检索结果-内蒙古大学图书馆

Classification of Camellia oleifera using a dual recognition strategy based on deep learning

Multimedia Tools and Applications 2025年第13期84卷 12219-12240页

作者： Meng, Zhichao Du, Xiaoqiang Yao, Xiaohua He, Leiying Lin, Lepeng School of Mechanical Engineering Zhejiang Sci-Tech University Hangzhou310018 China Zhejiang Key Laboratory of Transplanting Equipment and Technology Hangzhou310018 China Provincial Key Laboratory of Agricultural Intelligent Sensing and Robotics Zhejiang Province Hangzhou310018 China Research Institute of Subtropical Forestry Chinese Academy of Forestry Hangzhou311400 China School of Youjia Intelligent Manufacturing Hangzhou Vocational & Technical College Hangzhou310018 China

With the rapid growth of the Camellia oleifera industry, counterfeit seedlings have surfaced in the market. Authenticating the genuineness of Camellia oleifera species and cultivars has emerged as a substantial challenge. Present methods for discerning fake Camellia oleifera cultivars depend on manual expertise, which is both time-consuming and labor-intensive. Current deep learning-based crop classification methods can solely be used to classify training objectives and are unable to distinguish between authentic and counterfeit objectives. In light of this situation, it is crucial to establish an efficient recognition strategy to accurately recognize Camellia oleifera species and cultivars and thereby eliminate the distribution of fake seedlings in the market. Firstly, the image datasets of the foreside and backside of Camellia oleifera leaves are constructed in this study. The classification is accomplished using the EfficientNet model, and a dual recognition strategy is employed to verify fake Camellia oleifera. The experiments demonstrate that EfficientNet B4 achieves the best performance among the EfficientNet models. Furthermore, to further enhance the model's performance, the EfficientNet B4 model incorporates the Mish activation function, Multi-class Focal Loss, and SENet V2. The Improved EfficientNet B4 achieves an average T-acc of 97.5%, which is 0.89% higher than the original EfficientNet B4 for the training target classification. Additionally, for the non-training target, the Improved EfficientNet B4 achieves an average T-acc of 93.9%, surpassing the original EfficientNet B4 by 2.22%. The improved model is then compared with 11 other convolutional neural networks to highlight its superior performance. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Evaluating the Intensity of Bitter–Astringent Interactive Perception in Green Tea Based on the Weber–Fechner Law

引用

Chemosensors 2025年第4期13卷

作者： Li, Siying Du, Dongdong Cheng, Shaoming Wei, Zhenbo Department of Biosystems Engineering Zhejiang University 866 Yuhangtang Road Hangzhou310058 China Zhejiang Key Laboratory of Intelligent Sensing and Robotics for Agriculture Department of Biosystems Engineering and Food Science Zhejiang University Hangzhou310058 China

This study aimed to quantify the bitter–astringent interactive perceived intensity (BAIPI) of green tea based on the Weber–Fechner law by two-alternative forced choice test. Caffeine, (-)-epigallocatechin-3-gallate (EGCG), and Quercetin-3-o-rutinoside (Que-rut) were identified as the key bitter–astringent compounds. Then, BAIPI values were assessed based on the concentration ratios and ranges of caffeine, EGCG, and Que-rut in green tea according to the Weber–Fechner law. The BAIPI values were further fitted to a logarithmic curve: (Formula presented.). The constant terms (Formula presented.), (Formula presented.), and (Formula presented.) were derived from the ratio of caffeine to EGCG, as the slope and X-intercept of the curves correlated exponentially with this ratio. The initial BAIPI value (IV) was calculated as: (Formula presented.). Validation showed a standard error of 1.5865 between estimated and actual BAIPI values. This method enables the estimation of bitter–astringent intensity in green tea by analyzing the concentrations of caffeine, EGCG, and Que-rut. © 2025 by the authors.

