Lightweight Cross-Modal Multispectral Pedestrian Detection Based on Spatial Reweighted Attention Mechanism

Multispectral pedestrian detection technology leverages infrared images to provide reliable information for visible light images, demonstrating significant advantages in low-light conditions and background occlusion scenarios. However, while continuously improving cross-modal feature extraction and fusion, ensuring the model’s detection speed is also a challenging issue. We have devised a deep learning network model for cross-modal pedestrian detection based on Resnet50, aiming to focus on more reliable features and enhance the model’s detection efficiency. This model employs a spatial attention mechanism to reweight the input visible light and infrared image data, enhancing the model’s focus on different spatial positions and sharing the weighted feature data across different modalities, thereby reducing the interference of multi-modal features. Subsequently, lightweight modules with depthwise separable convolution are incorporated to reduce the model’s parameter count and computational load through channel-wise and point-wise convolutions. The network model algorithm proposed in this paper was experimentally validated on the publicly available KAIST dataset and compared with other existing methods. The experimental results demonstrate that our approach achieves favorable performance in various complex environments, affirming the effectiveness of the multispectral pedestrian detection technology proposed in this paper.

New regularization method and iteratively reweighted algorithm for sparse vector recovery

Motivated by the study of regularization for sparse problems,we propose a new regularization method for sparse vector recovery.We derive sufficient conditions on the well-posedness of the new regularization,and design an iterative algorithm,namely the iteratively reweighted algorithm(IR-algorithm),for efficiently computing the sparse solutions to the proposed regularization model.The convergence of the IR-algorithm and the setting of the regularization parameters are analyzed at length.Finally,we present numerical examples to illustrate the features of the new regularization and algorithm.

APPLICATION OF LEAST MEDIAN OF SQUARED ORTHOGONAL DISTANCE (LMD) AND LMD BASED REWEIGHTED LEAST SQUARES (RLS) METHODS ON THE STOCK RECRUITMENT RELATIONSHIP

Analysis of stock recruitment (SR) data is most often done by fitting various SR relationship curves to the data. Fish population dynamics data often have stochastic variations and measurement errors, which usually result in a biased regression analysis. This paper presents a robust regression method, least median of squared orthogonal distance (LMD), which is insensitive to abnormal values in the dependent and independent variables in a regression analysis. Outliers that have significantly different variance from the rest of the data can be identified in a residual analysis. Then, the least squares (LS) method is applied to the SR data with defined outliers being down weighted. The application of LMD and LMD based Reweighted Least Squares (RLS) method to simulated and real fisheries SR data is explored.

The construction of general basis functions in reweighting ensemble dynamics simulations: Reproduce equilibrium distribution in complex systems from multiple short simulation trajectories

Ensemble simulations, which use multiple short independent trajectories from dispersive initial conformations, rather than a single long trajectory as used in traditional simulations, are expected to sample complex systems such as biomolecules much more efficiently. The re-weighted ensemble dynamics（RED） is designed to combine these short trajectories to reconstruct the global equilibrium distribution. In the RED, a number of conformational functions, named as basis functions,are applied to relate these trajectories to each other, then a detailed-balance-based linear equation is built, whose solution provides the weights of these trajectories in equilibrium distribution. Thus, the sufficient and efficient selection of basis functions is critical to the practical application of RED. Here, we review and present a few possible ways to generally construct basis functions for applying the RED in complex molecular systems. Especially, for systems with less priori knowledge, we could generally use the root mean squared deviation（RMSD） among conformations to split the whole conformational space into a set of cells, then use the RMSD-based-cell functions as basis functions. We demonstrate the application of the RED in typical systems, including a two-dimensional toy model, the lattice Potts model, and a short peptide system. The results indicate that the RED with the constructions of basis functions not only more efficiently sample the complex systems, but also provide a general way to understand the metastable structure of conformational space.

