve. The importance of variables was identified based on each classifier and the shapley additive explanations approach. Using all available variables, all models for predicting risk of T2DM demonstrated strong predictive performance, with AUCs ranging between 0.811 and 0.872 using laboratory data and from 0.767 to 0.817 without laboratory data. Among them, the GBM model performed best (AUC: 0.872 with laboratory data and 0.817 without laboratory data). Performance of models plateaued when introduced 30 variables to each model except CART model. Among the top-10 variables across all methods were sweet flavor, urine glucose, age, heart rate, creatinine, waist circumference, uric acid, pulse pressure, insulin, and hypertension. New important risk factors (urinary indicators, sweet flavor) were not found in previous risk prediction methods, but determined by machine learning in our study. Through the results, machine learning methods showed competence in predicting risk of T2DM, leading to greater insights on disease risk factors with no priori assumption of causality.

Vision Foundation Models (VFMs) such as the Segment Anything Model (SAM) allow zero-shot or interactive segmentation of visual contents, thus they are quickly applied in a variety of visual scenes. However, their direct use in many Remote Sensing (RS) applications is often unsatisfactory due to the special imaging properties of RS images. In this work, we aim to utilize the strong visual recognition capabilities of VFMs to improve change detection (CD) in very high-resolution (VHR) remote sensing images (RSIs). We employ the visual encoder of FastSAM, a variant of the SAM, to extract visual representations in RS scenes. To adapt FastSAM to focus on some specific ground objects in RS scenes, we propose a convolutional adaptor to aggregate the task-oriented change information. Moreover, to utilize the semantic representations that are inherent to SAM features, we introduce a task-agnostic semantic learning branch to model the semantic latent in bi-temporal RSIs. The resulting method, SAM-CD, obtains superior accuracy compared to the SOTA fully-supervised CD methods and exhibits a sample-efficient learning ability that is comparable to semi-supervised CD methods. To the best of our knowledge, this is the first work that adapts VFMs to CD in VHR RS images.

Long-range contextual information is crucial for the semantic segmentation of high-resolution (HR) remote sensing images (RSIs). However, image cropping operations, commonly used for training neural networks, limit the perception of long-range contexts in large RSIs. To overcome this limitation, we propose a wide-context network (WiCoNet) for the semantic segmentation of HR RSIs. Apart from extracting local features with a conventional convolutional neural network (CNN), the WiCoNet has an extra context branch to aggregate information from a larger image area. Moreover, we introduce a context transformer to embed contextual information from the context branch and selectively project it onto the local features. The context transformer extends the vision transformer, an emerging kind of neural networks, to model the dual-branch semantic correlations. It overcomes the locality limitation of CNNs and enables the WiCoNet to see the bigger picture before segmenting the land-cover/land-use (LCLU) classes. Ablation studies and comparative experiments conducted on several benchmark datasets demonstrate the effectiveness of the proposed method. In addition, we present a new Beijing Land-Use (BLU) dataset. This is a large-scale HR satellite dataset with high-quality and fine-grained reference labels, which can facilitate future studies in this field.

Real-time functional magnetic resonance imaging neurofeedback (rtfMRI-NF) is a prospective tool to enhance the emotion regulation capability of participants and to alleviate their emotional disorders. The hippocampus is a key brain region in the emotional brain network and plays a significant role in social cognition and emotion processing in the brain. However, few studies have focused on the emotion NF of the hippocampus. This study investigated the feasibility of NF training of healthy participants to self-regulate the activation of the hippocampus and assessed the effect of rtfMRI-NF on the hippocampus before and after training. Twenty-six right-handed healthy volunteers were randomly assigned to the experimental group receiving hippocampal rtfMRI-NF (n = 13) and the control group (CG) receiving rtfMRI-NF from the intraparietal sulcus rtfMRI-NF (n = 13) and completed a total of 4 NF runs. The hippocampus and the intraparietal sulcus were defined based on the Montreal Neurological Institute (MNI) standard template, and NF signal was measured as a percent signal change relative to the baseline obtained by averaging the fMRI signal for the preceding 20 s long rest block. NF signal (percent signal change) was updated every 2 s and was displayed on the screen. The amplitude of low-frequency fluctuation and regional homogeneity values were calculated to evaluate the effects of NF on spontaneous neural activity in resting state fMRI. A standard general linear model (GLM) analysis was separately conducted for each fMRI NF run. Results showed that the activation of hippocampus increased after 4 NF training runs. The hippocampal activity of the experiment group participants was higher than that of the CG. They also showed elevated hippocampal activity and the greater amygdala-hippocampus connectivity. The anterior temporal lobe, parahippocampal gyrus, hippocampus, and amygdala of brain regions associated with emotional processing were activated during training. We presented a proof-of-concept study using rtfMRI-NF for hippocampus up-regulation in the recall of positive autobiographical memories. The current study may provide a new method to regulate our emotions and can potentially be applied to the clinical treatment of emotional disorders.

Recently, the field of hyperspectral image (HSI) classification is dominated by deep learning-based methods. However, training deep learning models usually needs a large number of labeled samples to optimize thousands of parameters. In this article, a deep multiview learning method is proposed to deal with the small sample problem of HSI. First, two views of an HSI scene are constructed by applying principal component analysis to different bands. Second, a deep residual network is designed to embed the different views of a sample to a latent space. The designed deep residual network is trained by maximizing agreement between differently augmented views of the same data sample via a contrastive loss in the latent space. Note that the training procedure of the designed deep residual network does not use labeled information. Therefore, the proposed method belongs to the category of unsupervised learning, which could alleviate the lack of labeled training samples. Finally, a conventional machine learning method (e.g., support vector machine) is used to complete the classification task in the learned latent space. To demonstrate the effectiveness of the proposed method, extensive experiments are carried on four widely used hyperspectral data sets. The experimental results demonstrate that the proposed method could improve the classification accuracy with small samples.

Personal e-health records (EHR) enable medical workers (e.g., doctors and nurses) to conveniently and quickly access each patient's medical history through the public cloud, which greatly facilitates patients' visits and makes telemedicine possible. Additionally, since EHR involve patients' personal privacy information, EHR holders would hesitate to directly outsource their data to cloud servers. A natural and favorite manner of conquering this issue is to encrypt these outsourced EHR such that only authorized medical workers can access them. Specifically, the ciphertext-policy attribute-based encryption (CP-ABE) supports fine-grained access over encrypted data and is considered to be a perfect solution of securely sharing EHR in the public cloud. In this paper, to strengthen the system security and meet the requirement of specific applications, we add functionalities of user revocation, secret key delegation and ciphertext update to the original ABE, and propose a revocable-storage hierarchical attribute-based encryption (RS-HABE) scheme, as the core building of establishing a framework for secure sharing of EHR in public cloud. The proposed RS-HABE scheme features of forward security (a revoked user can no longer access previously encrypted data) and backward security (a revoked user also cannot access subsequently encrypted data) simultaneously, and is proved to be selectively secure under a complexity assumption in bilinear groups, without random oracles. The theoretical analysis indicates that the proposed scheme surpasses existing similar works in terms of functionality and security, at the acceptable cost of computation overhead. Moreover, we implement the proposed scheme and present experiments to demonstrate its practicability.