Query：
学者姓名：黄世国
Refining：
Year
Type
Indexed by
Source
Complex
Co-Author
Language
Clean All
Abstract ：
Algal odors pose significant challenges to the sensory quality of food products, often affecting consumer acceptance and product marketability. Despite advancements in food engineering, research focused on effective odor control in algal-based functional foods remains limited. This study utilizes machine learning to analyze the molecular structures of algal-derived odor compounds, aiming to enhance sensory quality through targeted processing methods. Molecular descriptors were extracted using RDKit from compounds in the OlfactionBase and PubChem databases, followed by visualization with t-distributed stochastic neighbor embedding (t-SNE), revealing distinct clustering patterns for specific odor profiles. Six machine learning algorithms, including Gaussian Naive Bayes, Random Forest, Support Vector Machine, k-Nearest Neighbors, Stochastic Gradient Descent, and Gradient Boosting Decision Trees, were evaluated for classification accuracy. Initially, a binary classification model was constructed to differentiate between "Ammonia-like" and "Rancid" odors. To further enhance model generalizability, an additional odor class ("Other") was incorporated to reflect non-typical or mixed odor profiles, resulting in a multiclass classification task. Feature selection and dimensionality reduction were conducted using RFECV and PCA, respectively, followed by model training and validation. In the binary classification task, the k-Nearest Neighbors model demonstrated superior performance, achieving accuracies of 94.25 % and 93.11 % in 3-fold and 5-fold cross-validation, respectively. In multi-classification tasks, Stochastic Gradient Descent achieves the best results. This computational approach offers a novel framework for odor mitigation, aiding the development of engineered solutions that improve the sensory characteristics of algal-derived foods. Future research should focus on integrating machine learning models with physical, chemical, and biological odor-control methods for more comprehensive strategies in food engineering.
Keyword ：
Algal odor control Algal odor control Food engineering Food engineering Molecular structure analysis Molecular structure analysis Neural networks Neural networks Sensory quality enhancement Sensory quality enhancement
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | Sun, Yilan , Zhang, Qinhua , Lin, Hongkun et al. Engineering odor control in algal foods: Machine learning for quality enhancement [J]. | JOURNAL OF FOOD ENGINEERING , 2026 , 402 . |
| MLA | Sun, Yilan et al. "Engineering odor control in algal foods: Machine learning for quality enhancement" . | JOURNAL OF FOOD ENGINEERING 402 (2026) . |
| APA | Sun, Yilan , Zhang, Qinhua , Lin, Hongkun , Lu, Juehan , Zhang, Huiyue , Su, Che et al. Engineering odor control in algal foods: Machine learning for quality enhancement . | JOURNAL OF FOOD ENGINEERING , 2026 , 402 . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
Traditional image classification often misclassifies unknown samples as known classes during testing, degrading recognition accuracy. Open-set image recognition can simultaneously detect known classes (KCs) and unknown classes (UCs) but still struggles to improve recognition performance caused by open space risk. Therefore, we introduce a cosine distance loss function (CDLoss), which exploits the orthogonality of one-hot encoding vectors to align known samples with their corresponding one-hot encoder directions. This reduces the overlap between the feature spaces of KCs and UCs, mitigating open space risk. CDLoss was incorporated into both Softmax-based and prototype-learning-based frameworks to evaluate its effectiveness. Experimental results show that CDLoss improves AUROC, OSCR, and accuracy across both frameworks and different datasets. Furthermore, various weight combinations of the ARPL and CDLoss were explored, revealing optimal performance with a 1:2 ratio. T-SNE analysis confirms that CDLoss reduces the overlap between the feature spaces of KCs and UCs. These results demonstrate that CDLoss helps mitigate open space risk, enhancing recognition performance in open-set image classification tasks.
