Online reconstruction-free single-pixel image classification

In single-pixel imaging, a series of illumination patterns are projected onto an object and the reflected or transmitted light from the object is integrated by a photodetector (the single-pixel detector). Then, from the set of received photodetector signals, the image of the object can ultimately be reconstructed. However, this reconstruction is not only computationally expensive, but also unnecessary for purposes such as image classification tasks. This work proposes a reconstruction-free multi-class image classification framework that, unlike most of the existing approaches, exploits the sequential nature of the problem. Indeed, by accumulating evidence of the sequence of scalar values, a decision is made after each measurement on whether already classifying the object being imaged, or waiting for more measurements. This online decision relies on a mechanism to achieve a recognition-delay trade-off that induces behaviours within the conservative-to-aggressive spectrum, which suit distinct requirements in different applications. Additionally, the presentation order of the illumination patterns makes a difference in terms of the reconstruction quality (if required) and classification performance when a limited number of patterns is used. Nevertheless, in many cases, simple data- and task-agnostic orders, such as random or frequency-based orders, are commonly used. To address this, a novel sparse-representation-based strategy is presented that sorts the patterns according to their general and discriminability utilities. Both, the online classification framework including the recognition-delay trade-off mechanism, and the data- and task-aware pattern ordering proposed, are experimentally assessed, with encouraging results, on the MNIST digits and CalTech 101 Silhouettes datasets.     

Imperfect market,emissions trading scheme,and technology adoption: A case study of an energy-intensive sector

It is widely accepted that the firms included in an emissions trading scheme (ETS) come mostly from oligopolistic industries. The “exclusionary manipulation” of these heterogeneous emitters can distort both output and permit markets and lead to differences in abatement technology adoption. We studied the impacts of asymmetric firms' market power on the diffusion of abatement technologies. A model for technology adoption among heterogeneous firms has been established, which takes into account diversity in production capacity and the integration of firms' strategic behaviour in both the carbon permit and the output markets. Our model reveals that, considering the direct and strategic effects in adoption benefits, firms' production capacity can directly determine their sequence order of adoption, and their market power can accelerate the diffusion of a new abatement technology. A case study of an energy-intensive sector in China is illustrated to support the conclusions derived from the model and help policymakers better understand the diffusion of abatement technologies under imperfect market structure.     

Magnetic properties of melt-spun ribbons(Sm_(1-x)Zr_x)(Fe_(0.92)Ti_(0.08))_(10) with ThMn_(12) structure and their hydrides

The structure and magnetic hysteresis properties of the cast Sm_(1-x)Zr_x(Fe_(0.92)Ti_(0.08))_(10)(x = 0-0.3)alloys and melt-spun ribbons prepared from them were studied.In the cast alloy with x0.2, a considerable amount of the eutectic phase is found in the SEM micrographs.Analysis of the temperature dependences of the magnetic susceptibility and XRD patterns allows amorphous state in the as-spun ribbons with x0.2 to be determined.The specific magnetization measured in a field of 17 kOe and remanence decrease with increasing annealing temperature from 800 to 900 ℃ and weakly depend on Zr concentration.The maximal value of coercivity Hc = 4.7 kOe is obtained on the ribbons with x = 0.2 after annealing at 850℃ for 10 min.After additional hydrogenation of the ribbons,both the coercivity and remanence increase by 54% and 7%,respectively.     

基于深度概率图模型的鲁棒人脸画像合成

针对基于数据驱动的人脸画像合成算法像素特征缺乏对光照变化和复杂背景的鲁棒性,常合成低质量的画像的问题,文中提出基于深度概率图模型的鲁棒人脸画像合成算法.采用预处理方法调整测试照片的光照亮度和人脸姿态,使之与训练照片一致.采用深度特征代替像素特征进行近邻匹配,采用深度概率图模型对画像重建权重和深度特征权重联合建模,得到合成画像的最佳重构表示.为了提高画像合成速度,提出快速近邻搜索方法.实验验证文中算法的鲁棒性和快速性.     

