2024
2024
-
Record 277 of
Title:Extending UWOC System Applications through Photon Transmission Dynamics Study in Harbor Waters
Author Full Names:Chang, Chang; Han, Xiaotian; Li, Guangying; Li, Peng; Nie, Wenchao; Liao, Peixuan; Li, Cong; Wang, Wei; Xie, XiaopingSource Title:APPLIED SCIENCES-BASELLanguage:EnglishDocument Type:ArticleKeywords Plus:PHASE FUNCTION; TIME; LINKAbstract:Underwater wireless optical communication (UWOC) in harbor waters can facilitate real-time monitoring underwater instruments for environmental monitoring, underwater inspection, and maintenance tasks. This study delves into the complex dynamics of UWOC in four distinct harbor waters. The research employs Monte Carlo method incorporated with Fournier-Forand scattering phase function for simulating photon transmission. Key parameters such as the Transmitted full divergence angle, received aperture, and Field of View (FOV) are meticulously evaluated for their impact on power loss and time delay spread. Notably, the normalized power loss and time delay spread are found to be more significantly affected by communication distance than water quality, and the traditional Beer-Lambert law is ineffective in harbor water. The power loss of Harbor II, III, and IV are found to be 14.00 dB, 31.59 dB, and 41.59 dB lower than that of Harbor I, and the time delay spread of Harbor II, III, and IV is 30.56%, 9.67%, and 0.49% times that of the Harbor I under certain conditions. In addition, increasing the received aperture and FOV, particularly over longer distance, make little contribution to reduce the power loss and mitigate the time delay spread. Based on the fixed transmitted full divergence angle, the most applicable received FOV range is 1-3.2 rad, and the most ideal received aperture is 0.15-0.4 m. Under these conditions, the variation in normalized power loss is less than 2 dB. Additionally, the time delay spread remains within the same order of magnitude with the attenuation length (AL) held constant. These conclusions hold substantial technical relevance for the engineering design of UWOC systems in harbor waters.Addresses:[Chang, Chang; Han, Xiaotian; Li, Guangying; Li, Peng; Nie, Wenchao; Liao, Peixuan; Wang, Wei; Xie, Xiaoping] Chinese Acad Sci, Xian Inst Opt & Precis Mech, State Key Lab Transient Opt & Photon, Xian 710119, Peoples R China; [Chang, Chang; Han, Xiaotian; Liao, Peixuan; Xie, Xiaoping] Univ Chinese Acad Sci CAS, Beijing 100049, Peoples R China; [Li, Cong] CAST, Inst Telecommun & Nav Satellites, Beijing 100094, Peoples R ChinaAffiliations:State Key Laboratory of Transient Optics & Photonics; Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Chinese Academy of Sciences; University of Chinese Academy of Sciences, CASPublication Year:2024Volume:14Issue:6Article Number:2493DOI Link:http://dx.doi.org/10.3390/app14062493數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001191597300001 -
Record 278 of
Title:Saturable absorption properties and ultrafast photonics applications of HfS3
Author Full Names:Li, Lu; Xue, Ze; Pang, Lihui; Xiao, Xusheng; Yang, Huiran; Zhang, Jinniu; Zhang, Yarning; Zhao, Qiyi; Liu, WenjunSource Title:OPTICS LETTERSLanguage:EnglishDocument Type:ArticleAbstract:In this Letter, we focus on investigating the ultrafast photonics applications of two -layer HfS3 nanosheets. We prepared two -layer HfS3 nanosheets and carried out experiments to study their nonlinear saturable absorption properties. The results showed that the two -layer HfS3-based saturable absorber exhibited a modulation depth of 16.8%. Additionally, we conducted theoretical calculations using first principles to estimate the structural and electronic band properties of the two -layer HfS3 material. Furthermore, we utilized the two -layer HfS3 materials as SAs in an erbiumdoped fiber cavity to generate mode -locked laser pulses. We measured a repetition frequency of 8.74 MHz, a pulse duration of 540 fs, and a signal-to-noise ratio of 77 dB. Overall, our findings demonstrate that the two -layer HfS3 material can serve as a reliable saturable absorber, possessing properties comparable to currently used two-dimensional materials. This expands the application fields of HfS3 materials and highlights their potential for advanced optoelectronic devices. (c) 2024 Optica Publishing GroupAddresses:[Li, Lu; Xue, Ze; Yang, Huiran; Zhang, Jinniu; Zhang, Yarning; Zhao, Qiyi] Xian