Keyword ：
cosine distance loss cosine distance loss one-hot encoding one-hot encoding open-set image classification open-set image classification open space risk open space risk
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | Li, Xiaolin , Chen, Binbin , Li, Jianxiang et al. Cosine Distance Loss for Open-Set Image Recognition [J]. | ELECTRONICS , 2025 , 14 (1) . |
| MLA | Li, Xiaolin et al. "Cosine Distance Loss for Open-Set Image Recognition" . | ELECTRONICS 14 . 1 (2025) . |
| APA | Li, Xiaolin , Chen, Binbin , Li, Jianxiang , Chen, Shuwu , Huang, Shiguo . Cosine Distance Loss for Open-Set Image Recognition . | ELECTRONICS , 2025 , 14 (1) . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
【目的】针对蝴蝶样本存在类间和类内分布不平衡导致识别性能下降的问题，探索一种多损失融合的蝴蝶自动识别方法。【方法】利用开源的Butterfly-200图像数据集作为实验数据。该数据集包括200种蝴蝶，每种蝴蝶的图像数量从30～885不等。以交叉熵损失(cross-entropy loss)为基准损失，分别叠加对比损失(contrastive loss)、焦点损失(focal loss)、类平衡损失(class-balanced loss)、采样(sampling)、logit调整(logit adjustment),比较算法的识别性能。在此基础上，利用中心损失(center loss)有助于缓解类内不平衡而焦点损失有助于缓解类内和类间不平衡的特点，开展消融实验分析叠加中心损失和焦点损失对识别性能的影响，提出了融合上述这两种损失的蝴蝶自动识别新方法。【结果】交叉熵损失与其他单一损失(对比损失除外)结合时，算法的识别性能基本上呈现不同程度的下降。我们的算法在交叉熵损失基础上结合中心损失和焦点损失后，其识别性能均超过交叉熵损失及其与其他损失的组合，准确率、F1分值、查准率和召回率分别91.67%, 90.68%, 91.68%和90.38%。消融试验进一步证实了中心损失和焦点损失的互补性，同时使用这两种损失能明显提升识别性能。此外，不同权重的损失组合对识别性能也有明显影响。【结论】研究结果证明融合中心损失和焦点损失在一定程度上缓解了类间和类内分布不均衡的问题，能够有效提高蝴蝶识别的准确性，为生态环境监测提供了一种有效的辅助手段。
Keyword ：
中心损失 中心损失 交叉熵损失 交叉熵损失 分布不均衡 分布不均衡 图像分类 图像分类 焦点损失 焦点损失 蝴蝶 蝴蝶
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | 李小林 , 李建祥 , 陈彬彬 et al. 融合中心损失和焦点损失的蝴蝶自动识别 [J]. | 昆虫学报 , 2025 , 68 (02) : 223-230 . |
| MLA | 李小林 et al. "融合中心损失和焦点损失的蝴蝶自动识别" . | 昆虫学报 68 . 02 (2025) : 223-230 . |
| APA | 李小林 , 李建祥 , 陈彬彬 , 王荣 , 张飞萍 , 黄世国 . 融合中心损失和焦点损失的蝴蝶自动识别 . | 昆虫学报 , 2025 , 68 (02) , 223-230 . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
Pine wilt disease poses a significant threat to pine forests worldwide, necessitating efficient and accurate detection of dead pine wood for effective disease control and forest management. Traditional deep learning methods based on supervised closed-set paradigms often struggle to address unknown interfering objects, causing false positives, missed detection, and increased annotation burdens. To overcome these challenges, we propose SS-OPDet, a semi-supervised open-set detection framework that leverages a small amount of labeled data along with abundant unlabeled data. SS-OPDet integrates a Weighted Multi-scale Feature Fusion module to dynamically integrate global- and local-scale features, thereby significantly improving representational accuracy for dead pine wood. Additionally, a Dynamic Confidence Pseudo-Label Generation strategy categorizes predictions by confidence level, effectively reducing training noise and maximizing the use of reliable unlabeled data. Experimental results from 7733 UAV images demonstrate that SS-OPDet achieves an average precision (APK) of 84.73%, a recall (RK) of 94.48%, an Absolute Open-Set Error (AOSE) of 271 and a Wilderness Impact (WI) of 0.0917%. Cross-region validation further confirms the robustness and generalization capability of the proposed framework. The proposed method offers a cost-effective and accurate solution for timely detection of pine wilt disease, providing substantial benefits to forest monitoring and management.
