Robot Vision over CosGANs to Enhance Performance with Source-Free Domain Adaptation Using Advanced Loss Function

Domain shift is when the data used in training does not match the ones it will be applied to later on under similar conditions.Domain shift will reduce accuracy in results.To prevent this,domain adaptation is done,which adapts the pre-trained model to the target domain.In real scenarios,the availability of labels for target data is rare thus resulting in unsupervised domain adaptation.Herein,we propose an innovative approach where source-free domain adaptation models and Generative Adversarial Networks(GANs)are integrated to improve the performance of computer vision or robotic vision-based systems in our study.Cosine Generative Adversarial Network(CosGAN)is developed as a GAN that uses cosine embedding loss to handle issues associated with unsupervised source-relax domain adaptations.For less complex architecture,the CosGAN training process has two steps that produce results almost comparable to other state-of-the-art techniques.The efficiency of CosGAN was compared by conducting experiments using benchmarked datasets.The approach was evaluated on different datasets and experimental results show superiority over existing state-of-the-art methods in terms of accuracy as well as generalization ability.This technique has numerous applications including wheeled robots,autonomous vehicles,warehouse automation,and all image-processing-based automation tasks so it can reshape the field of robotic vision with its ability to make robots adapt to new tasks and environments efficiently without requiring additional labeled data.It lays the groundwork for future expansions in robotic vision and applications.Although GAN provides a variety of outstanding features,it also increases the risk of instability and over-fitting of the training data thus making the data difficult to converge.

Underdetermined DOA estimation and blind separation of non-disjoint sources in time-frequency domain based on sparse representation method

This paper deals with the blind separation of nonstation-ary sources and direction-of-arrival （DOA） estimation in the under-determined case, when there are more sources than sensors. We assume the sources to be time-frequency （TF） disjoint to a certain extent. In particular, the number of sources presented at any TF neighborhood is strictly less than that of sensors. We can identify the real number of active sources and achieve separation in any TF neighborhood by the sparse representation method. Compared with the subspace-based algorithm under the same sparseness assumption, which suffers from the extra noise effect since it can-not estimate the true number of active sources, the proposed algorithm can estimate the number of active sources and their cor-responding TF values in any TF neighborhood simultaneously. An-other contribution of this paper is a new estimation procedure for the DOA of sources in the underdetermined case, which combines the TF sparseness of sources and the clustering technique. Sim-ulation results demonstrate the validity and high performance of the proposed algorithm in both blind source separation （BSS） and DOA estimation.

Ghost imaging based on the control of light source bandwidth

A scheme to improve the quality in ghost imaging(GI)by controlling the bandwidth of light source(BCGI)is proposed.The theoretical and numerical results show that the reconstruction result with high quality can be obtained by adjusting the bandwidth range of the light source appropriately,and the selection criterion of the bandwidth is analyzed by the power distribution of the imaging target.A proof-of-principle experiment is implemented to verify the theoretical and numerical results.In addition,the BCGI also presents better anti-noise performance when compared with some popular GI methods.

Perfect plane-wave source for a high-order symplectic finite-difference time-domain scheme

The method of splitting a plane-wave finite-difference time-domain （SP-FDTD） algorithm is presented for the initiation of plane-wave source in the total-field / scattered-field （TF/SF） formulation of high-order symplectic finite- difference time-domain （SFDTD） scheme for the first time. By splitting the fields on one-dimensional grid and using the nature of numerical plane-wave in finite-difference time-domain （FDTD）, the identical dispersion relation can be obtained and proved between the one-dimensional and three-dimensional grids. An efficient plane-wave source is simulated on one-dimensional grid and a perfect match can be achieved for a plane-wave propagating at any angle forming an integer grid cell ratio. Numerical simulations show that the method is valid for SFDTD and the residual field in SF region is shrinked down to -300 dB.