Univ Posts & Telecommun, Sch Sci, Xian 710121, Peoples R China; [Pang, Lihui] Xi An Jiao Tong Univ, Affiliated Hosp 1, Shaanxi Prov Ctr Regenerat Med & Surg Engn, Xian 710061, Peoples R China; [Xiao, Xusheng] Chinese Acad Sci, Xian Inst Opt & Precis Mech, State Key Lab Transient Opt & Photon, Xian 710119, Shaanxi, Peoples R China; [Liu, Wenjun] Beijing Univ Posts & Telecommun, Sch Sci, State Key Lab Informat Photon & Opt Commun, Beijing 100876, Peoples R ChinaAffiliations:Xi'an University of Posts & Telecommunications; Xi'an Jiaotong University; Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; State Key Laboratory of Transient Optics & Photonics; Beijing University of Posts & TelecommunicationsPublication Year:2024Volume:49Issue:5Start Page:1293End Page:1296DOI Link:http://dx.doi.org/10.1364/OL.513573數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001202790500002 -
Record 279 of
Title:Computational Imaging: The Next Revolution for Biophotonics and Biomedicine
Author Full Names:Pan, An; Yao, Baoli; Zuo, Chao; Liu, Fei; Yang, Jiamiao; Cao, LiangcaiSource Title:CELLSLanguage:EnglishDocument Type:ArticleAddresses:[Pan, An; Yao, Baoli] Chinese Acad Sci, State Key Lab Transient Opt & Photon, Xi An Inst Opt & Precis Mech, Xian 710119, Peoples R China; [Zuo, Chao] Nanjing Univ Sci & Technol, Dept Elect & Opt Engn, Smart Computat Imaging Lab SCILab, Nanjing 210094, Peoples R China; [Liu, Fei] Xidian Univ, Sch Optoelect Engn, Xian 710071, Peoples R China; [Yang, Jiamiao] Shanghai Jiao Tong Univ, Dept Instrument Sci & Engn, Shanghai 200240, Peoples R China; [Cao, Liangcai] Tsinghua Univ, Dept Precis Instruments, Beijing 100084, Peoples R ChinaAffiliations:State Key Laboratory of Transient Optics & Photonics; Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Nanjing University of Science & Technology; Xidian University; Shanghai Jiao Tong University; Tsinghua UniversityPublication Year:2024Volume:13Issue:5Article Number:433DOI Link:http://dx.doi.org/10.3390/cells13050433數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001183318400001 -
Record 280 of
Title:High linear sensitivity humidity and temperature sensing based on chiral long-period fiber grating
Author Full Names:Ren, Kaili; Hu, Jiayue; Jia, Aochi; Xi, Jiawei; Ren, Yuchong; Yan, Xuewen; Li, Jinze; Dong, Jun; Liu, JihongSource Title:OPTICAL ENGINEERINGLanguage:EnglishDocument Type:ArticleKeywords Plus:FABRICATION; SENSORSAbstract:A highly sensitive, reversible, and linear sensor, exhibiting excellent stability in response to temperature and humidity, has been successfully proposed and demonstrated for the first time. This sensor is achieved by wrapping a polyvinyl alcohol/graphene nanofiber film onto a chiral long-period fiber grating (CLPG), which is fabricated by periodically twisting single mode fiber. In the experiment, the CLPG sensor demonstrates a temperature sensitivity of 74 pm/degrees C, which is approximately twice as high as that of conventional fiber grating sensors. Note that, by wrapping the graphene nanofiber film on CLPG, the temperature sensitivity of the sensor is up to 115.23 pm/degrees C in the range of 30 degrees C to 75 degrees C. In addition, CLPG using for humidity sensing is first demonstrated. The humidity sensitivity measures -9.92 pm/%RH with linearity of 0.995 during a change from 40%RH to 80%RH. In comparison to other humidity sensors, the sensitivity of the CLPG is comparable, whereas its sensing linearity stands out notably above the rest. The results show that CLPG has the characteristics of simple fabrication, easy combination with materials, stable performance, and high sensitivity and holds significant development potential in optical fiber sensing application fields. (c) 2024 Society of Photo-Optical Instrumentation Engineers (SPIE)Addresses:[Ren, Kaili; Hu, Jiayue; Jia, Aochi; Ren, Yuchong; Yan, Xuewen; Dong, Jun; Liu, Jihong] Xian Univ Posts & Telecommun, Sch Elect Engn, Xian, Peoples R China; [Ren, Kaili] Chinese Acad Sci, Xian Inst Opt & Precis Mech, State Key Lab Transient Opt & Photon, Xian, Peoples R China; [Xi, Jiawei; Li, Jinze] Xidian Univ, Sch Optoelect Engn, Xian, Peoples R ChinaAffiliations:Xi'an University of Posts & Telecommunications; State Key Laboratory of Transient Optics & Photonics; Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Xidian UniversityPublication Year:2024Volume:63Issue:3Article Number:36103DOI Link:http://dx.doi.org/10.1117/1.OE.63.3.036103數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001230092500034 -
