1.385 hasil
Menampilkan 1–20 · 0,53 detik · RSS
上海师范大学学报. 自然科学版
ISSN 1000-5137 · Academic Journals Center of Shanghai Normal University · 1.385 artikel di indeks
-
Artikel Jurnal
A nighttime image dehazing method integrating dual-domain feature extraction and mixture
For nighttime images, this paper proposes an end-to-end dehazing model NIDFEM-Net that combines dual-domain feature extraction and fusion. The model is trained on a nighttime hazy image dataset generated based on image generation technology, enabling it to remove nighttime haze and enhance the…
-
Artikel Jurnal
Research on high-precision fiber shape sensing optimization by integrating spectral attention with position-weighted loss
This study addresses the deficiencies of existing fiber shape sensing (FSS) methods in feature modeling and training target design, and proposes two improvement measures： (1) introducing a spectral specificity attention module to achieve adaptive recalibration of the importance of input spectral…
-
Artikel Jurnal
Research on innovative scenarios of AIGC-enabled performance management education
Addressing issues such as the disconnect between theory and practice, insufficient cultivation of high-order abilities, and monotonous evaluation methods in university performance management education, we reconstruct a curriculum objective system encompassing knowledge, abilities, skills, and…
-
Artikel Jurnal
Intestinal polyp segmentation network model based on local thresholding and dual attention
Early detection of colorectal cancer is crucial for improving the cure rate, where accurate polyp segmentation plays a pivotal role in disease quantification and treatment evaluation. However, due to significant variations in polyp size, shape, boundary clarity, and background complexity, automatic…
-
Artikel Jurnal
Research on mobile robot positioning based on zonotope set-membership estimation algorithm
This paper proposes a multi-convex set-membership estimation algorithm for mobile robot localization by integrating multi-sensor data. Compared to algorithms like Kalman filtering, which can only achieve single-point estimation and rely on specific noise statistical characteristics, the…
-
Artikel Jurnal
Domain generalization method for human activity recognition based on structured temporal Mixup and supervised contrastive learning
Human activity recognition (HAR) based on wearable sensors has garnered widespread attention in application scenarios such as intelligent health monitoring, smart homes, and human-computer interaction. However, due to significant differences in movement patterns, physiological characteristics, and…
-
Artikel Jurnal
YOLO-AVF: an improved forest fire detection algorithm based on YOLOv11
Based on the improved you only look once version 11 (YOLOv11) algorithm, this paper proposes a forest fire detection (YOLO-AVF) model. It introduces an attention-enhanced interactive feature integration (AIFI) module to strengthen high-level semantic feature expression. The model employs a…
-
Artikel Jurnal
Gesture recognition based on few-shot learning and sEMG
Surface electromyography (sEMG) signals can directly reflect human activity intentions and hold great potential for applications in fields such as medical rehabilitation and robotic arm teleoperation. Currently, most sEMG-based classifiers suffer from issues of limited flexibility and scalability…
-
Artikel Jurnal
Knowledge tracing under data sparsity: an agent-based student simulation data augmentation method
Knowledge tracing (KT), as an important technology, is widely applied in predicting students’ learning status. However, in the actual teaching process, data sparsity has always been a key factor affecting the performance of traditional KT models. The data sparsity issue often stems from students’…
-
Artikel Jurnal
A generative model enhancement method with identity-aware attention and frequency-domain reweighting
Addressing the issues of unstable semantic injection, decreased identity consistency, and difficulty in balancing structure and details that often arise in text-to-image diffusion models under complex semantic conditions, this paper proposes a model-level structure enhancement method within the…
-
Artikel Jurnal
An image super-resolution model based on improved latent space consistency training
In the field of image restoration, diffusion models can achieve high-quality reconstruction results, but they often come with high computational costs. Consistency models (CM) have alleviated this issue to some extent, but their research in image super-resolution tasks remains relatively limited…
-
Artikel Jurnal
Outage probability analysis of cooperative NOMA with lossy forward relaying
This paper investigates the outage probability performance of a cooperative non-orthogonal multiple access (CO-NOMA) system employing lossy forwarding (LF) relays in block Rayleigh fading channels. Specifically, the near-end user equipment (UE) in the NOMA system acts as a relay, forwarding the…
-
Artikel Jurnal
AI-generated image quality evaluation: a layered review from pixels to ethics
This paper summarizes a unified four-layer evaluation framework encompassing pixel (L1), structural (L2), semantic (L3), and value (L4) dimensions. Analysis reveals that compared to mature visual fidelity and semantic alignment metrics, geometric consistency and value alignment represent critical…
-
Artikel Jurnal
CFP-OCT multimodal glaucoma grading method based on uncertainty fusion
Glaucoma is an irreversible blinding eye disease characterized by progressive damage to the optic nerve structure. Accurate disease diagnosis and grading are crucial for clinical intervention and disease management. Color fundus photography and optical coherence tomography exhibit significant…
-
Artikel Jurnal
Light field image compression and reconstruction method based on three-branch interaction
Traditional light field image compression methods fail to fully utilize the spatial， angular， and geometric structure information of light field images， resulting in limited image reconstruction quality and compression efficiency. To address this， this paper proposes a light field image compression…
-
Artikel Jurnal
A preliminary study on pests of reishi mushroom in Shanghai: species diversity, life history, and occurrence patterns
In July 2025， a field survey conducted at a grape-Ganoderma lucidum Chinese name Lingzhi intercropping base in Jiading district， Shanghai， which revealed a diverse pest species belonging to various taxa. In addition to hygrophilous invertebrates such as slug Meghimatium bilineatum， snail Acusta…
-
Artikel Jurnal
A semi-supervised medical image segmentation method based on adversarial learning and cross-frequency interaction
In medical image segmentation tasks， obtaining high-quality pixel-level labeled data is costly. Existing semi-supervised medical image segmentation methods often rely on consistency regularization methods or pseudo-label learning， but the modeling ability of a single feature domain is limited，…
-
Artikel Jurnal
AGM lightweight occlusion awareness light field depth estimation network
Light field imaging captures scene information simultaneously in spatial and angular dimensions， providing rich geometric cues for high-precision disparity and depth estimation tasks. However， existing methods are prone to issues such as matching ambiguity and boundary errors due to occlusion，…
-
Artikel Jurnal
Improved VMD-BiLSTM blood glucose prediction model based on NOA optimization algorithm
For patients with type 1 diabetes， effective blood glucose management is crucial. Traditional insulin injection methods have limitations such as inconvenience in operation and insufficient accuracy， leading to a growing demand for reliable artificial pancreas （AP） systems. Accurate prediction of…
-
Artikel Jurnal
A multi-time scale heterogeneous expert fusion framework for bond yield prediction
Addressing the issues of inconsistent time scales of multi-source heterogeneous signals in financial time series forecasting， insufficient fusion capability of single models， and weak interpretability of results， this paper proposes a multi-time scale heterogeneous expert fusion forecasting…
Menampilkan maksimal 50 halaman. Persempit pencarian dengan filter untuk hasil yang lebih spesifik.