公开(公告)号：US20250157045A1

公开(公告)日：2025-05-15

申请号：US18518513

申请日：2023-11-23

Applicant: University Of Electronic Science And Technology Of China ,  Yangtze Delta Region Institute (Huzhou), University Of Electronic Science And Technology Of China

Inventor： Shihua Li ,  Shunda Zhao ,  Yuyang Guo ,  Fugui Luo ,  Minfeng Xing

IPC: G06T7/187 ,  G06T7/11 ,  G06T7/155 ,  G06T7/521

Abstract: Provided is an individual-tree segmentation method of UAV LiDAR point cloud based on canopy morphology. The method uses woodland data obtained by a UAV LiDAR to: extract initial canopies from a CHM based on a region growing algorithm, determine whether each initial canopy is a correct segmentation canopy according to the number of local density maximum points in each initial canopy, finely segment each wrong segmentation canopy according to canopy morphology to obtain an updated set of tree tops, and finally use each of the updated set of tree tops as a seed point for performing the region growing algorithm to thereby obtain a final individual-tree segmentation result. The method make full use of height information and density information contained in a tree point cloud; and under the guidance of the density information, a wrong segmentation tree can be more accurately identified.

公开(公告)号：US12051211B2

公开(公告)日：2024-07-30

申请号：US17989453

申请日：2022-11-17

Applicant: University of Electronic Science and Technology of China ,  Yangtze Delta Region Institute (Huzhou), University of Electronic Science and Technology of China

Inventor： Jiang Qian ,  Haitao Lyu ,  Junzheng Jiang ,  Minfeng Xing

IPC: G06K9/00 ,  G06T7/20

CPC classification number: G06T7/20 ,  G06T2207/20081

Abstract: A moving target focusing method and system based on a generative adversarial network are provided. The method includes: generating, using a Range Doppler algorithm, a two-dimensional image including at least one defocused moving target, as a training sample; generating at least one ideal Gaussian point in a position of at least one center of the at least one defocused moving target in the two-dimensional image, to generate a training label; constructing the generative adversarial network, the generative adversarial network includes a generative network and a discrimination network; inputting the training sample and the training label into the generative adversarial network to perform repeated training until an output of the generative network reaches a preset condition, to thereby obtain a trained network model; and inputting a testing sample into the trained network model, to output a moving target focused image.

公开(公告)号：US20220294605A1

公开(公告)日：2022-09-15

申请号：US17517661

申请日：2021-11-02

Applicant: University of Electronic Science and Technology of China ,  Yangtze Delta Region Institute (Huzhou), University of Electronic Science and Technology of China

Inventor： Chunxiang XU ,  Changsong JIANG ,  Yicong DU

IPC: H04L9/00 ,  H04L9/06

Abstract: A blockchain-based public parameter generation method against backdoor attacks, includes: acquiring the hash values of L latest confirmed blocks on a blockchain, and the hash values of the L blocks and a count variable for generation are mapped to an element in a set G via a specified mapping to obtain the generated public parameter; L≥φ, φ is the minimum number to guarantee blockchains' chain quality property; checking whether the generated parameter meets the condition, if not, discarding the parameter and updating the generated public parameter; if the condition is met, outputting the public parameter to the device that uses the public parameter. In this disclosure, the public parameters are random, since they are based on the latest confirmed blocks on the blockchain and are guaranteed by the computational power of the blockchain; the generation of public parameters is publicly verifiable and random.

公开(公告)号：US20250020772A1

公开(公告)日：2025-01-16

申请号：US18517235

申请日：2023-11-22

Applicant: UNIVERSITY OF ELECTRONIC SCIENCE AND TECHNOLOGY OF CHINA ,  Yangtze Delta Region Institute (Huzhou), University of Electronic Science and Technology of China

Inventor： Shihua Li ,  Ze He ,  Fugui Luo ,  Minfeng Xing

IPC: G01S7/41 ,  G01S13/88

Abstract: A rice-crop intensity identification method based on radar time series observation and temperature analysis is provided. Capturing of diversified periodic characteristics of time series backscatter and detection of backscatter troughs are achieved through time series reconstruction and trough identification; potential phenological phases corresponding to the backscatter troughs are determined through potential rice phenological phase estimation; and through temperature limitation of rice phenological phase, temperature suitability of potential rice phenological phases is evaluated, a backscatter trough that does not satisfy a temperature condition is removed by combining a rice growth mechanism and a regulation of rice-crop intensity, thereby realizing the identification of the rice troughs and the correction of the overestimation of the rice-crop intensity, and finally realizing the identification of rice-crop intensity.

公开(公告)号：US20230162373A1

公开(公告)日：2023-05-25

申请号：US17989453

申请日：2022-11-17

Applicant: University of Electronic Science and Technology of China ,  Yangtze Delta Region Institute (Huzhou), University of Electronic Science and Technology of China

Inventor： Jiang Qian ,  Haitao Lyu ,  Junzheng Jiang ,  Minfeng Xing

IPC: G06T7/20

CPC classification number: G06T7/20 ,  G06T2207/20081

Abstract: A moving target focusing method and system based on a generative adversarial network are provided. The method includes: generating, using a Range Doppler algorithm, a two-dimensional image including at least one defocused moving target, as a training sample; generating at least one ideal Gaussian point in a position of at least one center of the at least one defocused moving target in the two-dimensional image, to generate a training label; constructing the generative adversarial network, the generative adversarial network includes a generative network and a discrimination network; inputting the training sample and the training label into the generative adversarial network to perform repeated training until an output of the generative network reaches a preset condition, to thereby obtain a trained network model; and inputting a testing sample into the trained network model, to output a moving target focused image.