Record 281 of
Title:Hybrid Grid Pattern Star Identification Algorithm Based on Multi-Calibration Star Verification
Author Full Names:Shen, Chao; Ma, Caiwen; Gao, Wei; Wang, YuanboSource Title:SENSORSLanguage:EnglishDocument Type:ArticleKeywords Plus:CYCLIC FEATURES; ROBUSTAbstract:In order to solve the star identification problem in the lost space mode for scientific cameras with small fields of view and higher instruction magnitudes, this paper proposes a star identification algorithm based on a hybrid grid pattern. The application of a hybrid pattern generated by multi-calibration stars in the initial matching enables the position distribution features of neighboring stars around the main star to be more comprehensively described and avoids the interference of position noise and magnitude noise as much as possible. Moreover, calibration star filtering is adopted to eliminate incorrect candidates and pick the true matched navigation star from candidate stars in the initial match. Then, the reference star image is utilized to efficiently verify and determine the final identification results of the algorithm via the nearest principle. The performance of the proposed algorithm in simulation experiments shows that, when the position noise is 2 pixels, the identification rate of the algorithm is 96.43%, which is higher than that of the optimized grid algorithm by 2.21% and the grid algorithm by 4.05%; when the magnitude noise is 0.3 mag, the star identification rate of the algorithm is 96.45%, which is superior to the optimized grid algorithm by 2.03% and to the grid algorithm by 3.82%. In addition, in the actual star image test, star magnitude values of <= 12 mag can be successfully identified using the proposed algorithm.Addresses:[Shen, Chao; Ma, Caiwen; Gao, Wei; Wang, Yuanbo] Chinese Acad Sci, Xian Inst Opt & Precis Mech, Xian 710119, Peoples R ChinaAffiliations:Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CASPublication Year:2024Volume:24Issue:5Article Number:1661DOI Link:http://dx.doi.org/10.3390/s24051661數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001183048800001 -
Record 282 of
Title:Adaptive Dual Aggregation Network with Normalizing Flows for Low-Light Image Enhancement
Author Full Names:Wang, Hua; Cao, Jianzhong; Huang, JijiangSource Title:ENTROPYLanguage:EnglishDocument Type:ArticleKeywords Plus:HISTOGRAM EQUALIZATIONAbstract:Low-light image enhancement (LLIE) aims to improve the visual quality of images taken under complex low-light conditions. Recent works focus on carefully designing Retinex-based methods or end-to-end networks based on deep learning for LLIE. However, these works usually utilize pixel-level error functions to optimize models and have difficulty effectively modeling the real visual errors between the enhanced images and the normally exposed images. In this paper, we propose an adaptive dual aggregation network with normalizing flows (ADANF) for LLIE. First, an adaptive dual aggregation encoder is built to fully explore the global properties and local details of the low-light images for extracting illumination-robust features. Next, a reversible normalizing flow decoder is utilized to model real visual errors between enhanced and normally exposed images by mapping images into underlying data distributions. Finally, to further improve the quality of the enhanced images, a gated multi-scale information transmitting module is leveraged to introduce the multi-scale information from the adaptive dual aggregation encoder into the normalizing flow decoder. Extensive experiments on paired and unpaired datasets have verified the effectiveness of the proposed ADANF.Addresses:[Wang, Hua; Cao, Jianzhong; Huang, Jijiang] Chinese Acad Sci, Xian Inst Opt & Precis Mech, Xian 710119, Peoples R China; [Wang, Hua] Univ Chinese Acad Sci, Beijing 100049, Peoples R ChinaAffiliations:Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Chinese Academy of Sciences; University of Chinese Academy of Sciences, CASPublication Year:2024Volume:26Issue:3Article Number:184DOI Link:http://dx.doi.org/10.3390/e26030184數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001191588100001 -
Record 283 of
Title:A Lightweight Remote Sensing Aircraft Object Detection Network Based on Improved YOLOv5n