关键词： Tea

来源：评论

学校读者我要写书评

暂无评论

Recognizing and Localizing Chicken Behaviors in Videos Based on Spatiotemporal Feature Learning

SSRN

引用

SSRN 2025年

作者： Hu, Yilei Xu, Jinyang Gou, Zhichao Cui, Di College of Biosystems Engineering and Food Science Zhejiang University Hangzhou310058 China Zhejiang Key Laboratory of Intelligent Sensing and Robotics for Agriculture Hangzhou310058 China 866 Yuhangtang Rd. Zhejiang Hangzhou310058 China

Timely acquisition of chicken behavioral information is crucial for assessing their health status and production performance. Video-based behavior recognition methods have emerged as a primary technique for obtaining such information due to their accuracy and robustness. Video-based models generally predicted a single behavior from a single video segment of a fixed duration. However, during periods of high activity in poultry, behavior transition may occur within the video segment, and existing models often failed to effectively capture such transition. The limitation indicated the insufficient temporal resolution of video-based behavior recognition models. This study proposed a chicken behavior recognition and localization model, CBLFormer, based on spatiotemporal feature learning. The model was designed to recognize behaviors occurring before and after transition in the video segment and to localize the corresponding time interval for each behavior. An improved transformer block, Cascade Encoder-Decoder Network (CEDNet), a transformer-based head, and the Weighted Distance Intersection over Union (WDIoU) loss were integrated into CBLFormer to enhance the model's ability to distinguish between different behavior categories and locate behavior boundaries. For the training and testing of CBLFormer, the dataset was created by collecting videos from 320 chickens across different ages and rearing densities. The results showed that the CBLFormer achieved an mAP@0.5:0.95 of 98.34% on the test set. The integration of CEDNet contributed the most to the performance improvement of the CBLFormer. Compared to other temporal behavior localization models, CBLFormer struck a favorable balance between model complexity and performance, with a computational complexity of 0.57 GFLOPs and a parameter size of 16.14 MB. Furthermore, visualization results confirmed that the model effectively captured the behavioral boundaries of chickens and correctly recognized behavior categories. The pr

关键词： Ability testing

来源：评论

学校读者我要写书评

暂无评论

Integrated and confinable paper-based chip biosensor for all-in-one colorimetric detection of aflatoxin B1

引用

Biosensors and Bioelectronics 2025年 282卷

作者： Moshirian-Farahi, Sareh Sadat Rahmanian, Hamidreza Wu, Jianxiong Huang, Qiao Sun, Yuxin Ma, Tongtong Wu, Huajun Fu, Yingchun Cheng, Kejun Pan, Jinming College of Biosystems Engineering and Food Science Zhejiang Key Laboratory of Intelligent Sensing and Robotics for Agriculture Zhejiang University Hangzhou310058 China School of Pharmaceutical Sciences Fuchun Campus Zhejiang Chinese Medical University Zhejiang Province Hangzhou311402 China Qingyuan County Sanheyuan Agriculture Development Co. ltd. Qingyuan China

This study presents the development of a rapid and highly sensitive colorimetric biosensor for the detection of aflatoxin B1 (AFB1), utilizing a peroxidase-mimetic nanozyme in combination with a membrane-confined signal amplification strategy. The biosensor platform incorporates AFB1-specific aptamer-labeled Fe3+-doped mesoporous carbon nanospheres, which hybridize with a complementary strand immobilized on paper nanofibers. Upon binding with AFB1, the nanozyme detaches and is subsequently removed through a washing step. The remaining nanozyme catalyzes the oxidation of 3,3′,5,5′-tetramethylbenzidine in the presence of H2O2, generating a blue-colored signal. To facilitate real-time, quantitative signal analysis, a smartphone-based imaging strategy is employed. In contrast to conventional open-solution detection methods, this approach confines the catalytic reaction and its products within the membrane, thereby enhancing the signal intensity. The integration of the nanozyme's high catalytic efficiency with the signal amplification enabled by membrane confinement results in superior sensitivity, stability, and operational simplicity. The biosensor demonstrates a broad detection range from 0.01 to 1000 ng mL−1 and an exceptionally low detection limit of 3.9 pg mL−1, outperforming most analogous systems. Additionally, the biosensor exhibits excellent performance in complex sample matrices, such as chicken feed and traditional Chinese medicinal herbs. Through the combination of smartphone imaging for quantitative analysis and rapid, visual detection, this platform provides a versatile, user-friendly tool for real-time, on-site food safety monitoring and health surveillance. © 2025 Elsevier B.V.

