Tian Lan

Tian Lan

Auto-generated profile, based on publicly available information

Tian Lan

Beijing Institute of Technology

Research Focus & Metrics

Academic Career Length

12Years (2015-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 October 2022 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
WGWeicheng Gao
Weicheng Gao
XYXiaopeng Yang
Xiaopeng Yang
XQXiaodong Qu
Xiaodong Qu
TLTian Lan
Tian Lan
To solve the problems of reduced accuracy and prolonging convergence time of through-the-wall radar (TWR) human motion due to wall attenuation, multipath effect, and system interference, we propose a multilink auto-encoding neural network (TWR-MCAE) data augmentation method. Specifically, the TWR-MCAE algorithm is jointly constructed by a singular value decomposition (SVD)-based data preprocessing module, an improved coordinate attention module, a compressed sensing learnable iterative shrinkage threshold reconstruction algorithm (LISTA) module, and an adaptive weight module. The data preprocessing module achieves wall clutter, human motion features, and noise subspaces separation. The improved coordinate attention module achieves clutter and noise suppression. The LISTA module achieves human motion feature enhancement. The adaptive weight module learns the weights and fuses the three subspaces. The TWR-MCAE can suppress the low-rank characteristics of wall clutter and enhance the sparsity characteristics in human motion at the same time. It can be linked before the classification step to improve the feature extraction capability without adding other prior knowledge or recollecting more data. Experiments show that the proposed algorithm gets a better peak signal-to-noise ratio (PSNR), which increases the recognition accuracy and speeds up the training process of the back-end classifiers.

9 October 2018 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
TLTian Lan
Tian Lan
NLNa Liu
Na Liu
FHFeng Han
Feng Han
QLQing Huo Liu
Qing Huo Liu
A joint petrophysical and structural inversion method for electromagnetic (EM) and seismic data based on the volume integral equation (VIE) is proposed in this paper. In the forward EM problem, only the contrast of conductivity is solved by the electric field integral equation method. However, in the forward seismic problem, both the contrasts of velocity and mass density are solved by the combined field VIE method. Both forward solvers are accelerated by the fast Fourier transform. In the inversion problem, by using the petrophysical equations about the porosity and saturation and applying the chain rule, we fuse the EM and seismic data and construct the joint petrophysical inversion equations, which can be solved by the variational Born iteration method. Then, in order to further enhance the reconstructed results of the joint petrophysical inversion, we enforce the structural similarity constraint between porosity and water saturation and add the cross-gradient function to the joint petrophysical inversion cost function. Two typical geophysical models based on the remote sensing measurement are used to validate the proposed methods. One is the cross-well model, and the other is the marine surface exploration model. The advantage of the joint inversion compared with the separate inversion is evaluated based on the resolution and the data misfits of the reconstructed profiles as well as the antinoise ability.

4 August 2023
Verified
Conference Paper
Open access
YWYangyang Wang
Yangyang Wang
XZXiao Zhang
Xiao Zhang
MLMingyi Li
Mingyi Li
TLTian Lan
Tian Lan
HCHuashan Chen
Huashan Chen
Dongxiao YuDongxiao Yu
Dongxiao Yu
In this paper, we propose an adaptive learning paradigm for resource-constrained cross-device federated learning, in which heterogeneous local submodels with varying resources can be jointly trained to produce a global model. Different from existing studies, the submodel structures of different clients are formed by arbitrarily assigned neurons according to their local resources. Along this line, we first design a general resource-adaptive federated learning algorithm, namely RA-Fed, and rigorously prove its convergence with asymptotically optimal rate O(1/√Γ*TQ) under loose assumptions. Furthermore, to address both submodels heterogeneity and data heterogeneity challenges under non-uniform training, we come up with a new server aggregation mechanism RAM-Fed with the same theoretically proved convergence rate. Moreover, we shed light on several key factors impacting convergence, such as minimum coverage rate, data heterogeneity level, submodel induced noises. Finally, we conduct extensive experiments on two types of tasks with three widely used datasets under different experimental settings. Compared with the state-of-the-arts, our methods improve the accuracy up to 10% on average. Particularly, when submodels jointly train with 50% parameters, RAM-Fed achieves comparable accuracy to FedAvg trained with the full model.

