Abstract Ship classification is an important technique for enhancing maritime management and security. Visible and infrared sensors are generally employed to deal with the challenging problem and improve classification performance. Herein, a two‐branch feature fusion neural network structure is proposed to classify the visible and infrared maritime vessel images simultaneously. Specifically, in this two‐branch neural network, one branch is based on a deep convolutional neural network that is used to extract the visible image features, while the other is a hybrid network structure that is a multi‐scale patch embedding network called MPANet. The sub‐network MPANet can extract fine‐ and coarse‐grained features, in which the pooling operation instead of the multi‐head attention mechanism is utilized to reduce memory consumption. When there are infrared images, it is used to extract the infrared image features, otherwise, this branch is also utilized to extract visible image features. Therefore, this dual network is suitable with or without infrared images. The experimental results on the visible and infrared spectrums (VAIS) dataset demonstrate that the introduced network achieves state‐of‐the‐art ship classification performance on visible images and paired visible and infrared ship images.

Federated learning can effectively ensure data security and improve the problem of data islanding. However, the performance of federated learning-based schemes could be better due to the imbalance of image data. Therefore, this paper proposes a federated learning approach based on priori knowledge and a bilateral segmentation network for image edge extraction. First, federated learning can distribute training images for some special complex images due to the small sample and unshared data. Then, the image with similar edge information to the original image is learned to obtain prior knowledge, and the local uniform sparsity method is used to strengthen the detail features and weaken the background features. Based on the bilateral segmentation network, we introduce a dilated pyramid pooling layer and multi-scale feature fusion module to fuse the shallow detailed features in the context path with the deep abstract features obtained through the dilated pyramid pooling. The final result is obtained by fusing the result with prior knowledge and the result with the context path. Finally, we conduct experiments on some public datasets, and the results show that the proposed method greatly improves extraction accuracy compared with the traditional and the most advanced methods.

Person search is a challenging retrieval problem which aims at matching pedestrians with the same identity over non-overlapping camera views. In this paper, we adopt Swin Transformer as the backbone network to extract discriminative features. We propose a symmetric online instance matching loss which transfers the symmetric idea from KL divergence to the online instance matching loss. The purpose is to strengthen the robustness of the person search model under the condition of limited training identities. We compared with the state-of-the-arts on two mainstream benchmarks: CUHK-SYSU and PRW datasets. Experimental results demonstrate the effectiveness of our method. Especially, we achieve better performance on the PRW dataset with an improvement of 6.5% and 3.5% at the mAP and top-1 accuracy, respectively.

Long non-coding RNAs (lncRNAs) are key regulators of major biological processes and their functional modes are dictated by their subcellular localization. Relative nuclear enrichment of lncRNAs compared to mRNAs is a prevalent phenomenon but the molecular mechanisms governing their nuclear retention in cells remain largely unknown. Here in this study, we harness the recently released eCLIP data for a large number of RNA-binding proteins (RBPs) in K562 and HepG2 cells and utilize multiple bioinformatics methods to comprehensively survey the roles of RBPs in lncRNA nuclear retention. We identify an array of splicing RBPs that bind to nuclear-enriched lincRNAs (large intergenic non-coding RNAs) thus may act as trans-factors regulating their nuclear retention. Further analyses reveal that these RBPs may bind with distinct core motifs, flanking sequence compositions, or secondary structures to drive lincRNA nuclear retention. Moreover, network analyses uncover potential co-regulatory RBP clusters and the physical interaction between HNRNPU and SAFB2 proteins in K562 cells is further experimentally verified. Altogether, our analyses reveal previously unknown factors and mechanisms that govern lincRNA nuclear localization in cells.

Sensor data analysis is used in many application areas, for example, Artificial Intelligence of Things (AIoT), with the rapid developing of the deep neural network learning that promotes its application area. In this work, we propose the Depth and Width Changeable Deep Kernel Learning‐based hyperspectral sensing data analysis algorithm. Compared with the traditional kernel learning‐based hyperspectral data classification, the proposed method has its advantages on the hyperspectral data classification. With the deep kernel learning, the feature is mapped through many times mapping and has the more discriminative ability. So, the deep kernel learning has the better performance compared with the multiple kernels learning. And it has the ability to adjust the network architecture for hyperspectral data space, with the optimization equation of the span bound. The experiments are implemented to testified the feasibility and performance of the algorithms on the hyperspectral data analysis, with the classification accuracy of hyperspectral data. The comprehensive analysis of the experiments shows that the proposed algorithm is feasible to hyperspectral sensor data analysis and its promising classification method in many areas data analysis.

