AUTOMATIC MULTILEVEL THRESHOLDING METHOD BASED ON MAXIMUM ENTROPY

In the multilevel thresholding segmentation of the image, the classification number is always given by the supervisor. To solve this problem, a fast multilevel thresholding algorithm considering both the threshold value and the classification number is proposed based on the maximum entropy, and the self-adaptive criterion of the classification number is given. The algorithm can obtain thresholds and automatically decide the classification number. Experimental results show that the algorithm is effective.

The deconvolution of lunar brightness temperature based on the maximum entropy method using Chang'e-2 microwave data

A passive and multi-channel microwave sounder onboard the Chang＇e-2 orbiter has successfully acquired microwave observations of the lunar surface and subsurface structure. Compared with the Chang＇e-1 orbiter, the Chang＇e-2 orbiter obtained more accurate and comprehensive microwave brightness temperature data, which are helpful for further research. Since there is a close relationship between mi- crowave brightness temperature data and some related properties of the lunar regolith, such as the thickness, temperature and dielectric constant, precise and high resolution brightness temperature data are necessary for such research. However, through the detection mechanism of the microwave sounder, the brightness temperature data ac- quired from the microwave sounder are weighted by the antenna radiation pattern, so the data are the convolution of the antenna radiation pattern with the lunar brightness temperature. In order to obtain the real lunar brightness temperature, a deconvolution method is needed. The aim of this paper is to solve the problem associated with per- forming deconvolution of the lunar brightness temperature. In this study, we introduce the maximum entropy method （MEM） to process the brightness temperature data and achieve excellent results. The paper mainly includes the following aspects： first, we introduce the principle of the MEM; second, through a series of simulations, the MEM has been verified as an efficient deconvolution method; and third, the MEM is used to process the Chang＇e-2 microwave data and the results are significant.

PHASE EXTENSION AND STRUCTURE ANALYSIS BY THE MAXIMUM ENTROPY METHOD

Excellent results are obtained in structure analysis with jew phases of structure factors by the maximum-entropy method (MEM) for CaGaN PbCO3 and ReBe22 single crystals. The computation time and memory space are minimized by symmetry operations so that structure analysis by the MEM can be carried out with a personal computer.

A Microcomputer Fitted Method for Determination of the True ODF of Hexagonal Materials according to the Maximum-Entropy Estimate

A method which is especially suitable for microcomputer calculation of the true orientation distribution function (ODF) according to the maximum-entropy estimate is proposed for hexagonal system polycrystalline materials with physical symmetry.The resultant computational software system has been also designed and first carried out in a microcomputer PANAFACOM-U1200 being on line with the X-ray diffractometer D/max-3A.The simu- lated calculation shows that the method is concisely pragmatic and easily popularized,while the results obtained are trust worthy.

A Maximum-Entropy Method for Estimating the Spectrum

Based on the maximunl-entropy （ME） principle, a new power spectral estimator for random waves is derived in the form of S（ω）=a/8H^2^-（2π）^（d＋2）exp[-b（2π/ω）^n],1）y solving a variational problem subject to some quite general constraints. This robust method is comprehensive enough to describe the wave spectra even in extreme wave conditions and is superior to periodogranl method that is not suit＇able to process comparatively short or intensively unsteady signals for its tremendous boundary effect and some inherent defects of FKF. Fortunately, the newly derived method for spectral estimation works fairly well, even though the sample data sets are very short and unsteady, and the reliability and efficiency of this spectral estimator have been preliminarily proved.

Texture Analysis of Thin Titanium(TA1)Strip by Maximum-entropy Method

The Maximum-entropy Method(MEM)for determining the complete ODF(orientation distribution function),accompanied with the equal-volume partitioning technique for quantitative texture analysis,was first tested in analysing the texture of a commercial purity titanium strip.The experimentally measured results indi- cate that the rolling planes of most grains in this sample are parallel to the{1010}and the{1210}with about ~±10°spread while the rolling directions nearly distribute uniformly and their volume fractions are 19.46％ and 18.70％ respectively.Besides,there are still two weaker texture components,(7526)[1544]and (1105)[2311],with 3.24％and 4.17％respectively.

Generalized Maximum Fuzzy Entropy Methods with Applications on Wind Speed Data

This study is connected with new Generalized Maximum Fuzzy Entropy Methods （GMax（F）EntM） in the form of MinMax（F）EntM and MaxMax（F）EntM belonging to us. These methods are based on primary maximizing Max（F）Ent measure for fixed moment vector function in order to obtain the special functional with maximum values of Max（F）Ent measure and secondary optimization of mentioned functional with respect to moment vector functions. Distributions, in other words sets of successive values of estimated membership function closest to （furthest from） the given membership function in the sense of Max（F）Ent measure, obtained by mentioned methods are defined as （MinMax（F）Ent）m which is closest to a given membership function and （MaxMax（F）Ent）m which is furthest from a given membership function. The aim of this study consists of applying MinMax（F）EntM and MaxMax（F）EntM on given wind speed data. Obtained results are realized by using MATLAB programme. The performances of distributions （MinMax（F）En0m and （MaxMax（F）Ent）m generated by using Generalized Maximum Fuzzy Entropy Methods are established by Chi-Square, Root Mean Square Error criterias and Max（F）Ent measure.