3 November 2022 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
HSHaoran Sun
Haoran Sun
XYXiaopeng Yang
Xiaopeng Yang
JGJunbo Gong
Junbo Gong
XQXiaodong Qu
Xiaodong Qu
TLTian Lan
Tian Lan
To better reconstruct underground targets based on ground-penetrating radar (GPR) data, this paper proposes a joint physics and data driven full-waveform inversion (PDD-FWI) scheme. This scheme combines a physics-based non-iterative approach and a data-driven deep neural network (DNN) to reconstruct target location, shape and permittivity accurately. Firstly, the normalized range migration algorithm (RMA) is introduced to extract the target contour and location information, which not only improves the robustness of the proposed scheme, but also ensures adaptability to different GPR equipment. Then, the GPR dielectric target reconstruction network (GPRDtrNet) is developed based on the improved U-net structure, including reducing network layers and adding multi-scale additive spatial attention gates and skip-connection structures. Compared with previous DNN-based reconstruction methods, GPRDtrNet has the advantages of small data requirement, high accuracy, strong generalization and noise tolerance. Finally, the simulated and real dataset containing kinds of targets is constructed to train and test GPRDtrNet. The results show that the proposed method can reconstruct underground dielectric targets accurately with high robustness and noise tolerance.

29 November 2021 in IET Radar, Sonar & Navigation
Verified
Journal Article
Open access
PDF file
SGSheng Gao
Sheng Gao
XYXiaopeng Yang
Xiaopeng Yang
TLTian Lan
Tian Lan
BHBowen Han
Bowen Han
HSHaoran Sun
Haoran Sun
ZYZhichao Yu
Zhichao Yu
The main‐lobe jamming causes serious degradation of radar system performance in practical applications. In recent years, the Blind Source Separation (BSS) method has been applied to suppress the main‐lobe jamming. However, the received echoes of target and jamming cannot be effectively identified while the robustness of the conventional BSS method is poor due to the influence of system noise. To tackle these problems, a radar main‐lobe jamming suppression and identification method is proposed based on robust whitening BSS and Convolutional Neural Networks methods. First, the target echoes and jamming are separated by the proposed robust whitening BSS method. An iterative optimisation method is proposed to obtain the robust whitening matrix, and sufficient conditions are provided for solvability. Then, a Target echo and Jamming Identification Network (TJINet) is proposed to identify the signals after the robust whitening BSS. The proposed TJINet method uses residual blocks and branch structures to improve classification performance and utilises asymmetric convolution blocks to extract subtle features between the target echo and deceptive jamming. Simulation and experimental results show that the proposed method has better jamming suppression performance than the traditional methods under low signal‐to‐noise ratio (SNR) conditions, the algorithm converges faster, and also can stably identify the target echo and jamming signals.

22 February 2022 in IEEE Antennas and Wireless Propagation Letters
Verified
Journal Article
Closed access
YLYixuan Li
Yixuan Li
XYXiaopeng Yang
Xiaopeng Yang
TLTian Lan
Tian Lan
RLRenjie Liu
Renjie Liu
XQXiaodong Qu
Xiaodong Qu
Parameter estimation of layered media is an important application of ground penetrating radar (GPR). However, in the case of thin-layer detection, the limitation of GPR vertical resolution can lead to overlapping reflected signals, which seriously deteriorates the detection results. To solve this problem, a modified reflected signal reconstruction method based on reflected signal reconstruction is proposed in this letter. First, a forward model of the GPR reflected signal is constructed using the generalized reflection coefficient definition. Then, the parameter inversion is transformed into an optimization problem. The residue cost function is defined as the error square between the model and the actual reflected signal, which is minimized by using genetic algorithm. Furthermore, the generalized reflection coefficient spectrum is feasible to estimate the parameters of each layer as the initial values of the parameters required by the algorithm, thus improving the accuracy and convergence speed of the algorithm. Numerical, experimental, and field tests demonstrate that the method has a high time resolution and antinoise capability in the inversion of layered media parameters.