Author Full Names:Wang, Jiale; Bai, Zhe; Zhang, Ximing; Qiu, YuehongSource Title:REMOTE SENSINGLanguage:EnglishDocument Type:ArticleKeywords Plus:IMAGESAbstract:Due to the issues of remote sensing object detection algorithms based on deep learning, such as a high number of network parameters, large model size, and high computational requirements, it is challenging to deploy them on small mobile devices. This paper proposes an extremely lightweight remote sensing aircraft object detection network based on the improved YOLOv5n. This network combines Shufflenet v2 and YOLOv5n, significantly reducing the network size while ensuring high detection accuracy. It substitutes the original CIoU and convolution with EIoU and deformable convolution, optimizing for the small-scale characteristics of aircraft objects and further accelerating convergence and improving regression accuracy. Additionally, a coordinate attention (CA) mechanism is introduced at the end of the backbone to focus on orientation perception and positional information. We conducted a series of experiments, comparing our method with networks like GhostNet, PP-LCNet, MobileNetV3, and MobileNetV3s, and performed detailed ablation studies. The experimental results on the Mar20 public dataset indicate that, compared to the original YOLOv5n network, our lightweight network has only about one-fifth of its parameter count, with only a slight decrease of 2.7% in mAP@0.5. At the same time, compared with other lightweight networks of the same magnitude, our network achieves an effective balance between detection accuracy and resource consumption such as memory and computing power, providing a novel solution for the implementation and hardware deployment of lightweight remote sensing object detection networks.Addresses:[Wang, Jiale; Bai, Zhe; Zhang, Ximing; Qiu, Yuehong] Xian Inst Opt & Precis Mech CAS, Xian 710119, Peoples R China; [Wang, Jiale] Univ Chinese Acad Sci, Beijing 100049, Peoples R ChinaAffiliations:Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Chinese Academy of Sciences; University of Chinese Academy of Sciences, CASPublication Year:2024Volume:16Issue:5Article Number:857DOI Link:http://dx.doi.org/10.3390/rs16050857數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001183018500001 -
Record 284 of
Title:Design of Mantis-Shrimp-Inspired Multifunctional Imaging Sensors with Simultaneous Spectrum and Polarization Detection Capability at a Wide Waveband
Author Full Names:Wang, Tianxin; Wang, Shuai; Gao, Bo; Li, Chenxi; Yu, WeixingSource Title:SENSORSLanguage:EnglishDocument Type:ArticleKeywords Plus:COMMUNICATION; TRANSMISSION; EVOLUTION; SYSTEM; CAMERA; SIGNAL; LIGHTAbstract:The remarkable light perception abilities of the mantis shrimp, which span a broad spectrum ranging from 300 nm to 720 nm and include the detection of polarized light, serve as the inspiration for our exploration. Drawing insights from the mantis shrimp's unique visual system, we propose the design of a multifunctional imaging sensor capable of concurrently detecting spectrum and polarization across a wide waveband. This sensor is able to show spectral imaging capability through the utilization of a 16-channel multi-waveband Fabry-Perot (FP) resonator filter array. The design incorporates a composite thin film structure comprising metal and dielectric layers as the reflector of the resonant cavity. The resulting metal-dielectric composite film FP resonator extends the operating bandwidth to cover both visible and infrared regions, specifically spanning a broader range from 450 nm to 900 nm. Furthermore, within this operational bandwidth, the metal-dielectric composite film FP resonator demonstrates an average peak transmittance exceeding 60%, representing a notable improvement over the metallic resonator. Additionally, aluminum-based metallic grating arrays are incorporated beneath the FP filter array to capture polarization information. This innovative approach enables the simultaneous acquisition of spectrum and polarization information using a single sensor device. The outcomes of this research hold promise for advancing the development of high-performance, multifunctional optical sensors, thereby unlocking new possibilities in the field of optical information acquisition.Addresses:[Wang, Tianxin; Wang, Shuai; Gao, Bo; Li, Chenxi; Yu, Weixing] Chinese Acad Sci, Xian Inst Opt & Precis Mech, Key Lab Spectral Imaging Technol, Xian 710119, Peoples R China; [Wang, Tianxin; Yu, Weixing] Univ Chinese Acad Sci, Ctr Mat Sci & Optoelect Engn, Beijing 100049, Peoples R ChinaAffiliations:Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Chinese Academy of Sciences; University of Chinese Academy of Sciences, CASPublication Year:2024Volume:24Issue:5Article Number:1689DOI Link:http://dx.doi.org/10.3390/s24051689數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001183065400001 -