Reconstructing parton distribution function based on maximum entropy method

A new method based on the maximum entropy principle for reconstructing the parton distribution function(PDF)from moments is proposed.Unlike traditional methods,the new method does not require any artificial assumptions.For the case of moments with errors,we introduce Gaussian functions to soften the constraints of moments.Through a series of tests,the effectiveness and reconstruction efficiency of this new method are evaluated comprehensively,demonstrating that this method is reasonable and can achieve high-quality reconstruction with at least the first six moments as input.Finally,we select a set of lattice quantum chromodynamics(QCD)results regarding moments as input and provide reasonable reconstruction results for the pion.

ε-OPTIMAL SOLUTION FOR CONTINUOUS MINIMAX PROBLEMS WITH MAXIMUM-ENTROPY FUNCTION

This paper describes and explores a maximum-entropy approach to continuous minimax problem, which is applicable in many fields, such as transportation planning and game theory. It illustrates that the maximum entropy approcach has easy framework and proves that every accumulation of {x_k} generated by maximum-entropy programming is -optimal solution of initial continuous minimax problem. The paper also explains BFGS or TR method for it. Two numerical exam.ples for continuous minimax problem are

MAXIMUM ENTROPY DECONVOLUTION OF XPS PEAK

Necessity of XPS spectrum deconvolution, disadvantages of the traditional Fast Fourier Transform decon-volution method (FFT) , principle, method and advantages of Maximum Entropy Deconvolution Method (MEM) are de-scribed. Criteria for determing the number of data points sam-pled in MEM are the main point disccussed in the paper,some XPS deconvolution applications of our MEM software show that the MEM makes XPS deconvolution much easier than the traditional FFT method.

面向稀疏奖励的机器人操作技能学习

基于深度强化学习的机器人操作技能学习成为研究热点,但由于任务的稀疏奖励性质,学习效率较低.本文提出了基于元学习的双经验池自适应软更新事后经验回放方法,并将其应用于稀疏奖励的机器人操作技能学习问题求解.首先,在软更新事后经验回放算法的基础上推导出可以提高算法效率的精简值函数,并加入温度自适应调整策略,动态调整温度参数以适应不同的任务环境;其次,结合元学习思想对经验回放进行分割,训练时动态调整选取真实采样数据和构建虚拟数的比例,提出了DAS-HER方法;然后,将DAS-HER算法应用到机器人操作技能学习中,构建了一个稀疏奖励环境下具有通用性的机器人操作技能学习框架;最后,在Mujoco下的Fetch和Hand环境中,进行了8项任务的对比实验,实验结果表明,无论是在训练效率还是在成功率方面,本文算法表现均优于其他算法.

弧形绝缘子金属连接环局部机械损坏识别仿真

弧形绝缘子金属连接环通常安装在电力设备上,由于其所处的环境复杂,且存在各种环境干扰因素的影响,使得其局部出现机械损坏时,识别结果不准确。为了精准识别连接环局部机械损坏,提出一种弧形绝缘子金属连接环局部机械损坏识别方法。通过形态学开—闭重构运算对弧形绝缘子金属连接环图像展开建模,使用差分运算获取差分图像,并结合最大熵值法对其分割。根据弧形绝缘子金属连接环的特点,排除非损伤部分对图像识别的干扰。采用基于分形理论的差分盒维法,将图像划分为多个子块并计算其分形维数值,并判断图像是否在数值范围内,从而实现对连接环局部机械损坏的识别。实验结果表明,所提方法可以精准识别弧形绝缘子金属连接环局部机械损坏,具有比较强的适用性。

物流大通道的形成机理与识别方法

物流大通道是国家综合立体交通网的重要功能系统,是国内国际双循环相互促进的基础。为更加精准识别物流大通道,通过国内外概念辨析进一步明确物流大通道的定义,研究提出物流大通道的形成是以交通运输基础设施网络化发展为根本依托,是网络成分互补的横向整合和节点集聚的纵向整合共同作用的结果,其表征是枢纽点极化、通行量规模化。在此基础上,明确了“先找点、后定线”的识别思路,提出以区位熵定“点”、最大树(边)定“线”为核心的基于城市、交通和经济数据的识别方法,通过多源数据收集、节点识别、线路识别、通道汇集、通道识别,最终提出“八纵七横”的物流大通道布局。相比于传统方法,选点定线的识别方法精度更高,能够进一步展示货物流向、运输结构、货运规模等运行特征,并为开展物流大通道与经济发展相关分析提供了可能。

基于热释光剂量计的少道谱仪研究

针对脉冲X射线单次脉冲时间短,瞬时剂量率大的特点,研究了一种基于热释光剂量计(TLD)的少道谱仪,该谱仪可用于脉冲X射线的能谱测量。本研究对该谱仪的吸收片及探测器进行选型,结合蒙特卡罗方法模拟少道谱仪对不同能量单能光子的响应矩阵,分析其对谱仪能量分辨率的影响,使用最大熵法开展脉冲X射线解谱验证工作,分析初始迭代谱对解谱效果的影响。结果表明,基于热释光剂量计研制的少道谱仪,可以开展短脉冲、高剂量率脉冲X射线能谱测量工作,同时也适用于其他大剂量率辐射场的能谱测量。