激光脉冲编码抗有源干扰效果研究

激光脉冲编码是激光制导武器的抗干扰措施之一。角度欺骗式干扰和高重频干扰是目前半主动激光制导武器的主要有源干扰来源。为研究不同激光脉冲编码方式对激光半主动制导武器抗这两种干扰性能的影响,本文针对敌方激光告警机的识别算法与我方导引头的解码过程,提出自相关函数与归一化互相关函数评价方法,并对目前主要编码方式进行仿真,仿真结果表明:激光脉冲编码的抗角度欺骗式干扰能力受编码序列周期性与脉冲间隔随机性的影响；抗高重频激光干扰能力受编码序列脉冲间隔随机性的影响；LFSR状态码的抗角度欺骗式干扰与抗高重频干扰效果均优于其他编码方式。     

Effect of lanthanum doping on water content and diffusion properties of corundum for producing transparent ceramics

The effect of doping corundum with lanthanum on the content of water and other gas impurities, on the diffusion characteristics, on the unit cell size and static strength of fine-crystalline corundum synthesized under thermovapours conditions in a supercritical water fluid was studied.Content, release and diffusion coefficients of water and other volatile impurities in corundum were measured by kinetic thermodesorption mass spectrometry. Water is the main volatile impurity in fine-crystalline corundum. It was found that the introduction of lanthanum into the structure of corundum reduces several times the content of water and other gas-liquid impurities. An increase in the diffusion coefficient of water during the doping of corundum with lanthanum leads to an increase in the degassing rate of fine-crystalline raw materials. When doping corundum, lanthanum, being localized in structural defects, not only displaces structural hydroxyl groups with the formation of strong aluminate-like bonds, but also decreases in the unit cell and increases in the average static strength of the crystals. La-doped corundum, due to the reduced content of water and other gas impurities, is a promising raw material for the production of transparent ceramics.     

磁悬浮控制力矩陀螺高速转子的高精度位置控制

在受到陀螺效应、动框架效应等影响后产生的磁力非线性问题是磁悬浮控制力矩陀螺(MSCMG)高速转子位置精度下降的主要因素。为解决以上问题,提高转子位置精度,本文分析了转子所受磁力的特性,建立了转子系统非线性动力学模型,提出了神经网络滑模控制方法。设计滑模控制律,采用径向基函数神经网络逼近控制律中的非线性模型,自适应算法根据误差在线调整神经网络的权值,同时可以保证整个系统的稳定性。仿真和实验结果表明,所提出方法的转子位置精度达到99%,稳态误差为0.000 2 mm。神经网络滑模控制可以实现MSCMG转子系统的高精度位置控制。     

特种车辆舱室送风系统布局仿真优化

特种车辆由于其功能特殊性和封闭性，无法像民用车辆一样利用窗户的开闭来控制舱室通风，长时间驾驶特种车辆，驾驶舱面临空间小、冷源少、热负荷大、缺乏新鲜空气等问题，且外部环境条件复杂。这就要求特种车辆舱室送风系统精细设计，满足人员和多个装置的热舒适和热控需求，因此有必要基于特殊车辆的驾驶舱特点，合理布局并优化通风系统，保证整舱热要求。为了让座舱送风与舱内热源热交换尽量完全，带出更多的热量，首先针对某型特种车辆，建立了典型舱室和乘员的三维物理模型和仿真模型。针对10种送风模式进行了详细的气流组织仿真优化分析，获得了特种车辆舱室多物理场。气流组织优化是从风口形式和送风口位置两方面进行，风口形式优化中提出多种风口组合形式，分别进行仿真。以设备温控、人员热舒适和空气龄等为目标，利用评价函数对仿真结果进行评价，对10种类型的送风工况进行了仿真。从仿真结果中，以上述多优化目标为依据，进一步开展了针对初步优选的送风形式的风口位置优化。在风口形式优化结果的基础上，采用遗传算法，将位置参数设为优化参数，选取两个评估函数作为优化的目标函数，并将头足温差等约束条件设置到优化模型中，对仿真结果进行筛选和逆优化研究，最终得到最优风口位置，完成气流组织优化。研究对于全密封特种车辆的有限空间热舒适性和空气质量研究具有一定意义。