Search

Challenges in enhancing the multiclass segmentation of remotely sensed data include expensive and scarce labeled samples, complex geo-surface scenes, and resulting biases. The intricate nature of geographical surfaces, comprising varying elements and features, introduces significant complexity to the task of segmentation. The limited label data used to train segmentation models may exhibit biases due to imbalances or the inadequate representation of certain surface types or features. For applications like land use/cover monitoring, the assumption of evenly distributed simple random sampling may be not satisfied due to spatial stratified heterogeneity, introducing biases that can adversely impact the model's ability to generalize effectively across diverse geographical areas. We introduced two statistical indicators to encode the complexity of geo-features under multiclass scenes and designed a corresponding optimal sampling scheme to select representative samples to reduce sampling bias during machine learning model training, especially that of deep learning models. The results of the complexity scores showed that the entropy-based and gray-based indicators effectively detected the complexity from geo-surface scenes: the entropy-based indicator was sensitive to the boundaries of different classes and the contours of geographical objects, while the Moran's I indicator had a better performance in identifying the spatial structure information of geographical objects in remote sensing images. According to the complexity scores, the optimal sampling methods appropriately adapted the distribution of the training samples to the geo-context and enhanced their representativeness relative to the population. The single-score optimal sampling method achieved the highest improvement in DeepLab-V3 (increasing pixel accuracy by 0.3% and MIoU by 5.5%), and the multi-score optimal sampling method achieved the highest improvement in SegFormer (increasing ACC by 0.2% and MIoU by 2.4%). These findings carry significant implications for quantifying the complexity of geo-surface scenes and hence can enhance the semantic segmentation of high-resolution remote sensing images with less sampling bias. [ABSTRACT FROM AUTHOR]

Structured light can be encoded to carry information for free‐space optical communications with an extended degree of freedom to increase the capacity, however, the accuracy issue along with capacity increase is one of the biggest challenges that prevent practical applications. To achieve high accuracy with high capacity by a simple method, they propose the spatial nonlinear conversion of structured light into a communication network, especially, realizing an ultra‐high‐accuracy point‐to‐multipoint (PtoMP) information transmission link. A series of coherently superposed spatial modes and their spatial nonlinear conversion states are used as information carriers to replace the prior orbital angular momentum beams and greatly expand channel capacity within quite low spatial mode order. Through the spatial nonlinear conversion of simple dual‐mode superposition and a very basic neural network for machine learning‐based recognition, as high as 99.5% accuracy for more than 500 modes is obtained. By a combination of diffuse reflection screens and multiple CCDs, the large observation angle PtoMP information transmission is also proved to be feasible. This work paves the way for practical large‐scale multi‐party information networks using structured light. [ABSTRACT FROM AUTHOR]

In response to the personalized needs of the tourism industry, intelligent tourism planning services should be adopted. Through artificial intelligence and Big data technology, users' travel preferences, time, budget and other information are analyzed and mined in depth to provide users with more suitable travel routes and services. Therefore, according to the demand of tourist path recommendation in cultural scenic spots, this paper proposes a light sensor positioning method based on deep learning algorithm. This paper expounds the advantages of using optical sensor positioning for path recommendation, and describes the specific steps and processes of the optical sensor positioning method based on deep learning algorithm in detail. Finally, the feasibility and effect of this method are verified by experiments and data analysis. The results show that the localization of light sensor based on deep learning algorithm has high accuracy and practicability in the travel path recommendation of cultural scenic spots. Therefore, this method can provide accurate and reasonable path recommendation for tourists in cultural scenic spots and improve tourism experience and benefits. With the continuous development of the tourism market and the changing needs of users, this personalized tourism route planning and recommendation system will be increasingly widely applied and promoted. [ABSTRACT FROM AUTHOR]

This paper addressed the challenges associated with the complexity, numerous parameters, computational resource demands, and slow processing speed of transformer fault identification models based on deep learning technologies. Sparse convolutional neural network (CNN) approach is proposed for identifying faults related to dissolved gases in oil. Leveraging an improved Gramian angular field, one‐dimensional fault samples are converted into two‐dimensional feature images and data augmentation is implemented to meet the input requirements of deep learning models. Building upon visual geometry group (VGG)19 and residual networks (ResNet)50 networks for fault diagnosis, sparsity techniques are introduced through pruning, the fusion of convolution layers and batch normalization layers, and parameter quantization. Numerical experiments and performance evaluations on dissolved gas in transformer oil fault data demonstrate that the proposed method effectively reduced model complexity, minimized parameter count, conserved computational resources, and improved processing speed while maintaining a considerable level of fault identification accuracy. This made it applicable to edge computing platforms characterized by small form factors and low power consumption in the power industry. [ABSTRACT FROM AUTHOR]