Evaluating gravity gradient components based on a reweighted inversion method

In gravity gradient inversion,to choose an appropriate component combination is very important,that needs to understand the function of each component of gravity gradient in the inversion.In this paper,based on the previous research on the characteristics of gravity gradient components,we propose a reweighted inversion method to evaluate the influence of single gravity gradient component on the inversion resolution The proposed method only adopts the misfit function of the regularized equation and introduce a depth weighting function to overcome skin effect produced in gravity gradient inversion.A comparison between different inversion results was undertaken to verify the influence of the depth weighting function on the inversion result resolution.To avoid the premise of introducing prior information,we select the depth weighting function based on the sensitivity matrix.The inversion results using the single-prism model and the complex model show that the influence of different components on the resolution of inversion results is different in different directions,however,the inversion results based on two kind of models with adding different levels of random noise are basically consistent with the results of inversion without noises.Finally,the method was applied to real data from the Vinton salt dome,Louisiana,USA.

Iterative Reweighted <i>l</i>₁Penalty Regression Approach for Line Spectral Estimation

In this paper, we proposed an iterative reweighted l1?penalty regression approach to solve the line spectral estimation problem. In each iteration process, we first use the ideal of Bayesian lasso to update the sparse vectors;the derivative of the penalty function forms the regularization parameter. We choose the anti-trigonometric function as a penalty function to approximate the?l0? norm. Then we use the gradient descent method to update the dictionary parameters. The theoretical analysis and simulation results demonstrate the effectiveness of the method and show that the proposed algorithm outperforms other state-of-the-art methods for many practical cases.

Continuous Iteratively Reweighted Least Squares Algorithm for Solving Linear Models by Convex Relaxation

In this paper, we present continuous iteratively reweighted least squares algorithm (CIRLS) for solving the linear models problem by convex relaxation, and prove the convergence of this algorithm. Under some conditions, we give an error bound for the algorithm. In addition, the numerical result shows the efficiency of the algorithm.

Instance Reweighting Adversarial Training Based on Confused Label

Reweighting adversarial examples during training plays an essential role in improving the robustness of neural networks,which lies in the fact that examples closer to the decision boundaries are much more vulnerable to being attacked and should be given larger weights.The probability margin(PM)method is a promising approach to continuously and path-independently mea-suring such closeness between the example and decision boundary.However,the performance of PM is limited due to the fact that PM fails to effectively distinguish the examples having only one misclassified category and the ones with multiple misclassified categories,where the latter is closer to multi-classification decision boundaries and is supported to be more critical in our observation.To tackle this problem,this paper proposed an improved PM criterion,called confused-label-based PM(CL-PM),to measure the closeness mentioned above and reweight adversarial examples during training.Specifi-cally,a confused label(CL)is defined as the label whose prediction probability is greater than that of the ground truth label given a specific adversarial example.Instead of considering the discrepancy between the probability of the true label and the probability of the most misclassified label as the PM method does,we evaluate the closeness by accumulating the probability differences of all the CLs and ground truth label.CL-PM shares a negative correlation with data vulnerability:data with larger/smaller CL-PM is safer/riskier and should have a smaller/larger weight.Experiments demonstrated that CL-PM is more reliable in indicating the closeness regarding multiple misclassified categories,and reweighting adversarial training based on CL-PM outperformed state-of-the-art counterparts.

A Reweighted Total Variation Algorithm with the Alternating Direction Method for Computed Tomography

A variety of alternating direction methods have been proposed for solving a class of optimization problems. The applications in computed tomography (CT) perform well in image reconstruction. The reweighted schemes were applied in l1-norm and total variation minimization for signal and image recovery to improve the convergence of algorithms. In this paper, we present a reweighted total variation algorithm using the alternating direction method (ADM) for image reconstruction in CT. The numerical experiments for ADM demonstrate that adding reweighted strategy reduces the computation time effectively and improves the quality of reconstructed images as well.