Tomographic inversion for microseismic source parameters in mining

We propose a new method for inverting source function of microseismic event induced in mining. The observed data from microseismic monitoring during mining are represented by a wave equation in a spherical coordinate system and then the data are transformed from the time-space domain to the time-slowness domain based on tomographic principle, from whichwe can obtain the signals related to the source in the time-slowness domain. Through analyzing the relationship between the signal located at the maximum energy and the source function, we derive the tomographic equations to compute the source function from the signals and to calculate the effective radiated energy based on the source function. Moreover, we fit the real amplitude spectrum of the source function computed from the observed data into the co-2 model based on the least squares principle and determine the zero-frequency level spectrum and the corner frequency, finally, the source rupture radius of the event is calculated and The synthetic and field examples demonstrate that the proposed tomographic inversion methods are reliable and efficient

Enhancing source domain availability through data and feature transfer learning for building power load forecasting

During the initial phases of operation following the construction or renovation of existing buildings,the availability of historical power usage data is limited,which leads to lower accuracy in load forecasting and hinders normal usage.Fortunately,by transferring load data from similar buildings,it is possible to enhance forecasting accuracy.However,indiscriminately expanding all source domain data to the target domain is highly likely to result in negative transfer learning.This study explores the feasibility of utilizing similar buildings(source domains)in transfer learning by implementing and comparing two distinct forms of multi-source transfer learning.Firstly,this study focuses on the Higashita area in Kitakyushu City,Japan,as the research object.Four buildings that exhibit the highest similarity to the target buildings within this area were selected for analysis.Next,the two-stage TrAdaBoost.R^(2) algorithm is used for multi-source transfer learning,and its transfer effect is analyzed.Finally,the application effects of instance-based(IBMTL)and feature-based(FBMTL)multi-source transfer learning are compared,which explained the effect of the source domain data on the forecasting accuracy in different transfer modes.The results show that combining the two-stage TrAdaBoost.R^(2) algorithm with multi-source data can reduce the CV-RMSE by 7.23%compared to a single-source domain,and the accuracy improvement is significant.At the same time,multi-source transfer learning,which is based on instance,can better supplement the integrity of the target domain data and has a higher forecasting accuracy.Overall,IBMTL tends to retain effective data associations and FBMTL shows higher forecasting stability.The findings of this study,which include the verification of real-life algorithm application and source domain availability,can serve as a theoretical reference for implementing transfer learning in load forecasting.

A NOVEL SWEPT LASER SOURCE BASED ON COMBINED TUNABLE FILTERS FOR OCT

A novel broad tunable bandwidth and narrow instantaneous line-width linear swept laser source using combined tunable filters working at 1,300 nm center wavelength is proposed.The combined filters consist of a fiber Fabry
Perot tunable filter and a tunable filter based on diffractive grating with scanning polygon mirror.In contrast to traditional method using single tunable filter,the trade-off between bandwidth and instantaneous line-width is alleviated.Parallel implementation of two semiconductor optical amplifiers with different wavelength range is adopted in the laser resonator for broadband light amplification.The Fourier domain mode locking swept laser source with combined tunable filters offers broadband tunable range with narrow instantaneous line-width,which is especially benefiting for high-quality optical frequency domain imaging.The proposed Fourier domain mode locking swept laser source provides a tuning range of 160 nm with instantaneous line-width of about 0.01nm at sweeping rate of 15 kHz,a finesse of 16,000 is thus achieved.

Renal function and choroidal thickness using swept-source optical coherence tomography in diabetic patients

AIM: To assess the relationship between choroidal thickness and renal function in diabetic patients. METHODS: Cross-sectional retrospective clinical study of 42 eyes of 21 ocular treatment-na?ve diabetic patients. Demographic data included: age, sex, type and course of diabetes. Ocular data included: severity of diabetic retinopathy;retinal thickness at the central macular region, as well as choroidal thickness at the central and paracentral quadrants, using automatically generated maps by swept-source optical coherence tomography;presence of cystic macular edema;and ocular axial length(AXL). Lab-test parameters included: glycated hemoglobin(HbA1c), albuminuria, albumin/creatinine ratio in urine, and glomerular filtration rate. RESULTS: A significant negative correlation was mainly observed between several choroidal thicknesses, age(P<0.020) and ocular AXL(P<0.030). On the contrary, a significant positive correlation was found between all choroidal thicknesses, HbA1 c(P<0.035) and albuminuria(P<0.040). CONCLUSION: Choroidal thickness can represent an additional tool to help clinicians predicting the renal status in ocular treatment-na?ve diabetic patients.