[Display omitted] • The synergistic effect of saltiness-enhancing peptides was demonstrated by S-curves. • The 3D model of TMC4 with ten transmembranes was constructed by homology modeling. • The binding mechanism of saltiness-enhancing peptides with TMC4 was explored. • The active sites of NEFGYSNR were calculated by frontier molecular orbitals analysis. The previously obtained chicken-derived umami peptides in the laboratory were evaluated for their saltiness-enhancing effect by sensory evaluation and S-curve, and the results revealed that peptides TPPKID, PKESEKPN, TEDWGR, LPLQDAH, NEFGYSNR, and LPLQD had significant saltiness-enhancing effects. In the binary solution system with salt, the ratio of the experimental detection threshold (129.17 mg/L) to the theoretical detection threshold (274.43 mg/L) of NEFGYSNR was 0.47, which had a synergistic saltiness-enhancing effect with salt. The model of transmembrane channel-like protein 4 (TMC4) channel protein was constructed by homology modeling, which had a 10-fold transmembrane structure and was well evaluated. Molecular docking and frontier molecular orbitals showed that the main active sites of TMC4 were Lys 471, Met 379, Cys 475, Gln 377, and Pro 380, and the main active sites of NEFGYSNR were Tyr, Ser and Asn. This study may provide a theoretical reference for low-sodium diets. [ABSTRACT FROM AUTHOR]

• CV-based contactless BWIM (cBWIM) system is proposed. • Only two consumer cameras are used to realize real-time monitoring of bridge displacement and vehicle trajectory. • It is applicable for non-uniform motion and vehicle occlusion cases. • The proposed method is verified by laboratory tests and field measurement. Bridge weigh-in-motion (BWIM) system plays increasingly important role in vehicle load monitoring and bridge maintenance due to advantages of high efficiency and little impact on traffic. However, the contact-type sensors in a traditional BWIM system should be installed in advance and is environment-sensitive, which hinders its application. To this end, a vision-based contactless BWIM (cBWIM) system is proposed. The core idea of the current method is to replace the traditional contact-type sensors by two consumer cameras to realize real-time monitoring of bridge displacement and vehicle trajectory. The cBWIM system consists of three modules, including the axle weight identification algorithm, the bridge displacement monitoring system and the vehicle trajectory tracking system. The influence line-based Moses algorithm is utilized for identification of axle weight. For bridge displacement monitoring, the minimum bounding rectangle-based similarity matching method is proposed to achieve high accuracy measurement of displacement and real-time data processing. For vehicle trajectory tracking, a deep learning-based YOLOv5s algorithm is adopted as a detector, combined with DeepSort algorithm to realize real-time tracking of vehicle position. The cBWIM system is applicable for non-uniform motion and vehicle occlusion cases. Laboratory tests covering various cases (i.e. uniform motion, non-uniform motion and visual occlusion) and field measurement on a full-scale bridge have demonstrated that the proposed method is effective and reliable and can provide alternative tool to the current vehicle weighing methods. [ABSTRACT FROM AUTHOR]

Objectives: To develop an interactive, non-invasive artificial intelligence (AI) system for malignancy risk prediction in cystic renal lesions (CRLs). Methods: In this retrospective, multicenter diagnostic study, we evaluated 715 patients. An interactive geodesic-based 3D segmentation model was created for CRLs segmentation. A CRLs classification model was developed using spatial encoder temporal decoder (SETD) architecture. The classification model combines a 3D-ResNet50 network for extracting spatial features and a gated recurrent unit (GRU) network for decoding temporal features from multi-phase CT images. We assessed the segmentation model using sensitivity (SEN), specificity (SPE), intersection over union (IOU), and dice similarity (Dice) metrics. The classification model's performance was evaluated using the area under the receiver operator characteristic curve (AUC), accuracy score (ACC), and decision curve analysis (DCA). Results: From 2012 to 2023, we included 477 CRLs (median age, 57 [IQR: 48–65]; 173 men) in the training cohort, 226 CRLs (median age, 60 [IQR: 52–69]; 77 men) in the validation cohort, and 239 CRLs (median age, 59 [IQR: 53–69]; 95 men) in the testing cohort (external validation cohort 1, cohort 2, and cohort 3). The segmentation model and SETD classifier exhibited excellent performance in both validation (AUC = 0.973, ACC = 0.916, Dice = 0.847, IOU = 0.743, SEN = 0.840, SPE = 1.000) and testing datasets (AUC = 0.998, ACC = 0.988, Dice = 0.861, IOU = 0.762, SEN = 0.876, SPE = 1.000). Conclusion: The AI system demonstrated excellent benign-malignant discriminatory ability across both validation and testing datasets and illustrated improved clinical decision-making utility. Critical relevance statement: In this era when incidental CRLs are prevalent, this interactive, non-invasive AI system will facilitate accurate diagnosis of CRLs, reducing excessive follow-up and overtreatment. Key Points: The rising prevalence of CRLs necessitates better malignancy prediction strategies. The AI system demonstrated excellent diagnostic performance in identifying malignant CRL. The AI system illustrated improved clinical decision-making utility. [ABSTRACT FROM AUTHOR]

