Search

Air-conditioning systems account for 40–60% of the energy consumption of buildings, and most of this figure corresponds to the cooling and dehumidification process of air-conditioning units. Compared with traditional compressed air-conditioning systems, solid adsorption dehumidification systems possess good potential to improve indoor air quality and reduce buildings’ energy consumption. However, there are still some problems that prevent the use of solid adsorption dehumidification systems, such as their complexity and high regeneration temperatures. The key to solving the above problem is the discovery of more suitable adsorbents. In this paper, poly N-isopropylacrylamide/polypyrrole (PNIPAm/PPy) hydrogel was selected as the research object, and the performance of the dehumidification material and its potential for application in solid dehumidification systems were studied. It was found that the pore structure of PNIPAm/PPy was relatively complex and that there were abundant pores with uneven pore sizes. The minimum pore size was about 4 μm, while the maximum pore size was about 25 μm, and the pore sizes were mostly distributed between 8 and 20 μm. Abundant and dense pores ensure good hygroscopic and water-releasing properties of the resulting hydrogel. The PPy inside the hydrogel acts as both a hygroscopic and photothermal agent. In an environment with a relative humidity of 90%, 60%, and 50%, the hygroscopic efficiency of PNIPAm/PPy reached 80% in about 75, 100, and 120 min, and the corresponding unit equilibrium hygroscopic capacity values were 3.85 g/g, 3.72 g/g, and 3.71 g/g, respectively. In the initial stage, the moisture absorption increased with the increase in time; then, the increase in moisture absorption decreased. When the temperature was below 40 °C, the hygroscopic performance of PNIPAm/PPy was almost temperature-independent. The PNIPAm/PPy with different thicknesses showed similar moisture absorption efficiency. The lowest desorption temperature of PNIPAm/PPy was 40 °C, which indicates that low-grade energy can be used for material desorption. And the higher the temperature, the faster the desorption rate of PNIPAm/PPy and the higher the desorption amount. It can be seen that the PNIPAm/PPy hydrogel presents good desorption performance and can be used repeatedly.

Objectives Electroporation, the breakdown of the biomembrane induced by external electric fields, has increasingly become a research hotspot for its promising related methods in various kinds of cancers. Content In this article, we utilized CiteSpace 6.1.R2 to perform a bibliometric analysis on the research foundation and frontier of electroporation-based applications in cancer therapy. A total of 3,966 bibliographic records were retrieved from the Web of Science Core Collection for the bibliometric analysis. Sersa G. and Mir L. M. are the most indispensable researchers in this field, and the University of Ljubljana of Slovenia is a prominent institution. By analyzing references and keywords, we found that, with a lower recurrence rate, fewer severe adverse events, and a higher success rate, irreversible electroporation, gene electrotransfer, and electrochemotherapy are the three main research directions that may influence the future treatment protocol of cancers. Summary This article visualized relevant data to synthesize scientific research on electroporation-based cancer therapy, providing helpful suggestions for further investigations on electroporation. Outlook Although electroporation-based technologies have been proven as promising tools for cancer treatment, its radical mechanism is still opaque and their commercialization and universalization need further efforts from peers.

Neural networks have dominated the research of hyperspectral image classification, attributing to the feature learning capacity of convolution operations. However, the fixed geometric structure of convolution kernels hinders long-range interaction between features from distant locations. In this article, we propose a novel spectral-spatial transformer network (SSTN), which consists of spatial attention and spectral association modules, to overcome the constraints of convolution kernels. Also, we design a factorized architecture search (FAS) framework that involves two independent subprocedures to determine the layer-level operation choices and block-level orders of SSTN. Unlike conventional neural architecture search (NAS) that requires a bilevel optimization of both network parameters and architecture settings, the FAS focuses only on finding out optimal architecture settings to enable a stable and fast architecture search. Extensive experiments conducted on five popular HSI benchmarks demonstrate the versatility of SSTNs over other state-of-the-art (SOTA) methods and justify the FAS strategy. On the University of Houston dataset, SSTN obtains comparable overall accuracy to SOTA methods with a small fraction (1.2%) of multiply-and-accumulate operations compared to a strong baseline spectral-spatial residual network (SSRN). Most importantly, SSTNs outperform other SOTA networks using only 1.2% or fewer MACs of SSRNs on the Indian Pines, the Kennedy Space Center, the University of Pavia, and the Pavia Center datasets.