形象与态度——党的二十大报告的概念隐喻研究

概念隐喻广泛存在于人类表述特殊抽象概念的话语之中,它不仅是一种修辞手法,更是一种认知方式。习近平总书记在党的二十大报告中多次运用概念隐喻表达抽象概念。报告中的语言将理论阐释与通俗表达有机结合,既体现权威性,又不乏大众性。论文以党的二十大报告为语料,以概念隐喻理论为框架,运用隐喻识别程序MIPVU,考察了语料中的概念隐喻类型、分布和表达特点。研究发现,党的二十大报告中概念隐喻的使用频率较高,类型包括旅程隐喻、建筑隐喻、军事隐喻、自然隐喻、方位隐喻和拟人隐喻,其中旅程隐喻使用频率最高。语篇通过运用不同的概念隐喻类型塑造了国家的正面形象,表达了中国人民必将团结一致、砥砺前行的态度。

天津某环境背景噪声分析

连续波形检查中发现,天津尔王庄(EWZ)等台明显受到某未知噪声的干扰,且EWZ台受影响程度最大,此噪声常年如此,不随昼夜时间发生变化,且已存在多年。长期观测还发现,此噪声每年会有几天观测不到,但时间并不确定。与通常的环境背景噪声不同,此噪声具有一些固定特征。本文对其从本身特征、影响范围、频谱特性、对地震的影响等几个方面进行了研究。研究表明,噪声不受天气、季节等自然因素影响,具有随距EWZ台距离增大而振幅变小的变化特征,其主要频率范围在1~2 Hz之间。噪声的影响范围较广泛,可被周围12个台站记录到,最远到达YGZ台,距离EWZ台约58 km。通过偏振分析,发现噪声的传播具有特定的方向性,初步判断此噪声具有固定源噪声的特点。此噪声会对地震记录造成影响,使震相拾取率降低,甚至完全淹没在噪声中,通过带阻频率滤波会有所改善。

基于时域有限差分法的电性源感应-极化效应三维数值模拟

时域电性源电磁探测法是一种可以有效快速探测矿产资源的方法。极化效应会导致电性源电磁响应快速衰减,发生符号反转现象。本文采用Cole-Cole模型描述极化效应,利用整数阶有理逼近算法实现任意分数阶Cole-Cole模型有理化;采用Yee氏网格对仿真区域进行剖分,基于时域有限差分(finite-difference time-domain, FDTD)法实现电性源感应-极化效应三维数值模拟。本文对三种典型模型(均匀半空间模型、极化半空间模型和三维极化体模型)的电磁响应进行数值模拟,结果表明:均匀半空间模型的电磁响应与解析解基本吻合,并且相对误差小于10%,证明了三维数值模拟方法的正确性;极化半空间模型与三维极化体模型的电磁响应均在晚期出现了负响应,与极化理论结果相符合。

多源频率域地震勘探技术及应用

在城市地球物理勘探中,因受到城市中各种强电磁机械干扰和交通建筑阻隔,诸多地球物理勘探技术勘探效果不佳。本文基于瞬态面波、微动台阵和微动谱比,采用多点三分量低频传感器获取的多源频率域地震数据,综合提取两种特征曲线后进行联合反演,实现融合3种技术的多源频率域地震勘探。其中,提出了双源综合频散谱提取技术、双源特征曲线联合反演技术和基于非均匀介质的等效均匀介质正演技术,勘探成果的精度和效率均得到了提高,在工程实践中得到验证。