Record 285 of
Title:Location and spectral extraction algorithm for a static broadband snapshot imaging spectrometer
Author Full Names:Wang, Zhipeng; Yang, Qinghua; Li, Bingbin; Wang, WeiqiangSource Title:OPTICAL ENGINEERINGLanguage:EnglishDocument Type:ArticleKeywords Plus:MICHELSON INTERFEROMETER; RESOLUTION; FILTERAbstract:A location algorithm and a spectral extraction algorithm for a static broadband snapshot imaging spectrometer (SBSIS) are presented. The high-energy target is dispersed into a V-shaped pseudo-image (VPI) in the focal plane of the SBSIS. The location algorithm accurately calculates the target azimuth based on the one-to-one mapping relationship between the intersection position of the extension lines of the two arms of the VPI and the azimuth of the target. The spectral extraction algorithm is described to extract the characteristic spectrum of the target based on the azimuth angle of the target and the VPI.Addresses:[Wang, Zhipeng; Yang, Qinghua; Li, Bingbin] Xidian Univ, Sch Optoelect Engn, Xian, Peoples R China; [Wang, Weiqiang] Chinese Acad Sci, Xian Inst Opt & Precis Mech, State Key Lab Transient Opt & Photon, Xian, Peoples R ChinaAffiliations:Xidian University; State Key Laboratory of Transient Optics & Photonics; Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CASPublication Year:2024Volume:63Issue:3DOI Link:http://dx.doi.org/10.1117/1.OE.63.3.035103數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001230092500012 -
Record 286 of
Title:On-orbit calibration of space camera lens distortion using a single image
Author Full Names:Zhang, Gaopeng; Wang, Feng; Zhang, Guangdong; Zhang, Zhe; Du, Hubing; Zhao, Zixin; Wang, Changqing; Cao, Jianzhong; Zhao, Jingwei; Li, Yanjie; Lu, RongSource Title:OPTICS AND LASERS IN ENGINEERINGLanguage:EnglishDocument Type:ArticleKeywords Plus:MODELAbstract:Since space cameras need to withstand the harsh mechanical and thermal conditions in the space environment for a long time, it is necessary to calibrate them in orbit. However, existing calibration methods have various disadvantages, making them impossible to use in orbit. To address this problem, we present an on-orbit calibration of space camera lens distortion with the vanishing points obtained from a single image of the solar panel. First, we propose a parallel-line-extraction method based on collinear constraints to obtain the parallel lines. Then, we train the optimal vanishing point using the common point constraint method. Using the optimal vanishing point, we establish the optimization function of lens distortion based on vanishing point consistency. Finally, we present an improved genetic algorithm to solve the optimization function. Simulations and experiments show that the proposed method is flexible and robust.Addresses:[Zhang, Gaopeng; Wang, Feng; Zhang, Guangdong; Zhang, Zhe; Cao, Jianzhong; Lu, Rong] Chinese Acad Sci, Xian Inst Opt & Precis Mech, Xian 710119, Shaanxi, Peoples R China; [Zhang, Guangdong; Wang, Changqing] Northwestern Polytech Univ, Sch Automat, Xian 710072, Shaanxi, Peoples R China; [Du, Hubing; Zhao, Jingwei; Li, Yanjie] Xian Technol Univ, Sch Mechatron Engn, Xian 710021, Shaanxi, Peoples R China; [Zhao, Zixin] Xi An Jiao Tong Univ, State Key Lab Mfg Syst Engn, Xian 710049, Shaanxi, Peoples R ChinaAffiliations:Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Northwestern Polytechnical University; Xi'an Technological University; Xi'an Jiaotong UniversityPublication Year:2024Volume:177Article Number:108140DOI Link:http://dx.doi.org/10.1016/j.optlaseng.2024.108140數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001207533300001 -
Record 287 of
Title:Interaction semantic segmentation network via progressive supervised learning