Continuous Mixed p-norm Control Scheme with Reweighted L_(0) norm Variable Step Size for Mitigating Power Quality Problems of Grid Coupled Solar PV Systems

In this paper,the performance of a two-stage three-phase grid coupled solar photovoltaic generating system(SPVGS)is analyzed by using a novel reweighted Lo norm variable step size continuous mixed p-norm(RLo-VSSCMPN)of a voltage source inverter(VSI)control scheme.The efficacy of the system is determined by considering unbalanced grid voltage,DC offset,voltage sag and swell,unbalanced load and variations in solar insolation.RLo-VSSCMPN is used for inverter control and it ex-tracts fundamental components of load current for generating the reference grid current with a faster convergence rate and lesser steady state oscillations.With the proposed control,harmonics in the grid current follows the IEEE-519 norm and the performance is also satisfactory under varying environmental/load conditions.The power generated from SPvGS is transferred optimally using a DC-DC boost converter utilizing the incremental conductance(INC)maximum power point technique.The proposed system is simulated using MATLAB/Simulink 2018a and test results are verified experimentally using dSPACE1202 in the laboratory to ensure the validity of a novel proposed robust RLo-VSSCMPN.Index Terms-INC maximum power point tracker,power quality,reweighted LoVSSCMPN algorithm,solar PV generating system,total harmonic distortion,voltage source inverter.

Iterative-Reweighting-Based Robust Iterative-Closest-Point Method

In point cloud registration applications,noise and poor initial conditions lead to many false matches.False matches significantly degrade registration accuracy and speed.A penalty function is adopted in many robust point-to-point registration methods to suppress the influence of false matches.However,after applying a penalty function,problems cannot be solved in their analytical forms based on the introduction of nonlinearity.Therefore,most existing methods adopt the descending method.In this paper,a novel iterative-reweighting-based method is proposed to overcome the limitations of existing methods.The proposed method iteratively solves the eigenvectors of a four-dimensional matrix,whereas the calculation of the descending method relies on solving an eight-dimensional matrix.Therefore,the proposed method can achieve increased computational efficiency.The proposed method was validated on simulated noise corruption data,and the results reveal that it obtains higher efficiency and precision than existing methods,particularly under very noisy conditions.Experimental results for the KITTI dataset demonstrate that the proposed method can be used in real-time localization processes with high accuracy and good efficiency.

Rapid fatty acids detection of vegetable oils by Raman spectroscopy based on competitive adaptive reweighted sampling coupled with support vector regression

Objectives:The composition and content of fatty acids are critical indicators of vegetable oil quality.To overcome the drawbacks of traditional detection methods,Raman spectroscopy was investigated for the fast determination of the fatty acids composition of oil.Materials and Methods:Rapeseed and soybean oil at different depths of the oil tank at different storage times were collected and an eighth-degree polynomial function was used to fit the Raman spectrum.Then,the multivariate scattering correction,standard normal variable transformation(SNV),and Savitzky–Golay convolution smoothing methods were compared.Results:Polynomial fitting combined with SNV was found to be the optimal pretreatment method.Characteristic wavelengths were selected by competitive adaptive reweighted sampling.For monounsaturated fatty acids(MUFAs),polyunsaturated fatty acids(PUFAs),and saturated fatty acids(SFAs),44,75,and 92 characteristic wavelengths of rapeseed oil,and 60,114,and 60 characteristic wavelengths of soybean oil were extracted.Support vector regression was used to establish the prediction model.The R^(2)values of the prediction results of MUFAs,PUFAs,and SFAs for rapeseed oil were 0.9670,0.9568,and 0.9553,and the root mean square error(RMSE)values were 0.0273,0.0326,and 0.0340,respectively.The R^(2)values of the prediction results of fatty acids for soybean oil were respectively 0.9414,0.9562,and 0.9422,and RMSE values were 0.0460,0.0378,and 0.0548,respectively.A good correlation coefficient and small RMSE value were obtained,indicating the results to be highly accurate and reliable.Conclusions:Raman spectroscopy,based on competitive adaptive reweighted sampling coupled with support vector regression,can rapidly and accurately analyze the fatty acid composition of vegetable oil.

Threshold reweighted Nadaraya-Watson estimation of jump-diffusion models

In this paper,we propose a new method to estimate the diffusion function in the jump-diffusion model.First,a threshold reweighted Nadaraya-Watson-type estimator is introduced.Then,we establish asymptotic normality for the estimator and conduct Monte Carlo simulations through two examples to verify the better finite-sampling properties.Finally,our estimator is demonstrated through the actual data of the Shanghai Interbank Offered Rate in China.