Range-spread target (RST) detection is an important issue for high-resolution radar (HRR). Traditional detectors relying on manually designed detection statistics have their performance limitations. Therefore, in this work, two deep learning-based detectors are proposed for RST detection using HRRPs, i.e., an NLS detector and DFCW detector. The NLS detector leverages domain knowledge from the traditional detector, treating the input HRRP as a low-level feature vector for target detection. An interpretable NLS module is designed to perform noise reduction for the input HRRP. The DFCW detector takes advantage of the extracted high-level feature map of the input HRRP to improve detection performance. It incorporates a feature cross-weighting module for element-wise feature weighting within the feature map, considering the channel and spatial information jointly. Additionally, a nonlinear accumulation module is proposed to replace the conventional noncoherent accumulation operation in the double-HRRP detection scenario. Considering the influence of the target spread characteristic on detector performance, signal sparseness is introduced as a measure and used to assist in generating two datasets, i.e., a simulated dataset and measured dataset incorporating real target echoes. Experiments based on the two datasets are conducted to confirm the contribution of the designed modules to detector performance. The effectiveness of the two proposed detectors is verified through performance comparison with traditional and deep learning-based detectors. [ABSTRACT FROM AUTHOR]

Understanding the genetic architecture of brain structure is challenging, partly due to difficulties in designing robust, non-biased descriptors of brain morphology. Until recently, brain measures for genome-wide association studies (GWAS) consisted of traditionally expert-defined or software-derived image-derived phenotypes (IDPs) that are often based on theoretical preconceptions or computed from limited amounts of data. Here, we present an approach to derive brain imaging phenotypes using unsupervised deep representation learning. We train a 3-D convolutional autoencoder model with reconstruction loss on 6130 UK Biobank (UKBB) participants' T1 or T2-FLAIR (T2) brain MRIs to create a 128-dimensional representation known as Unsupervised Deep learning derived Imaging Phenotypes (UDIPs). GWAS of these UDIPs in held-out UKBB subjects (n = 22,880 discovery and n = 12,359/11,265 replication cohorts for T1/T2) identified 9457 significant SNPs organized into 97 independent genetic loci of which 60 loci were replicated. Twenty-six loci were not reported in earlier T1 and T2 IDP-based UK Biobank GWAS. We developed a perturbation-based decoder interpretation approach to show that these loci are associated with UDIPs mapped to multiple relevant brain regions. Our results established unsupervised deep learning can derive robust, unbiased, heritable, and interpretable brain imaging phenotypes. A study utilizing unsupervised deep learning to generate interpretable brain imaging phenotypes from brain T1 and T2-FLAIR MRI identified 97 genetic loci enhancing understanding of brain structure genetics. [ABSTRACT FROM AUTHOR]

The matching of SAR and optical images is crucial for various remote sensing applications, such as monitoring natural disasters and change detection. However, the significant differences in geometric and radiometric properties between these two sensors pose challenges for robust and accurate matching. Recent deep learning-based approaches mitigate modality differences by aligning all contents on a single pixel-level feature map, leading to limited robustness to content differences and resolution variations. In this paper, we propose a novel robust SAR-optical matching network to address these challenges. To enhance robustness against noise and resolution changes, we align and match on feature maps of multiple granularities simultaneously. Further, we introduce the novel multi-granularity matching strategy called "look closer to match better" to reduce the computational burden of global matching across multiple granularities. This strategy employs coarse-grained features to quickly narrow down the search range, followed by the use of finer-grained features to gradually locate the finer matching position within a reduced search range, improving both efficiency and performance. Additionally, we address the issue of treating all regions equally during feature alignment by proposing a Non-Shared Contents Filtering (NSCF) module. This module adaptively filters out non-shared regions that are difficult to align, thereby avoiding its interference with the similarity measure of the consistent representation and enhancing robustness to content differences. We evaluate our framework on various satellite datasets. Experiments show that our method achieves the best performance on the SEN1-2 dataset and competitive generalization ability on other unseen satellite datasets. • Propose novel SAR-optical image matching network. • Adaptive selection of multi-granularity shared contents for robust matching. • Introduce multi-granularity matching strategy. • Flexible fusion strategy for accurate results with reduced costs. • NSCF module enhances robustness to content difference. [ABSTRACT FROM AUTHOR]