Author Full Names:Zhao, Ruini; Xie, Meilin; Feng, Xubin; Guo, Min; Su, Xiuqin; Zhang, PingSource Title:MACHINE VISION AND APPLICATIONSLanguage:EnglishDocument Type:ArticleKeywords Plus:IMAGE SEGMENTATIONAbstract:Semantic segmentation requires both low-level details and high-level semantics, without losing too much detail and ensuring the speed of inference. Most existing segmentation approaches leverage low- and high-level features from pre-trained models. We propose an interaction semantic segmentation network via Progressive Supervised Learning (ISSNet). Unlike a simple fusion of two sets of features, we introduce an information interaction module to embed semantics into image details, they jointly guide the response of features in an interactive way. We develop a simple yet effective boundary refinement module to provide refined boundary features for matching corresponding semantic. We introduce a progressive supervised learning strategy throughout the training level to significantly promote network performance, not architecture level. Our proposed ISSNet shows optimal inference time. We perform extensive experiments on four datasets, including Cityscapes, HazeCityscapes, RainCityscapes and CamVid. In addition to performing better in fine weather, proposed ISSNet also performs well on rainy and foggy days. We also conduct ablation study to demonstrate the role of our proposed component. Code is available at: https://github.com/Ruini94/ISSNetAddresses:[Zhao, Ruini; Xie, Meilin; Feng, Xubin; Guo, Min; Su, Xiuqin] Chinese Acad Sci, Xian Inst Opt & Precis Mech, Xian 710119, Peoples R China; [Zhang, Ping] Changan Univ, Xian 710064, Peoples R ChinaAffiliations:Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Chang'an UniversityPublication Year:2024Volume:35Issue:2Article Number:26DOI Link:http://dx.doi.org/10.1007/s00138-023-01500-4數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001156034900001 -
Record 288 of
Title:A semi-supervised cross-modal memory bank for cross-modal retrieval
Author Full Names:Huang, Yingying; Hu, Bingliang; Zhang, Yipeng; Gao, Chi; Wang, QuanSource Title:NEUROCOMPUTINGLanguage:EnglishDocument Type:ArticleKeywords Plus:NETWORKAbstract:The core of semi -supervised cross -modal retrieval tasks lies in leveraging limited supervised information to measure the similarity between cross -modal data. Current approaches assume an association between unlabelled data and pre -defined k -nearest neighbour data, relying on classifier performance for this selection. With diminishing labelled data, classifier performance weakens, resulting in erroneous associations among unlabelled instances. Moreover, the lack of interpretability in class probabilities of unlabelled data hinders classifier learning. Thus, this paper focuses on learning pseudo -labels for unlabelled data, providing pseudosupervision to aid classifier learning. Specifically, a cross -modal memory bank is proposed, dynamically storing feature representations in a common space and class probability representations in a label space for each cross -modal data. Pseudo -labels are derived by computing feature representation similarity and adjusting class probabilities. During this process, imposing constraints on the classification loss between labelled data and contrastive losses between paired cross -modal data is a prerequisite for the successful learning of pseudolabels. This procedure significantly contributes to enhancing the credibility of these pseudo -labels. Empirical findings demonstrate that using only 10% labelled data, compared to prevailing semi -supervised techniques, this method achieves improvements of 2.6%, 1.8%, and 4.9% in MAP@50 on the Wikipedia, NUS -WIDE, and MS-COCO datasets, respectively.Addresses:[Huang, Yingying; Zhang, Yipeng; Gao, Chi; Wang, Quan] Chinese Acad Sci, Key Lab Spectral Imaging Technol, Xian Inst Opt & Precis Mech, Xian 710119, Shaanxi, Peoples R China; [Huang, Yingying; Zhang, Yipeng; Gao, Chi] Univ Chinese Acad Sci, Beijing 100049, Peoples R China; [Huang, Yingying; Hu, Bingliang; Zhang, Yipeng; Gao, Chi; Wang, Quan] Key Lab Biomed Spect, Xian 710119, Shaanxi, Peoples R ChinaAffiliations:Chinese Academy of Sciences; Xi'an Institute of Optics & Precision Mechanics, CAS; Chinese Academy of Sciences; University of Chinese Academy of Sciences, CASPublication Year:2024Volume:579Article Number:127430DOI Link:http://dx.doi.org/10.1016/j.neucom.2024.127430數(shù)據(jù)庫(kù)ID（收錄號(hào)）:WOS:001198409500001