Recent geometric deep learning works define convolution operations in local regions and have enjoyed remarkable success on non-Euclidean data, including graph and point clouds. However, the high-level geometric correlations between the input and its neighboring coordinates or features are not fully exploited, resulting in suboptimal segmentation performance. In this article, we propose a novel graph convolution architecture, which we term as Taylor Gaussian mixture model (GMM) network (TGNet), to efficiently learn expressive and compositional local geometric features from point clouds. The TGNet is composed of basic geometric units, TGConv, that conduct local convolution on irregular point sets and are parametrized by a family of filters. Specifically, these filters are defined as the products of the local point features and the neighboring geometric features extracted from local coordinates. These geometric features are expressed by Gaussian weighted Taylor kernels. Then, a parametric pooling layer aggregates TGConv features to generate new feature vectors for each point. TGNet employs TGConv on multiscale neighborhoods to extract coarse-to-fine semantic deep features while improving its scale invariance. Additionally, a conditional random field (CRF) is adopted within the output layer to further improve the segmentation results. Using three point cloud data sets, qualitative and quantitative experimental results demonstrate that the proposed method achieves 62.2% average accuracy on ScanNet, 57.8% and 68.17% mean intersection over union (mIoU) on Stanford Large-Scale 3D Indoor Spaces (S3DIS) and Paris-Lille-3D data sets, respectively.

Recently, the advancement of deep learning in discriminative feature learning from 3D LiDAR data has led to rapid development in the field of autonomous driving. However, automated processing uneven, unstructured, noisy, and massive 3D point clouds is a challenging and tedious task. In this paper, we provide a systematic review of existing compelling deep learning architectures applied in LiDAR point clouds, detailing for specific tasks in autonomous driving such as segmentation, detection, and classification. Although several published research papers focus on specific topics in computer vision for autonomous vehicles, to date, no general survey on deep learning applied in LiDAR point clouds for autonomous vehicles exists. Thus, the goal of this paper is to narrow the gap in this topic. More than 140 key contributions in the recent five years are summarized in this survey, including the milestone 3D deep architectures, the remarkable deep learning applications in 3D semantic segmentation, object detection, and classification; specific datasets, evaluation metrics, and the state of the art performance. Finally, we conclude the remaining challenges and future researches., 21 pages, submitted to IEEE Transactions on Neural Networks and Learning Systems

In this paper, we address the hyperspectral image (HSI) classification task with a generative adversarial network and conditional random field (GAN-CRF) -based framework, which integrates a semi-supervised deep learning and a probabilistic graphical model, and make three contributions. First, we design four types of convolutional and transposed convolutional layers that consider the characteristics of HSIs to help with extracting discriminative features from limited numbers of labeled HSI samples. Second, we construct semi-supervised GANs to alleviate the shortage of training samples by adding labels to them and implicitly reconstructing real HSI data distribution through adversarial training. Third, we build dense conditional random fields (CRFs) on top of the random variables that are initialized to the softmax predictions of the trained GANs and are conditioned on HSIs to refine classification maps. This semi-supervised framework leverages the merits of discriminative and generative models through a game-theoretical approach. Moreover, even though we used very small numbers of labeled training HSI samples from the two most challenging and extensively studied datasets, the experimental results demonstrated that spectral-spatial GAN-CRF (SS-GAN-CRF) models achieved top-ranking accuracy for semi-supervised HSI classification., Comment: Accepted by IEEE T-CYB