Keyword ：
dead pine wood dead pine wood dynamic confidence pseudo-label generation dynamic confidence pseudo-label generation open-set detection open-set detection semi-supervised learning semi-supervised learning weighted multi-scale feature fusion weighted multi-scale feature fusion
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | Lu, Xiaojian , Huang, Shiguo , Wu, Songqing et al. SS-OPDet: A Semi-Supervised Open-Set Detection Framework for Dead Pine Wood Detection [J]. | SENSORS , 2025 , 25 (11) . |
| MLA | Lu, Xiaojian et al. "SS-OPDet: A Semi-Supervised Open-Set Detection Framework for Dead Pine Wood Detection" . | SENSORS 25 . 11 (2025) . |
| APA | Lu, Xiaojian , Huang, Shiguo , Wu, Songqing , Zhang, Feiping , Weng, Mingqing , Luo, Jianlong et al. SS-OPDet: A Semi-Supervised Open-Set Detection Framework for Dead Pine Wood Detection . | SENSORS , 2025 , 25 (11) . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
藻类富含多种营养物质，是重要的水产食品来源之一。然而，在加工过程中，藻类会因产生腥味而影响产品的食用品质。为解决此问题，通常采用物理、化学、生物及复合方法进行脱腥处理。目前，机器学习和多变量分析方法已被引入腥味物质的分类和预测研究，在腥味处理方面展现了较大的应用潜力。该文系统综述藻类腥味的形成机制、智能化检测及脱腥方法的最新研究进展。
Keyword ：
智能检测 智能检测 机器学习 机器学习 海洋藻类 海洋藻类 脱腥处理 脱腥处理 腥味形成机制 腥味形成机制 食品加工 食品加工
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | 张钦华 , 洪婉馨 , 胡知文 et al. 藻类腥味智能检测与脱腥策略的研究进展 [J]. | 食品研究与开发 , 2025 , 46 (20) : 182-194 . |
| MLA | 张钦华 et al. "藻类腥味智能检测与脱腥策略的研究进展" . | 食品研究与开发 46 . 20 (2025) : 182-194 . |
| APA | 张钦华 , 洪婉馨 , 胡知文 , 雒艳颢 , 于子洋 , 王月光 et al. 藻类腥味智能检测与脱腥策略的研究进展 . | 食品研究与开发 , 2025 , 46 (20) , 182-194 . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
本综述对机器学习在花椒中麻味物质的应用进行了总结。市面上不同品种的花椒含有不同的麻味物质且其含量亦不尽相同，同时花椒麻味物质的组成及其含量的传统检测与分析方法有诸多局限性，因此引入机器学习算法为这一领域带来了新的可能性。使用机器学习算法建立花椒品质的预测模型与干燥模型，综合对花椒的感官评价，建立花椒种质资源库，对花椒的遗传育种具有重大意义。本综述系统地回顾了不同花椒品种中麻味分子的组成及其含量，同时分析了机器学习算法在品质预测模型、干燥模型与麻味物质数据分析中的应用情况。通过整合机器学习技术，研究人员能够更深入地了解目前已建立的模型，基于麻味物质为花椒的产量与品质的优化提供支持。
Keyword ：
机器学习 机器学习 模型 模型 花椒 花椒 麻味物质 麻味物质
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | 王月光 , 李小林 , 王芹 et al. 花椒中麻味物质的研究及机器学习方法应用进展 [J]. | 食品科学 , 2024 , 45 (18) : 282-289 . |
| MLA | 王月光 et al. "花椒中麻味物质的研究及机器学习方法应用进展" . | 食品科学 45 . 18 (2024) : 282-289 . |
| APA | 王月光 , 李小林 , 王芹 , 苏澈 , 张钦华 , 黄世国 et al. 花椒中麻味物质的研究及机器学习方法应用进展 . | 食品科学 , 2024 , 45 (18) , 282-289 . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
This study explores the predictive analysis of fishy odors in food, a crucial factor in sensory quality, using machine learning to analyze molecular structures. We compiled a novel dataset of fishy odor molecules from the OlfactionBase database, supplemented with PubChem data via web crawler techniques. Using RDKit for chemoinformatics, we extracted molecular structural features and applied t-distributed stochastic neighbor embedding (t-SNE) for visualization, revealing distinct molecular clusters corresponding to various fishy odor types and highlighting unique molecular descriptors. We employed six machine learning algorithms (GNB, RF, SVC, KNN, SGD, and GBDT) to investigate the relationship between molecular structure and odor. The KNN model demonstrated superior classification performance, achieving accuracies of 94.25% and 93.11% in 3-fold and 5-fold cross-validation, respectively. Our findings offer valuable insights into the molecular basis of different fishy odor types, significantly enhancing the understanding of flavor mechanisms in food science. © 2024, The Authors. All rights reserved.