关键词： Aflatoxins

来源：评论

学校读者我要写书评

暂无评论

Research on Design and Experiment of Rear-Drive Apple Harvesting Robotic Arm Based on Obstacle Avoidance Posture Conditions

SSRN

引用

SSRN 2025年

作者： Zhao, Xiong Liang, Hongxuan Liu, Ziwei Wang, Lei Du, Xiaoqiang Faculty of Mechanical Engineering and Automation Zhejiang Sci-Tech University Hangzhou310018 China Key Laboratory of Transplanting Equipment and Technology of Zhejiang Province Hangzhou310018 China Provincial Key Laboratory of Agricultural Intelligent Sensing and Robotics Zhejiang Province Hangzhou310018 China Key Laboratory of Agricultural Equipment for Hilly and Mountainous Areas in Southeastern China Ministry of Agriculture and Rural Affairs Hangzhou310018 China

Aiming at the problems of small space coverage of fruit area and high equipment price in apple picking operation using traditional industrial manipulator, this paper designs a driving rear-mounted apple picking manipulator based on obstacle avoidance posture condition in combination with the growth environment and picking demand of fruit. Based on the investigation of the environment of Red Fuji orchard in Weihai, Shandong Province, a design scheme of driving rear-mounted apple picking manipulator suitable for large space is proposed, and the modular design of each part of the manipulator is carried out. Aiming at the problem that the horizontal picking attitude of the three-claw end-effector is easy to interfere with the 'connecting branch', an obstacle avoidance attitude evaluation function is proposed to solve a set of adaptive obstacle avoidance picking attitude angles, which are 43° for upward picking and 49° for downward picking. Finally, the designed mechanical structure is processed and assembled, and the motion control system of the manipulator is designed based on the CANopen communication sub-protocol DS301 and the control system sub-protocol DS402. The final picking test demonstrated a success rate of 82.21% for upward picking with an obstacle avoidance posture angle of 43° and 57.9% for downward picking with an obstacle-avoidance posture angle of 49°, validating the rationality of the obstacle avoidance posture angles. In multi-fruit picking tests, the overall success rate reached 86%, with an average harvesting rate of 11.5 seconds per fruit, achieving continuous harvesting functionality for the robotic arm. © 2025, The Authors. All rights reserved.

关键词： Robotic arms

来源：评论

学校读者我要写书评

暂无评论

Boosting photocatalytic desulfurization efficiency via in-situ Ti-MOF disassembly and reconstruction strategies on S-scheme M-TiO2@Cu2S heterojunction

引用

Chemical Engineering Journal 2025年 514卷

作者： Wang, Leiping Zhang, Qiyu Man, Zun Meng, Yang Wang, Wei Dai, Xiaorong Xiao, Hang Liu, Dezhao College of Biosystems Engineering and Food Science Zhejiang University Hangzhou310058 China Key Laboratory of Urban Environment and Health Ningbo Observation and Research Station Institute of Urban Environment Chinese Academy of Sciences Xiamen361021 China Ningbo Key Laboratory of Urban Environmental Pollution Control CAS Haixi Industrial Technology Innovation Center in Beilun Ningbo315830 China Key Laboratory of Equipment and Informatization in Environment Controlled Agriculture from Ministry of Agriculture and Rural Affairs of China Key Laboratory of Intelligent Equipment and Robotics for Agriculture of Zhejiang Province Hangzhou310058 China College of Biological and Environmental Sciences Zhejiang Wanli University Ningbo315100 China

Space charge transfer, which can be provided by heterogeneous structure constructed based on metal sulfides, is the key to efficient photocatalytic sulfur removal processes. However, how to design an ideal interface for efficient photocatalytic carrier transfer/separation remains critical for improving photocatalytic performance. To address this issue, S-scheme M-TiO2@Cu2S heterojunctions (TCMx) with dense interfaces were constructed. The key to the strategy in this study was the induction of uniform Cu loading on the precursor Ti-MOF (NH2-MIL-125, TM), based on the affinity of amino groups for metals, and the subsequent simultaneous MOF pyrolysis and Cu-site vulcanization. The preservation of the pore structure and spatially dispersed catalytic sites promotes the adsorption and degradation of hydrogen sulfide, enabling TCM2 to exhibit excellent adsorption performance and long-term cycling stability. Importantly, photovoltaic performance characterizations confirm that band-aligned M-TiO2@Cu2S provides abundant active centers, enhanced light absorption in the visible region and accelerated interfacial charge transfer. In situ X-ray photoelectron spectroscopy (XPS) and density-functional theory (DFT) calculations further provide important information for explaining the formation of the internal electric field as well as for understanding S-scheme charge transfer. The results showed that TCM2 exhibited the highest H2S removal capacity, with a dark adsorption capacity of H2S 17 times higher than that of TM, while the photocatalytic degradation efficiency was improved by 57 % and 85 % compared to Cu2S and M-TiO2, respectively. This study provides insights into the rational design of S-scheme heterojunctions for efficient carrier transport and their application in environmental remediation. © 2025 Elsevier B.V.