MEMS陀螺仪冗余配置故障检测方法

针对姿态测量系统的可靠性问题,提出了一种冗余配置方式的微机电系统(micro-electro-mechanical system,MEMS)陀螺仪故障检测方法。首先,基于三轴MEMS陀螺仪研究了冗余配置的形式,给出了相应的故障评价函数,并给出了最优配置角度的计算方法;其次,根据冗余配置建立陀螺仪故障检测数学模型,设计了MEMS陀螺仪硬故障检测方法,并进一步证明了陀螺仪的软故障和陀螺仪输出噪声的相关性,通过最大似然估计给出了软故障检测判决函数;最后,利用自主构建的故障检测硬件平台进行了实验验证,成功实现了对发生硬故障及软故障陀螺仪敏感轴的检测。该方法将MEMS陀螺仪的冗余配置扩展到了敏感轴水平,提高了MEMS陀螺仪的数据利用率,增加了系统可靠性。

熵权模糊改进法在宁夏黄河水质评价中的应用

为了解宁夏境内黄河流域的水质状况,选取银古公路桥、平罗黄河大桥、中卫香山湖、吴忠金沙湾、固原沟圈5个监测断面在2021年4月、6月、8月、12月4个时期的溶解氧、高锰酸盐指数、氨氮、总磷、总氮5项评价指标,共获得80组水样。针对熵权模糊评价法在权重趋于1时其微小变化会引起熵权成倍变化的病态问题和由于最大隶属度原则不能普遍适用而导致评价失效的问题,提出了一种熵权模糊改进的评价方法,通过改进熵权法确定权重和加权等级,建立改进的模糊综合评价模型,对水质样品进行水质分析和综合评价。结果表明,在权重分配方面,相比于隶属度法和传统的熵值法,改进熵权法分配的权重更为合理;在水质评价方面,加权等级法评定出的结果更符合研究区域的真实水质状况;宁夏境内黄河流域Ⅰ类和Ⅱ类水质占比超过90%,水质状况良好,5个断面4个时期的水质综合评价结果显示,12月份的水质最优,Ⅰ类水质占比85%;对水质的时空分析显示氨氮为全局性污染因子,溶解氧为区域性污染因子。改进的模糊综合评价模型对宁夏黄河水质的评价结果更为客观合理。

基于双边滤波和Frangi滤波的织物疵点检测方法

为提高织物疵点检测效率,提出了一种基于双边滤波和多尺度线性滤波(Frangi)相结合的织物疵点检测方法。首先,采用双边滤波对织物图像进行处理,以减少织物背景纹理信息对检测结果的影响;然后,应用多尺度线性滤波器对经双边滤波后的织物图像进行二次滤波,增强织物疵点区域信息;最后,使用最大熵法对增强后的织物图像进行分割,定位出织物疵点区域。通过对油污、破洞、线头、断经、双纬、节纱、吊经织物图像进行实验,发现该方法能快速有效检测出疵点信息。

基于极大熵方法的反向散射通信资源分配算法

针对因反向散射通信节点存在重叠、混乱分布情况,导致通信效率降低、资源分布不均匀问题,提出基于极大熵方法的反向散射通信资源分配算法。以资源累积奖励期望作为目标函数,利用极大熵方法求得资源节点在不同状态下的局部最优值,采用贝尔曼迭代逐次递归求解资源分配最大熵值参数。基于分配信息熵概念,根据信息熵与资源均匀性间的正比关系,将反向散射通信不同区域节点的信息熵值和时延变化参数作为约束条件。将反向散射系数、传输功率、反射端子载波数的实时状态输入约束函数,获取最优资源分配因子,按照分配因子对通信参数逐一调试直至满足最佳资源分配条件。实验结果证明,所提方法性能表现优异,应用后通信系统能效得到了明显改善,资源分配效果更优。

MEM在铁路UM2000轨道电路信号接收中的应用

最大熵法(MEM)是一种常用的现代谱估计法,它与经典谱估计法相比,分辨率比较高.UM2000是法国应用于高速铁路中的一种数字编码轨道电路,其信号特点是频率间隔比较小,所以要求其频谱的分辨率高,同时实时性要好.本文研究了用MEM在UM2000轨道电路信号分析中的应用.

Risk calculation method for complex engineering system

This paper presents a rapid and simple risk calculation method for large and complex engineering systems, the simulated maximum entropy method （SMEM）, which is based on integration of the advantages of the Monte Carlo and maximum entropy methods, thus avoiding the shortcoming of the slow convergence rate of the Monte Carlo method in risk calculation. Application of SMEM in the calculation of reservoir flood discharge risk shows that this method can make full use of the known information under the same conditions and obtain the corresponding probability distribution and the risk value. It not only greatly improves the speed, compared with the Monte Carlo method, but also provides a new approach for the risk calculation in large and complex engineering systems.