The recent integration of attention mechanisms into segmentation networks improves their representational capabilities through a great emphasis on more informative features. However, these attention mechanisms ignore an implicit sub-task of semantic segmentation and are constrained by the grid structure of convolution kernels. In this paper, we propose a novel squeeze-and-attention network (SANet) architecture that leverages an effective squeeze-and-attention (SA) module to account for two distinctive characteristics of segmentation: i) pixel-group attention, and ii) pixel-wise prediction. Specifically, the proposed SA modules impose pixel-group attention on conventional convolution by introducing an 'attention' convolutional channel, thus taking into account spatial-channel inter-dependencies in an efficient manner. The final segmentation results are produced by merging outputs from four hierarchical stages of a SANet to integrate multi-scale contexts for obtaining an enhanced pixel-wise prediction. Empirical experiments on two challenging public datasets validate the effectiveness of the proposed SANets, which achieves 83.2% mIoU (without COCO pre-training) on PASCAL VOC and a state-of-the-art mIoU of 54.4% on PASCAL Context., Comment: Accepted to CVPR 2020

The prevailing framework consisted of complex feature extractors following by conventional classifiers. Nevertheless, the high spatial and high spectral dimensionality of each pixel in the hyperspectral imagery hinders the development of hyperspectral image classification. Fortunately, since 2012, deep learning models, which can extract the hierarchical features of large amounts of daily three-channel optical images, have emerged as a better alternative to their shallow learning counterparts. Within all deep learning models, convolutional neural networks (CNNs) exhibit convincing and stunning ability to process a huge mass of data. In this paper, the CNNs have been adopted as an end-to-end pixelwise scheme to classify the pixels of hyperspectral imagery, in which each pixel contains hundreds of continuous spectral bands. According to the preliminarily qualitative and quantitative results, the existing CNN models achieve promising classification accuracy and process effectively and robustly on the University of Pavia dataset.

Deep neural networks can learn deep feature representation for hyperspectral image (HSI) interpretation and achieve high classification accuracy in different datasets. However, counterintuitively, the classification performance of deep learning models degrades as their depth increases. Therefore, we add identity mappings to convolutional neural networks for every two convolutional layers to build deep residual networks (ResNets). To study the influence of deep learning model size on HSI classification accuracy, this paper applied ResNets and CNNs with different depth and width using two challenging datasets. Moreover, we tested the effectiveness of batch normalization as a regularization method with different model settings. The experimental results demonstrate that ResNets mitigate the declining-accuracy effect and achieved promising classification performance with 10% and 5% training sample percentages for the University of Pavia and Indian Pines datasets, respectively. In addition, t-Distributed Stochastic Neighbor Embedding (t-SNE) provides a direct view of the extracted features through dimensionality reduction.

The main research direction of intelligent control technology is analyzed and discussed in detail, including fuzzy model and basic control algorithm, fuzzy predictive control, stability analysis of fuzzy control system. The status and function of intelligent control are discussed, and the application and development of intelligent control in medicine are summarized and prospected in this paper.

Available big geoscientific data and modern powerful computation hardware have laid a solid foundation for the prevailing deep learning models in the field of image classification, detection and segmentation. In these models, fully convolutional networks achieve unprecedented success in image segmentation tasks [6]. In this paper, we apply the contemporary image segmentation models in the context of extracting buildings and roads from high spatial resolution imagery. We estimate the influence of filter stride, learning rate, input data size, training epoch and fine-tuning on model performance. Selected Massachusetts road and building datasets are used for training, validation, and testing the performance of the models with different parameters. As a result of combining shallow fine-grained pooling layer outputs with the deep final-score layer or abandoning coarse-grained pooling layers, the extraction precision rate of the best modified model improves significantly to over 78%.

Unmanned Aircraft System (UAS), a low cost and time efficient remote sensing platform, can help public agencies obtain useful urban aerial imagery of cities and update land use information frequently, especially in natural colour bands. In this paper, a decision tree based land use classification approach using optical aerial imagery is proposed. First, land cover information is extracted through the Maximum Likelihood Classifier and tabulated with an Ownership parcel map. Second, a decision tree is generated to establish the relationship between land cover and land use. Taking advantage of the geometric characteristics of parcels, an organized land use parcel map is produced. Afterwards, by resampling the aerial imagery from 20 cm, 50 cm and 100 cm resolution, effects of spatial resolution in this classification approach are discussed and determined. This land use classification method is flexible and can be widely used in urban planning and landscape monitoring.