It is well known that functional modes of lncRNAs are intimately associated with their subcellular localization1and relative nuclear enrichment of lncRNAs compared to mRNAs is a prevalent phenomenon2. As RBPs control the production, maturation, localization, translation, and degradation of cellular RNAs3, we reason that it is important to uncover partner RBPs that can bind and facilitate lncRNA nuclear localization. In this study, we thus harnessed the recently released large scale of eCLIP data4and subcellular RNA-seq data5available in K562 and HepG2 cell lines to characterize lncRNA-RBP interactome and uncovered potential factors and associated mechanisms determining lncRNA nuclear retention. Analyses of the subcellular RNA-seq data identified nuclear enriched lncRNAs (nuc-lncRNAs) and confirmed that lncRNAs and eRNAs (enhancer associated lncRNAs) are relatively nuclear enriched in both cells. By integrating the RBP binding profiles, we next generated RBP/nuc-lncRNA interaction map to identify RBPs associated with nuc-lncRNAs including HNRNPU, SAFB2, KHSRP and KHDRBS1 in K562 and HRNPNPC as well as HNRNPL in HepG2 cell lines. To further confirm the above findings, HNRNPU was knocked down and which led to nuclear retention of a panel of lncRNAs.

In this study, the authors present a novel level set method for infrared image segmentation. Local region‐based models can fit intensity inhomogeneity partly but they are sensitive to local window scale. To deal with it, they embed an heat diffusion process in conventional level set evolution and convert heat to a part of data term in level set energy function. Besides, bias field model can extract the local intensity clustering property of the image. Therefore, the proposed method can deal with the interference of intensity inhomogeneity and complex background if appropriate seeded pixels are selected. Finally, the energy functional is minimised by a combinatorial optimal algorithm in a graph model to get a global optimal solution and accelerate the level set evolution implementation. The experiments show that the proposed method is robust to parameter setting, noise, and initial contour position. The comparisons on a large quantity of infrared image datasets with standard level set methods also demonstrate the efficiency of the proposed method.

Intelligent internet data mining is an important application of AIoT (Artificial Intelligence of Things), and it is necessary to construct large training samples with the data from the internet, including images, videos, and other information. Among them, a hyperspectral database is also necessary for image processing and machine learning. The internet environment provides abundant hyperspectral data resources, but the hyperspectral data have no class labels and no so high value for applications. So, it is important to label the class information for these hyperspectral data through machine learning-based classification. In this paper, we present a quasiconformal mapping kernel machine learning-based intelligent hyperspectral data classification algorithm for internet-based hyperspectral data retrieval. The contributions include three points: the quasiconformal mapping-based multiple kernel learning network framework is proposed for hyperspectral data classification, the Mahalanobis distance kernel function is as the network nodes with the higher discriminative ability than Euclidean distance-based kernel function learning, and the objective function of measuring the class discriminative ability is proposed to seek the optimal parameters of the quasiconformal mapping projection. Experiments show that the proposed scheme is effective for hyperspectral image classification and retrieval.

Dynamic texture (DT) classification has attracted extensive attention in the field of image sequence analysis. The probability distribution model, which has been used to analysis DT, can describe well the distribution property of signals. Here, the authors introduce the finite mixtures of Gumbel distributions (MoGD) and the corresponding parameter estimation method based on expectation–maximisation algorithm. Then, the authors propose the DT features based on MoGD model for DT classification. Specifically, after decomposing DTs with the dual‐tree complex wavelet transform (DT‐CWT), the median values of complex wavelet coefficient magnitudes of non‐overlapping blocks in detail subbands are modelled with MoGDs. The model parameters are accumulated into a feature vector to describe DT. During the classification, a variational approximation version of the Kullback–Leibler divergence is used to measure the similarity between different DTs. The experimental evaluations on two popular benchmark DT data sets (UCLA and DynTex++) demonstrate the effectiveness of the proposed approach.

The stochastic signal model, hidden Markov model (HMM), is a probabilistic function of the Markov chain. In this letter, we propose a general nth-order HMM based dynamic texture description and classification method. Specifically, the pixel intensity sequence along time of a dynamic texture is modeled with a HMM that encodes the appearance information of the dynamic texture with the observed variables, and the dynamic properties over time with the hidden states. A new dynamic texture sequence is classified to the category by determining whether it is the most similar to this category with the probability that the observed sequence is produced by the HMMs of the training samples. The experimental results demonstrate the arbitrary emission probability distribution and the higher-order dependence of hidden states of a higher-order HMM result in better classification performance, as compared with the linear dynamical system (LDS) based method.