Keyword ：
Learning algorithms Learning algorithms Machine learning Machine learning Molecular structure Molecular structure Odors Odors Quality control Quality control Sensory analysis Sensory analysis Stochastic systems Stochastic systems Web crawler Web crawler
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | Sun, Yilan , Lin, Hongkun , Lu, Juehan et al. Deciphering Fishy Odor Profiles in Food: A Machine Learning Approach to Molecular Structure Analysis [J]. | SSRN , 2024 . |
| MLA | Sun, Yilan et al. "Deciphering Fishy Odor Profiles in Food: A Machine Learning Approach to Molecular Structure Analysis" . | SSRN (2024) . |
| APA | Sun, Yilan , Lin, Hongkun , Lu, Juehan , Zhang, Qinhua , Su, Che , Huang, Shiguo et al. Deciphering Fishy Odor Profiles in Food: A Machine Learning Approach to Molecular Structure Analysis . | SSRN , 2024 . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
[目的]昆虫常在色彩、纹理或形态上和背景相似,具有伪装性,识别难度大.本研究旨在探索基于深度学习的伪装昆虫前背景自动分割方法.[方法]将显著目标检测算法(salient object detection algorithm)、大模型 图像分割算法(large-scale model-based image segmentation algorithm)以及伪装目标检测算法(camouflaged object detection algorithm)应用于伪装昆虫数据集,该数据集包括10类昆虫共1 900张图片;并进一步针对现有伪装目标检测算法的不足,提出了一种基于DGNet(deep-gradient network)的网络模型改进方法,即 ZDNet(zoom-deep gradient network).在构建该模型时,充分运用图像特征增强、交错图像金字塔、梯度诱导和跳跃式特征融合等技术.利用伪装目标检测公开数据集COD10K与CAMO构建了包含螽斯、蜘蛛等10个目昆虫的图像数据集,结合迁移学习进行网络训练,将经过训练的模型用于分割伪装昆虫.[结果]现有的伪装目标检测模型用于伪装昆虫前背景分割时,其分割性能明显优于显著目标检测模型和大模型分割图像.同时,ZDNet在性能上也明显优于现有的伪装目标检测算法,获得的S度量值、最大F度量值、平均F度量值、最大E度量值、平均E度量值和平均绝对误差(mean absolute error,MAE)分别为0.890,0.865,0.824,0.966,0.951和0.020.[结论]研究结果证明了 ZDNet网络模型能够获得很好的伪装昆虫前背景分割结果,有利于提高昆虫识别的性能,也进一步拓宽了伪装目标检测方法的应用范围.
Keyword ：
伪装 伪装 图像分割 图像分割 昆虫 昆虫 深层梯度网络 深层梯度网络 深度学习 深度学习 目标检测 目标检测
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | 范炬臣 , 李小林 , 任昊杰 et al. 一种伪装昆虫图像的前背景自动分割算法 [J]. | 昆虫学报 , 2024 , 67 (8) : 1127-1136 . |
| MLA | 范炬臣 et al. "一种伪装昆虫图像的前背景自动分割算法" . | 昆虫学报 67 . 8 (2024) : 1127-1136 . |
| APA | 范炬臣 , 李小林 , 任昊杰 , 王荣 , 张飞萍 , 黄世国 . 一种伪装昆虫图像的前背景自动分割算法 . | 昆虫学报 , 2024 , 67 (8) , 1127-1136 . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