公开(公告)号：US20250139491A1

公开(公告)日：2025-05-01

申请号：US18383908

申请日：2023-10-26

Applicant: YANGTZE DELTA REGION INSTITUTE (HUZHOU), UNIVERSITY OF ELECTRONIC SCIENCE AND TECHNOLOGY OF CHINA

Inventor： Huanlong Zhang ,  Yanchun Zhao ,  Panyun Wang ,  Fengli Shen ,  Fusheng Li ,  Pengpeng Zhi ,  Weiqiang Fu ,  Xiangbo Yang

IPC: G06N20/00

Abstract: The present disclosure discloses a single-target tracking method based on a credit allocation network. The credit allocation network is at first provided, which can be updated online using a guiding focus loss function. The credit allocation network generates credit scores for the tracking results by learning features of the target object, which ensures that the reliable samples are updated for storage in the memory pool. In order to better adapt to the changes in the appearance of the target during the tracking process, a new memory selection strategy is provided to collect high-quality tracking results as the memory samples during the tracking process, which further improves the reliability and adaptability of the memory pool. The present disclosure uses the guiding focus loss function to update the credit allocation network online, which can select more reliable memory samples for the memory pool and improve the robustness of the tracking results.

公开(公告)号：US12267100B2

公开(公告)日：2025-04-01

申请号：US17877896

申请日：2022-07-30

Applicant: University of Electronic Science and Technology of China

Inventor： Bo Zhang ,  Zhongqian Niu ,  Xiaobo Yang ,  Bingli Dai ,  Yi Hu ,  Jicong Zhang ,  Yong Fan ,  Ke Liu ,  Zhi Chen

IPC: H04B1/40 ,  H03D7/16

Abstract: The present disclosure provides a hetero-integrated terahertz low-noise miniaturized image frequency rejection transceiver front-end, including an intermediate frequency circuit and a terahertz circuit arranged up and down, where the terahertz circuit includes a local oscillator frequency tripler, a 135° 3 dB filter coupler, a radio frequency waveguide power divider, and two quartz hetero-integrated subharmonic mixers; resonant cavities of an input unit, a first output unit, an isolation unit, and a second output unit of the 135° 3 dB filter coupler are sequentially coupled through resonant grooves to form a ring structure, a cavity length of the resonant cavity of the input unit is twice that of the resonant cavities of the other three units, and an electrical length of a waveguide of the first output unit is 45° different from that of a waveguide of the second output unit.

公开(公告)号：US20250103952A1

公开(公告)日：2025-03-27

申请号：US18521883

申请日：2023-11-28

Applicant: Hebei University of Technology ,  Zhejiang Lab ,  Shenzhen Institute for Advanced Study, University of Electronic Science and Technology of China ,  DALIAN UNIVERSITY OF TECHNOLOGY

Inventor： Bin CAO ,  Zeyu JIANG ,  Xin LIU ,  Wen CHENG ,  Yun LI ,  Rensheng SHEN ,  Yuchun CHANG

IPC: G06N20/00

Abstract: The present invention provides a federated large model adaptive learning system. Based on the combination of multiobjective optimization and incremental learning, multiple optimization indexes are constructed, and adaptive mini model incremental learning is designed. A gradient scaling method of mini models is proposed for data privacy protection under federated learning, to make full use of gradient information. A correlation between the generalization ability and sampling data is revealed to propose a generalization ability evaluation function. With respect to the real problems of performance degradation and fault faced by industrial equipment during operation, multiple optimization objectives are designed in combination with the generalization ability evaluation function, and the models are updated and repaired adaptively through multiobjective evolutionary learning, to improve the usability of large models in real industrial scenarios. Finally, the adaptive accurate update of the large models and mini models is realized to improve the generalization ability of the models.

公开(公告)号：US20250081553A1

公开(公告)日：2025-03-06

申请号：US18388870

申请日：2023-11-13

Applicant: University of Electronic Science and Technology of China

Inventor： Ming QIAO ,  Jue LI ,  Zesheng SHI ,  Daoming SHEN ,  Bo ZHANG

IPC: H01L29/06

Abstract: A power semiconductor device, including a cell region, a transition region, and a terminal region. The transition region is located between the cell region and the terminal region of the device. A first conduction type substrate, a first conduction type epitaxial layer located above the first conduction type substrate, and a first conduction type buffer layer located in the first conduction type epitaxial layer are jointly arranged at the bottoms of the cell region, the transition region, and the terminal region of the device. In a high-current application, since the cell region occupies the largest area of a chip, in a case that breakdown can occur in the cell region and the current can be discharged through the cell region. On the basis of ensuring the BV of the terminal region, a silicon layer step is formed by elevating the position of a top structure of the terminal region.

公开(公告)号：US20250068798A1

公开(公告)日：2025-02-27

申请号：US18383419

申请日：2023-10-24

Applicant: University of Electronic Science and Technology of China, Yangtze River Delta Research Institute

Inventor： Rongjiang Tang ,  Lu Gan ,  Fusheng Li ,  Fengli Shen

IPC: G06F30/27

Abstract: The present disclosure belongs to the field of geophysical exploration, and discloses a method, a system, a medium, a device and a terminal for predicting the probing depth of transient electromagnetic. Based on the existing published resistivity model database, the transient electromagnetic field in layered medium is calculated, and the probing depth is calculated based on Jacobian matrix to establish a training data set. The simulated induced electromotive force is used as the input of the neural network, and the calculated probing depth is used as the output of the network. A rapid mapping between observation data and probing depth is established by using residual neural network.