关键词： Titanium dioxide

来源：评论

学校读者我要写书评

暂无评论

Explicable Attention Mechanism for Diameter Correction in Predicting Soluble Solids Content of Fruits

SSRN

引用

SSRN 2025年

作者： Sun, Zhizhong Yang, Jie Zheng, Yingjie Liu, Penghui Ma, Chan Hu, Dong Ying, Yibin Xie, Lijuan College of Biosystems Engineering and Food Science Zhejiang University Hangzhou310058 China College of Chemistry and Materials Engineering Zhejiang A&F University Hangzhou311300 China Key Laboratory of Intelligent Equipment and Robotics for Agriculture of Zhejiang Province Hangzhou310058 China College of Optical Mechanical and Electrical Engineering Zhejiang Agriculture and Forestry University Hangzhou311300 China Department of Bioproducts and Biosystems Engineering University of Minnesota Saint PaulMN55108 United States

Varying fruit diameters significantly influences the optical path length within the sample, thereby impacting the predictive precision of the soluble solids content (SSC) of the fruits. This study develops a deep learning model, the attention convolutional neural network (Attention-CNN), which integrates diameter and Vis/NIR data to enhance SSC prediction for apples and pears. By employing an attention mechanism, the model autonomously adjusts the weight coefficients of each wavelength relative to diameter, correcting for diameter-related influences. Results showed the Attention-CNN model reduced RMSE in predicting SSC by 43.5% for apples and 18.2% for pears compared to PLS model, and by 9.7% and 6.1%, respectively, compared to 1D-CNN. Visualization through the SHAP method highlighted the model’s ability to capture diameter information more effectively than PLS and 1D-CNN. In summary, the use of attention mechanisms can enhance the performance of convolutional neural networks-based models in predicting fruit SSC. © 2025, The Authors. All rights reserved.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

A colorimetric-bionic sensor for multimodal molecular sensing using hierarchical metamaterials

引用

Device 2025年第5期3卷

作者： Geng, Wenhui Zhou, Ying Huang, Kang Ying, Yibin Xie, Lijuan College of Biosystems Engineering and Food Science Zhejiang University 866 Yuhangtang Road Hangzhou 310058 China The National Key Laboratory of Agricultural Equipment Technology Beijing 100083 China Zhejiang Key Laboratory of Intelligent Sensing and Robotics for Agriculture Hangzhou 310058 China Hangzhou Customs Technology Center 398 Jianshe San Road Xiaoshan District Hangzhou City China Department of Biological Systems Engineering Washington State University Pullman 99164 WA United States

Metamaterials can enhance a local electric field for molecular sensing. However, the difficulty in specifically recognizing target molecules and integrating multiple functionalities restricts their development. Here, we present a hierarchical metamaterial-based colorimetric-bionic sensor (HMm-CBs) using a layer-by-layer assembly method for multimodal molecular sensing. The CBs layers offer molecular recognition ability by enabling the specific perception of analytes, and the HMm layers provide coupling of vibrations between analyte molecules at the terahertz (THz) frequency. The HMm-CBs have broadband adaptability in controlling electromagnetic waves from optical to THz frequency. As a proof of principle, we designed a sensor for analyzing tetracycline hydrochlorides (TCHs). The sensor, in conjunction with a back propagation neural network (BPNN), identifies four types of TCHs. As a demonstration, a multichannel indicator is created for real-time monitoring of shrimp freshness and volatile metabolites. © 2025 The Author(s)

关键词： antibiotics carbon nanotubes chemometric algorithms colorimetric sensor array DTI-3: Develop metamaterials

来源：评论

学校读者我要写书评

暂无评论

DCDNet: A deep neural network for dead chicken detection in layer farms

引用

Computers and Electronics in agriculture 2025年 237卷

作者： Dihua Wu Yibin Ying Mingchuan Zhou Jinming Pan Di Cui College of Biosystems Engineering and Food Science Zhejiang University Hangzhou 310058 Zhejiang China Key Laboratory of Intelligent Equipment and Robotics for Agriculture of Zhejiang Province Science Technology Department of Zhejiang Province China Key Laboratory of On-Site Processing Equipment for Agricultural Products Ministry of Agriculture and Rural Affairs China