基于多源域适应的缺陷类别预测方法

随着规模和复杂性的迅猛膨胀,软件系统中不可避免地存在缺陷.近年来,基于深度学习的缺陷预测技术成为软件工程领域的研究热点.该类技术可以在不运行代码的情况下发现其中潜藏的缺陷,因而在工业界和学术界受到了广泛的关注.然而,已有方法大多关注方法级的源代码中是否存在缺陷,无法精确识别具体的缺陷类别,从而降低了开发人员进行缺陷定位及修复工作的效率.此外,在实际软件开发实践中,新项目通常缺乏足够的缺陷数据来训练高精度的深度学习模型,而利用已有项目的历史数据训练好的模型往往在新项目上无法达到良好的泛化性能.因此,首先将传统的二分类缺陷预测任务表述为多标签分类问题,即,使用CWE(common weakness enumeration)中描述的缺陷类别作为细粒度的模型预测标签.为了提高跨项目场景下的模型性能,提出一种融合对抗训练和注意力机制的多源域适应框架.该框架通过对抗训练来减少域(即软件项目)差异,并进一步利用域不变特征来获得每个源域和目标域之间的特征相关性.同时,该框架还利用加权最大均值差异作为注意力机制,以最小化源域和目标域特征之间的表示距离,从而使模型可以学习到更多的域无关特征.在构建的包含8个真实世界开源项目的数据集上的实验表明,所提方法对比最先进的基线方法取得了显著的性能提升.

不同工况及类别下热力系统故障诊断的多源域自适应方法

针对不同负荷工况下,热工参数数据分布差异大且故障类别不一致的问题,提出了一种基于多源样本加权域对抗网络的热力系统故障诊断方法。首先,构建领域共享的一维卷积神经网络以提取多个源域和目标域的深度判别特征;其次,引入加权机制和域一致性损失度量样本,以降低仅存在于源域的故障类别的负迁移影响;然后,通过多域判别器的对抗学习实现每对源域和目标域的特征差异对齐;最后,构建多分类器对齐模块以提高预测的一致性,从而实现多源域不同工况下热力系统故障的准确诊断。借助某600MW超临界机组全范围仿真系统进行故障仿真实验,结果验证了所提方法的鲁棒性和优越性。

基于多源域迁移学习的带式输送机剩余寿命预测方法

煤矿开采过程中,带式输送机运行环境恶劣、工况复杂,致使获得的传感监测数据量有限且存在大量噪声干扰,严重限制了其剩余寿命预测的准确度。针对该问题,提出了一种多源域迁移学习剩余寿命预测方法,充分利用煤矿运输过程中积累的带式输送机多工况数据,以达到准确预测其关键零部件托辊轴承剩余寿命的目的。首先构建集成多尺度卷积神经网络和双向门控循环单元(MCNN-BiGRU)的设备退化特征提取模型,对单工况数据进行特征提取挖掘,并使用PSO算法确定模型超参数。在此基础上,加入多源域迁移学习(MDT)方法,利用多个工况数据进行剩余寿命预测,通过最大均值差异(MMD)与相互关系对齐(CORAL)联合损失拉近各源域数据分布差异,解决因数据量少导致的模型训练精度不高的问题。最后以煤矿实际生产数据集为例进行实验,结果表明:MDT-MCNN-BiGRU模型的预测效果较好,Savitzky-Golay滤波去噪后模型性能得以进一步提升;使用IMS数据集与现有方法进行比较,发现所提方法预测准确度较高,对煤矿运输设备健康管理具有一定的指导意义。

汽车组合仪表有意电磁干扰空时域联合分离系统设计

有意电磁波与汽车仪器电磁传输波产生的混合干扰,降低了汽车组合仪表的稳定运行能力,造成仪表设备失灵;为避免上述情况的发生,设计汽车组合仪表有意电磁干扰空时域联合分离系统;在步进电机驱动电路中,设置汽车组合仪表控制器、分离电源与PCB电路板、有意电磁干扰数字传感器;分配有意电磁波空时域源地址,通过计算干扰信号平坦度的方式,确定脉冲参数的取值范围,实现汽车组合仪表有意电磁的空时域干扰效应分析;估算有意电磁波干扰信号的数目,按照信号分解的处理原则,求解联合分离修正函数,实现电磁干扰的联合分离;在设置频率范围在100 MHz~1 GHz之间、幅度范围为0.5~5 V、脉冲宽度范围为10~100 ns区间后,通过实验结果表明,设计系统应用后可以在混合信号中,按照空域、时域标准的不同分离有意电磁波,且分离后的信号波长、信号频率依然在1.0μm、300 Hz左右,不会因混合干扰降低汽车组合仪表的稳定运行能力。