[目的]昆虫常在色彩、纹理或形态上和背景相似,具有伪装性,识别难度大。本研究旨在探索基于深度学习的伪装昆虫前背景自动分割方法。[方法]将显著性目标检测算法(salient object detection algorithm)、大模型图像分割算法(large-scale model-based image segmentation algorithm)以及伪装目标检测算法(camouflaged object detection algorithm)应用于伪装昆虫数据集,该数据集包括10个类昆虫共1900张图片;并进一步针对现有伪装目标检测算法的不足,提出了一种基于DGNet(deep gradient network)的网络模型改进方法,即ZDNet (zoom-deep gradient network)。在构建该模型时,充分运用图像特征增强、交错图像金字塔、梯度诱导和跳跃式特征融合等技术。利用伪装目标检测公开数据集COD10K与CAMO构建了包含螽斯、蜘蛛等10个类的昆虫图像数据集。并结合迁移学习进行网络训练,将经过训练的模型用于分割伪装昆虫。[结果]现有的伪装目标检测模型用于伪装昆虫前背景分割时,其分割性能明显优于显著目标检测模型、分割大模型。同时,ZDNet在性能上也明显优于现有的伪装目标检测算法,它在S度量值、最大F度量值、平均F度量值、最大E度量值、平均E度量值和平均绝对误差(mean absolute error,MAE)上分别为0.890,0.865,0.824,0.966,0.951和0.020。[结论]研究结果证明了ZDNet网络模型能够获得很好的伪装昆虫前背景分割结果,有利于提高昆虫识别的性能,也进一步拓宽了伪装目标检测方法的应用范围。
Keyword ：
伪装 伪装 图像分割 图像分割 昆虫 昆虫 深层梯度神经网络 深层梯度神经网络 深度学习 深度学习 目标检测 目标检测
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | 范炬臣 , 李小林 , 任昊杰 et al. 一种伪装昆虫图像的前背景自动分割算法——ZDNet [J]. | 昆虫学报 , 2024 , (08) : 2-11 . |
| MLA | 范炬臣 et al. "一种伪装昆虫图像的前背景自动分割算法——ZDNet" . | 昆虫学报 08 (2024) : 2-11 . |
| APA | 范炬臣 , 李小林 , 任昊杰 , 王荣 , 张飞萍 , 黄世国 . 一种伪装昆虫图像的前背景自动分割算法——ZDNet . | 昆虫学报 , 2024 , (08) , 2-11 . |
| Export to | NoteExpress RIS BibTex |
Version ：
Abstract ：
[目的]准确鉴别蝴蝶种类,动态观测蝴蝶群落多样性变化对生境质量评估、生态环境恢复等方面具有重要意义。针对现有蝴蝶识别方法仅依靠整体特征,忽略了局部特征导致识别生态图像能力不足的问题,本研究旨在开发一种Local-Global-VIT细粒度分类算法的蝴蝶识别方法。[方法]本研究以5科200种共计25 279张蝴蝶图像为识别对象,采用多种数据增强方法扩充图像数据;通过视觉Transformer(vision transformer,VIT)层级结构及自注意力机制逐层选择局部令牌并保留至最后一层学习蝴蝶局部判别部位信息;聚合高层全局令牌消除复杂背景干扰;通过对比损失拉大类间距提高区分度。除此之外,使用合理的学习率调整策略和迁移学习方法,优化了模型收敛过程。[结果] Local-Global-VIT算法在大规模细粒度公开数据集Butterfly-200上识别准确率达91.20%,较改进前提升了1.15%,比最优的一般害虫识别算法EfficientNet＿b0和细粒度分类算法TransFG准确率分别高了1.83%和0.64%,F1分值分别提高了1.89%和0.88%。[结论]LocalGlobal-VIT算法以细粒度识别方式有效解决了蝴蝶类内差异大、类间差异小的分类难题,能准确地识别蝴蝶种类,有助于高效评估生境质量。
Keyword ：
vision transformer vision transformer 全局令牌聚合 全局令牌聚合 图像识别 图像识别 局部令牌选择 局部令牌选择 细粒度 细粒度 蝴蝶 蝴蝶
Cite：
Copy from the list or Export to your reference management。
| GB/T 7714 | 李建祥 , 李小林 , 王荣 et al. 基于Local-Global-VIT细粒度分类算法的蝴蝶识别 [J]. | 昆虫学报 , 2024 , (09) : 2-12 . |
| MLA | 李建祥 et al. "基于Local-Global-VIT细粒度分类算法的蝴蝶识别" . | 昆虫学报 09 (2024) : 2-12 . |
| APA | 李建祥 , 李小林 , 王荣 , 张元孜 , 陈淑武 , 张飞萍 et al. 基于Local-Global-VIT细粒度分类算法的蝴蝶识别 . | 昆虫学报 , 2024 , (09) , 2-12 . |
| Export to | NoteExpress RIS BibTex |
Version ：
Export
| Results： |
Selected to |
| Format： |