Detecting deceased layers is vital in farm inspections. Manual inspections are inefficient and pose bio-security risks. Deep learning excels on large datasets, yet accurate dead chicken detection is challenging due to data scarcity, imbalance, visual similarity, and irregular morphology. To achieve desirable performance in distinguishing dead and normal chickens, a novel deep neural network, DCDNet, was proposed in this study. The pipeline consisted of the following three modules: Poisson fusion-based data augmentation (PFDA) module seamlessly integrated the area of the deceased layer into the new background, generating a more realistic image that alleviates sample scarcity; the designed DCDNet was utilized to accurately identify dead and normal layers by extracting and fusing features more efficiently, thus better suiting their irregular body shapes; the non-monotonic dynamic focusing (NDF) sliding weight loss function was proposed to enhance the contribution of difficult samples in model training flexibly, reducing bias caused by unbalanced data. Extensive experiments have been conducted on our dead-chicken dataset constructed on a commercial farm. The results revealed that the proposed method achieved a mean average precision (mAP) of 97.5%, outperforming the state-of-the-art methods reported thus far. Moreover, the average precision (AP) difference between dead and normal chickens is only 0.1%. The proposed dead chicken detection approach, based on DCDNet, was effective in dealing with sample scarcity and dataset imbalance. This may provide some reference for other researchers on other similar tasks.

关键词：

来源：评论

学校读者我要写书评

暂无评论

BAN: A Universal Paradigm for Cross-Scene Classification Under Noisy Annotations From RGB and Hyperspectral Remote sensing Images

引用

IEEE Transactions on Geoscience and Remote sensing 2025年 63卷

作者： Chen, Wentang Wen, Yibin Zheng, Juepeng Huang, Jianxi Fu, Haohuan Sun Yat-sen University School of Artificial Intelligence Zhuhai519082 China Zhejiang University State Key Laboratory of Fluid Power and Mechatronic Systems Engineering Research Center for Design Engineering and Digital Twin of Zhejiang Province School of Mechanical Engineering Robotics Institute Hangzhou310027 China Southwest Jiaotong University Faculty of Geosciences and Engineering Chengdu610032 China China Agricultural University College of Land Science and Technology Beijing100107 China Ministry of Agriculture and Rural Affairs Key Laboratory of Remote Sensing for Agri-Hazards Beijing100125 China Tsinghua University Tsinghua Shenzhen International Graduate School Shenzhen518071 China Tsinghua University Ministry of Education Key Laboratory for Earth System Modeling Department of Earth System Science Beijing100190 China National Supercomputer Center in Shenzhen Shenzhen518071 China

While domain adaptation (DA) methods have made significant strides in remote sensing community, most current works assume that the source domain labels are accurate. However, limited emphasis has been placed on the scenario where source data are mislabeled with noisy annotations, which is more common in real applications and referred to as noisy DA (NDA). This article formulates remote sensing cross-scene classification on NDA scenarios and proposes a novel network called bilateral adaptation network (BAN), which consists of two parts: 1) forward learning (FL), which utilizes a model learning from the noisy source domain and transfers knowledge to target domain;and 2) backward learning (BL), which utilizes a dual model to acquire knowledge from the target domain and transfer it to source domain. We conduct two parts alternately and adopt a symmetrical Kullback-Leibler (KL) loss to align predictions of the model and its dual model in the same domain. This interactive strategy is able to explore bilateral relationships between domains, implicitly reducing label noise in the source domain. In addition, BAN could serve as a universal paradigm to not only improve the existing NDA methods but also enhance recent DA approaches. Comprehensive evaluations on three publicly available RGB-band remote sensing datasets and two hyperspectral datasets validate the superior effectiveness of our proposed BAN. BAN improves the average accuracy by 6.70%-15.70% on RGB datasets and overall accuracy (OA) by 1.36%-3.14% on hyperspectral datasets with flip-20% noise compared to other state-of-the-art DA and NDA approaches. Promising results indicate the potential of our approach in tackling more general and practical problems with noisy source domain. © 2025 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：