Feng He

Infrared dim and small target detection aims to accurately localize targets within complex backgrounds or clutter. However, under extremely low signal-to-noise ratio (SNR) conditions, single-frame detection methods often fail to effectively detect such targets. In contrast, multi-frame detection can exploit temporal cues to significantly improve the probability of detection (Pd) and reduce false alarms (Fa). Existing multi-frame approaches often employ 3D convolutions/RNNs to implicitly extract temporal features. However, they typically lack explicit modeling of target motion. To address this, we propose a Dual Encoder–Decoder Multi-Frame Infrared Small Target Detection Network with Motion Encoding (DEMNet) that explicitly incorporates motion information into the detection process. The first multi-level encoder–decoder module leverages spatial and channel attention mechanisms to fuse hierarchical features across multiple scales, enabling robust spatial feature extraction from each frame of the temporally aligned input sequence. The second encoder–decoder module encodes both inter-frame target motion and intra-frame target positional information, followed by 3D convolution to achieve effective motion information fusion. Extensive experiments demonstrate that DEMNet achieves state-of-the-art performance, outperforming recent advanced methods such as DTUM and SSTNet. For the DAUB dataset, compared to the second-best model, DEMNet improves Pd by 2.42 percentage points and reduces Fa by 4.13 × 10⁻⁶ (a 68.72% reduction). For the NUDT dataset, it improves Pd by 1.68 percentage points and reduces Fa by 0.67 × 10⁻⁶ (a 7.26% reduction) compared to the next-best model. Notably, DEMNet demonstrates even greater advantages on test sequences with SNR ≤ 3.

Infrared dim and small target detection aims to accurately localize targets within complex backgrounds or clutters. However, under extremely low signal-to-noise ratio (SNR) conditions, single-frame detection methods often fail to effectively detect such targets. In contrast, multi-frame detection can exploit temporal cues to significantly improve Pd (probability of detection) and reduce Fa (false alarms). Existing multi-frame approaches typically rely on 3D convolutions/RNNs to implicitly extract temporal features, yet they typically neglect the explicit modeling of target motion. In this paper, we propose a Dual-Encoder-Decoder Multi-Frame Infrared Small Target Detection Network with Motion Encoding (DEMNet) that explicitly incorporates motion information into the detection process. The first multi-level encoder-decoder module applies spatial and channel attention mechanisms to fuse hierarchical features across multiple scales, enabling robust spatial feature extraction from each frame of the temporally aligned input sequence. The second encoder-decoder module encodes both inter-frame target motion and intra-frame target position information, followed by 3D convolution to achieve effective motion information fusion. Extensive experiments demonstrate that DEMNet achieves state-of-the-art performance, outperforming recent advanced methods such as DTUM and SSTNet. On DAUB dataset, compared to the second-best model, DEMNet improves Pd by 2.42 percentage points and reduces Fa by 4.13×10⁻⁶ (reduced by 68.72%). On the NUDT dataset, it improves Pd by 1.68 percentage points, reduces Fa by 0.67×10⁻⁶ (reduced by 7.26%) compared to the next-best model. Notably, on the test sequences with SNR≤ 3, DEMNet’s advantages are even more pronounced.

In the existing studies on remote sensing image scene classification, the supervised learning methods which are fine-tuned from pre-trained model require a large amount of labeled training data and parameters, while unsupervised learning methods do not make full use of label information, and the classification performance could be improved. In this paper, we introduced semi-supervised learning into generative adversarial network (GAN), so the discriminator learned more discriminative features from labeled data and unlabeled data. Moreover, the mixup data augmentation method was introduced into our classification model to augment the data and stabilized the training process. We carried out extensive experiments for both UC-Merced and NWPU-RESISC45 datasets with a 5-fold cross-validation protocol using a linear SVM as classifier. We trained the proposed method on UC-Merced dataset and achieve an average overall accuracy of 94.05% under 80% training ratio. When trained on NWPU-RESISC45 dataset, the proposed method reached an average overall accuracy of 83.12% and 92.78% under the training ratios of 20% and 80% respectively, which achieves the state-of-the-art deep learning methods without pre-training.

This communication addresses the thinned linear array design for the optimization of radiometric sensitivity and angular resolution, both of which are crucial characteristics of interferometric aperture synthesis radiometers (IASRs). First, we establish a multi-objective optimization model concerning their tradeoff. Then, three state-of-the-art search algorithms, namely, multi-objective particle swarm optimization (MOPSO), nondominated sorting genetic algorithm II (NSGA-II), and nondominated neighbor immune algorithm (NNIA), are implemented for the design of linear interferometric arrays. Further, a constraint principle of interferometric arrays (CPIA) is proposed to reveal the inherent relation among array parameters closely related to IASR performances, which would be useful for the multi-objective optimization of IASR arrays.

Biometrics identification uses human biometric such as human faces, fingerprints, irises for authentication. Electroencephalograph (EEG) provides a detectable neuroelectrophysiological signal that could reflect psychological or behavioral characteristics. In the past, EEG -based biometrics were usually used with machine learning based procedure for identification. Recently Convolutional Neural Networks (CNN) was used as a new tool for biometric automatic feature exaction and classification. This paper aims at investigating CNN's performance on EEG -based human identification with the increase number of subjects. P300 -speller was introduced in this study to induce Event -related potential (ERP). 33 healthy subjects participated in the experiment. Result shows that the CNN -based Biometric System reached a high degree of accuracy (99.9%) after for 8 -class classification , 99.3% for 10 -class and 99.3% for 13 -class. Then, we made some finetuning to the network structure of the 10 -class and the 13 class classification, Then,we also improved the convergence speed of the network and reduced the time to get the best classification model by changing the structure of the CNN that we introduced here, while ensuring that the accuracy and the loss function did not change significantly. The finding indicts that CNN get good performance in EEG -based biometric identification.

Double difference bases (DDBs) and construction of thinned linear arrays with twofold redundancy are discussed in this communication. We present an analytical method, based on the combinatorial tools, to generate twofold redundancy arrays by employing regular DDB patterns separately or combining them with cyclic difference sets. Besides, the bound for DDBs is derived that would be useful to research on thinned array geometries.

There have been some algorithms that use image processing for power quality disturbances identification. This algorithms firstly converts 1-D power signal to 2-D image, then use image algorithms to classify power quality disturbances. Different from these methods, a new method is presented here. Converts 2-D texture image algorithms to 1-D waveform algorithms, then use a classifier(here is SVM) for recognize power quality disturbances. These 1-D features include level Co-occurrence Matrix, Markov random field, voltage values and gradient firstly. The experiment show that the algorithm can reach a good result.