Software Defect Prediction Method Based on Stable Learning

The purpose of software defect prediction is to identify defect-prone code modules to assist software quality assurance teams with the appropriate allocation of resources and labor.In previous software defect prediction studies,transfer learning was effective in solving the problem of inconsistent project data distribution.However,target projects often lack sufficient data,which affects the performance of the transfer learning model.In addition,the presence of uncorrelated features between projects can decrease the prediction accuracy of the transfer learning model.To address these problems,this article propose a software defect prediction method based on stable learning(SDP-SL)that combines code visualization techniques and residual networks.This method first transforms code files into code images using code visualization techniques and then constructs a defect prediction model based on these code images.During the model training process,target project data are not required as prior knowledge.Following the principles of stable learning,this paper dynamically adjusted the weights of source project samples to eliminate dependencies between features,thereby capturing the“invariance mechanism”within the data.This approach explores the genuine relationship between code defect features and labels,thereby enhancing defect prediction performance.To evaluate the performance of SDP-SL,this article conducted comparative experiments on 10 open-source projects in the PROMISE dataset.The experimental results demonstrated that in terms of the F-measure,the proposed SDP-SL method outperformed other within-project defect prediction methods by 2.11%-44.03%.In cross-project defect prediction,the SDP-SL method provided an improvement of 5.89%-25.46% in prediction performance compared to other cross-project defect prediction methods.Therefore,SDP-SL can effectively enhance within-and cross-project defect predictions.

基于重加权L1的ATpV正则化叠前反演方法

地震叠前反演能够准确获取地下储层介质的各类参数,是油气的勘探与开发中重要技术之一。然而,地震反演是典型的病态问题,为了克服此问题,通常使用正则化约束目标函数,来减轻反演问题的病态性。但是正则化约束忽略了地层边界的振幅信息,使用重加权方法可以很好地克服这一问题,更好地恢复稀疏性。提出了一种基于重加权L1的ATpV正则化叠前三参数反演方法(ATpV-L1方法),首次将重加权L1方法与ATpV方法结合,并引入到叠前反演中。采用交替方向乘子算法(ADMM)建立反演框架,对目标函数进行分块优化,有效提高了收敛速度。首先,介绍ATpV-L1方法,建立了基于ATpV-L1的叠前反演目标函数;然后,应用理论模拟数据对比新方法和ATpV方法反演结果,验证了方法的效果;最后,使用实际数据进行实验分析,进一步验证了ATpV-L1方法的反演精度及可行性。实验结果表明,提出的ATpV-L1方法可以有效恢复反演结果的稀疏性,提高反演精度。

ZY1-02DAHSI影像归一化阴影植被指数NSVI的波段选择及其构建

高光谱影像具有连续的地物光谱信息,在阴影检测方面具有巨大的潜力,而波段冗余度高需进行波段优选。归一化阴影植被指数(NSVI)能够扩大光谱差异,在高光谱影像中应用NSVI将更有效地识别阴影。资源一号02D卫星是我国首颗自主研发并成功运行的高光谱业务卫星,数据信噪比大、覆盖能力强,对该高光谱影像进行准确的阴影检测具有重要意义。以ZY1-02DAHSI影像为试验数据,提取并分析明亮区植被、阴影区植被及水体的光谱反射率;结合竞争自适应重加权采样(CARS)和连续投影算法(SPA)筛选能够有效区分典型地物的主要波段,综合考虑算法的特性进一步选出特征波段构建NSVI;通过步长法确定最佳阈值对影像进行分类,从像元值分布情况、分类精度和光谱增强效果等对比出构建NSVI的最佳波段,并结合不同的阴影指数、波段和影像进行综合评价,验证该方法的意义及普适性。结果表明:波段32和波段73是构建NSVI的最佳波段,分别对应红光波段和近红外波段;不同波段构建的NSVI分类精度均高于90%,由最佳波段构建的NSVI分类精度为94.33%,Kappa系数为0.8328,分类效果最优;NSVI能够增强典型地物间的光谱差异并缓解归一化植被指数的“易饱和”现象,在该影像中因水体累积产生的小波峰有助于提取水体;在ZY1-02DAHSI影像中NSVI的分类效果优于归一化阴影指数和阴影指数,于另一景影像的分类精度也达到93.55%,Kappa系数为0.8167。由算法筛选出的波段具有一定的代表性,最佳波段构建的NSVI在ZY1-02DAHSI影像中具有较好的阴影检测能力,对高光谱影像阴影检测及构建植被指数具有一定的借鉴和参考意义。