基于多源域自适应残差网络的滚动轴承故障诊断

针对传统无监督领域自适应方法扩展到多工况滚动轴承故障诊断场景适用性较弱的问题,提出了一种多源域自适应残差网络(multi-source domain adaptive residual network,MDARN),通过对齐来自多个源域的相关子域,从而提高模型在多工况下的故障诊断性能。首先,利用ResNeXt残差网络从源域和目标域充分提取可迁移特征;然后,引入局部最大平均差异(local maximum mean difference,LMMD)准则,以两个源域的子域为基础对齐目标域中相关子域,减少相关子域间和全局域间的分布差异;最后,利用美国凯斯西储大学轴承数据集和MFS机械综合故障试验台产生的真实的轴承振动数据集,对所提方法进行了试验验证。结果表明,该方法在多工况下的平均故障诊断精度高达99.76%。与现有代表性方法相比,所提方法具有更好的故障诊断效果。

基于多源域数据与机器学习算法的转子不平衡故障诊断

国内能源生产装置规模大型化发展趋势明显,与其配套的旋转机械设备发生故障导致的非计划停机将会造成严重的经济损失与重大安全问题。转子不平衡贯穿了旋转机械设备的整个生命周期,服役转子的状态诊断格外重要。针对大型旋转机械振动测点较多,振动信号具有非平稳特征等问题,提出基于多源域数据提取与机器学习算法的转子不平衡故障诊断模型。首先以多源振动监测数据为驱动,根据互相关系数提取故障信息丰富的振动信号,融合时域、频域、时频域等多域特征构建高维混合特征空间;其次利用基于t分布的随机邻域嵌入方法揭示高维空间的特征信息,反映为可视化的三维空间;最终通过最邻近节点算法进行故障分类,判断转子的不平衡质量与相位。本文提出利用互相关系数表征多源数据的故障信息丰富程度,并结合机器学习手段判断转子不平衡类型。通过设计不同附加质量的转子在多转速下不平衡状态实验,验证了所提模型的有效性,解决了转子在线诊断和现场动平衡问题。

基于深度学习的跨域辐射源个体识别综述

基于深度学习的辐射源个体识别已经成为识别和认证无线设备的主要研究方法之一。然而,基于传统深度学习的辐射源个体识别算法不能直接应用于跨域(跨信道条件、跨接收机、跨接收时间等)辐射源个体识别场景,因为一个域上的数据训练的模型用于另一个域上进行推理,其效果通常会变差。调研了现有基于对比学习、迁移学习、域适应等先进深度学习的跨域辐射源个体识别方法,整理和归纳了跨域辐射源个体识别相关的开源数据集。分析了跨域辐射源个体识别存在的难题与挑战,展望了跨域辐射源个体识别发展趋势及未来研究方向,以助力深度学习在复杂电磁环境下辐射源个体识别的实用化。

基于多源域适应JYPLS迁移的间歇过程质量预测

针对间歇过程数据不足,单源域迁移存在模型偏移,跨域信息损失导致建模效果不佳、负迁移等问题,结合域适应学习和多源域学习方法的优势,提出一种基于多源域适应联合Y偏最小二乘(joint-Y partial least squares,JYPLS)迁移的间歇过程质量预测方法。该方法通过迁移学习使用相似旧过程的数据辅助新过程建模,提高建模效率和模型预测精度;采用多源域适应的方式,通过引入多个源域,有效避免了负迁移;基于域适应思想减少源域和目标域之间的边缘概率分布差异,使得源域知识在目标域更好地泛化。最后,通过青霉素发酵过程的仿真案例验证了所提方法的有效性。