基于CWT-sCARS的土壤铜含量高光谱反演

光谱变量的有效程度与土壤铜含量的反演精度密切相关。基于原始反射率以及不同分解尺度下的小波系数,本研究采用连续小波变换(CWT)算法、稳定性竞争自适应重加权采样(sCARS)算法和随机森林(RF)算法对土壤铜含量进行了反演与验证。研究结果表明:连续小波变换可以有效提高光谱特征与土壤铜含量之间的相关性,不同分解尺度对应的最大相关系数中,最大值位于Scale 8分解尺度下1343 nm处,相关系数为0.60;使用sCARS算法可以显著减少特征变量的数量,结合CWT变换和sCARS算法可以显著减轻数据冗余,提高土壤Cu含量的反演精度。该研究可为利用高光谱遥感技术,快速、高精度反演土壤Cu含量提供重要参考。

基于特征重加权的小样本遥感图像目标检测算法

针对遥感图像具有目标尺度多变、目标模糊、背景复杂的特点,提出了一种基于特征重加权的遥感小样本目标检测算法RE-FSOD;该模型包括3部分:元特征提取器、特征重加权提取器、预测模块,其中元特征提取器由CSPDarknet-53、FPN以及PAN构成,负责提取数据的元特征;特征重加权提取器用于生成特征重加权向量,用于调整元特征来强化对于检测新类有帮助的特征;预测模块由YOLOv3的预测模块构成,在此基础上将定位损失函数替换为CIOU损失函数,提升模型的定位精度;最后在NWPU VHR-10遥感数据集上进行了训练和测试,实验结果表明,该方法相较于基线方法FSODM的在3-shot、5-shot、10-shot情况下分别提升了约19%、11%、8%。

面向激光测振应用的改进相位生成载波解调算法

为了解决相位生成载波解调方案应用于激光测振技术时,系统中存在的非线性失真问题,提出一种改进相位生成载波解调算法。该算法采用低频、大幅值的相位调制和迭代重加权椭圆特殊拟合对两路含有非线性误差的正交信号进行校正以抑制非线性失真。其中,低频相位调制由三角波信号驱动压电换能器生成,确保小信号情况下椭圆拟合结果的准确性。迭代重加权椭圆特殊拟合可以减小离群数据的影响并避免拟合结果退化为双曲线,具有精度高、鲁棒性好和计算效率高的优点。实验结果表明改进相位生成载波解调算法可以有效抑制激光测振实验系统的非线性失真,在不同相位调制深度(0.8~3.4 rad)下解调信号的信纳比和总谐波失真稳定,对应的标准差分别为0.55 dB和0.03%。系统的响应线性度优于99.99%,动态范围可达103.9 dB@500 Hz,总谐波失真为1%且工作频宽为20~8000 Hz。与传统相位生成载波解调方案相比,该算法不仅显著抑制了非线性失真,还克服了椭圆拟合算法在小信号下无法工作的缺点。

基于稀疏增强重加权与掩码块张量的红外弱小目标检测

高度异构的复杂背景破坏了场景的低秩性,现有算法难以利用低秩稀疏恢复方法从背景中分离出小目标。为了解决上述问题,本文将小目标检测问题转化为张量模型的凸优化函数求解问题,提出基于稀疏增强重加权与掩码块张量的检测模型。首先,将掩码块图像以堆叠方式扩展至张量空间,并构建掩码块张量模型以筛选候选目标。在此基础上,利用结构张量构建稀疏增强重加权模型以抑制背景杂波,克服凸优化函数求解过程中设定加权参数的缺陷。实验表明本文检测算法在背景抑制因子及信杂比增益两方面都优于新近代表性算法,证明该算法的有效性。