基于Deep-Semi-NMF的苹果斑点落叶病检测方法

[目的/意义]苹果斑点落叶病易导致苹果树叶过早脱落,从而影响苹果品质和产量。因此,如何准确检测此病一直是苹果树病害精准防治的热点问题。由于逆光等因素影响,传统基于图像分割的病斑检测方法难以在复杂背景下准确检测病斑区域边界,亟需发展苹果斑点落叶病检测新方法,助力苹果树病害精准防治。[方法]针对上述问题,本研究从图像异常检测的角度出发,考虑复杂背景干扰,采用深度半非负矩阵分解理论,结合鲁棒性好的马氏距离度量,提出一种新的深度半非负矩阵分解的马氏距离异常检测方法(Deep Semi-Non-Negative Ma⁃trix Factorization-Based Mahalanobis Distance-Anomaly Detector,DSNMFMAD)。该方法首先利用深度非负矩阵分解(Deep Semi-Non-Negative Matrix Factorization,DSNMF)提取图像中低秩的背景部分和稀疏的异常部分。然后采用基于奇异值分解特征子空间的马氏距离构建病斑检测器,检测器通过计算异常部分每个像元的异常度来标记病斑。最后,分别构建了实验室和自然条件下的两个苹果斑点落叶病数据集,用以验证提出方法的有效性。[结果和讨论]DSNMFMAD在实验室条件和自然条件下对苹果斑点落叶病的识别准确率分别达到了99.8%和87.8%;平均检测速度为0.087和0.091 s/幅。相较于4种经典的异常检测方法和1种深度学习模型,本研究所提出方法的检测准确率在实验室条件下分别提高了0.2%、37.9%、10.3%、0.4%和24.5%;在自然条件下分别提高了2.5%、32.7%、5%、14.8%和3.5%。[结论]本研究提出的DSNMFMAD能够通过DSNMF有效地将图像中的异常部分提取出来,并利用构建的病斑检测器准确地将苹果斑点落叶病位置检测出来。即使在复杂背景条件下,该方法亦获得了比对比方法更高的检测准确度,展现出了优异的病斑检测性能,为苹果斑点落叶病的检测与防治提供了技术参考依据。

基于改进AdvSemiSeg的半监督遥感影像作物制图方法

作物精准遥感制图对于农业资源调查与管理具有重要意义。深度学习为实现精准高效作物制图提供了技术支持。为了缓解深度学习对标记样本的依赖,本文提出了一种改进AdvSemiSeg的半监督遥感影像作物制图方法。所提方法引入STMF-DeepLabv3+作为对抗学习中的生成网络,通过Swin Transformer(ST)和多尺度特征融合(Multi-scale fusion, MF)模块提高生成网络特征编码能力和语义表达能力,改善遥感影像作物分割效果;此外,在判别网络中引入通道注意力(Efficient channel attention, ECA)模块,对不同通道特征图的表征信息进行自适应学习,增强判别网络对不同通道特征的感知能力。模型训练过程中,判别网络为生成网络提供高质量的伪标签和对抗损失,有效提高生成网络的泛化能力。采用所提方法与几种先进的半监督语义分割方法对内蒙古河套灌区遥感影像种植信息进行提取,本文方法性能最优。

Reversible Semi-Fragile Watermarking Technique for Integrity Control of Relational Database

Reversible watermarking schemes for relational database are usually classified into two groups: robust schemes and fragile schemes. The main limitation of existing reversible fragile methods is that they cannot differentiate between legal and malicious modifications. In this paper, we introduce a novel lossless semi-fragile scheme based on prediction-error expansion for content protection of relational database. In the proposed method, all attributes in a database relation are first classified according to their sensitivity to legitimate updates. Then, the watermark is embedded by expanding the prediction error of the two least significant digits of securely selected attributes. At watermark extraction, the proposed method has the ability to fully restore the original data while detecting and localizing tampering. The applicability of our method is demonstrated theoretically and experimentally.

基于路由协议的Semi-TCP网络层实现方法

为提高移动自组网的网络性能和推进Semi-TCP方法的实际应用,提出基于路由协议的Semi-TCP网络层逐跳拥塞控制实现方法。通过在网络层的路由协议中加入拥塞信息广播机制,控制节点对数据包的发送频率,保持链路高速传输数据的同时,降低拥塞程度。在EXata仿真平台进行了验证,在设计的动态拓扑实验中,文中方法的吞吐量相比TCP New Reno提高了169.57%,在引入多TCP应用竞争链路后比原Semi-TCP方法提高19.52%,证明了文中方法的有效性。

Clinical Profile of Juvenile Idiopathic Arthritis (JIA) in a Semi-Rural Ivorian Environment

Objectives: To describe the clinical profile of juvenile idiopathic arthritis (JIA) in a semi-rural Ivorian setting: the rheumatology department of the Bouaké University Hospital (CHU.B). Patients and methods: This was a descriptive cross-sectional study from January 2018 to December 2023 in the rheumatology department of the CHU.B. The study focused on the epidemiological, diagnostic and therapeutic data of children followed for JIA on the basis of the classification criteria of the International League of Associations for Rheumatology (ILAR). Excel 2019 software was used to analyse the data. Results: Over a 6-year period, JIA represented 0.29% (11/3780) of all rheumatological conditions and 1.40% (11/782) of childhood rheumatic diseases. The mean age of the children at diagnosis was 13.09 ± 3.01 years, and the mean age at onset of the disease was 8.54 years. There were 6 girls and 5 boys. Joint involvement was the main reason for consultation and hospitalisation, accounting for 81.81% (9) of cases. There was a family history in 3 children (27.3%). The clinical forms were as follows: systemic 27.27% (3), oligoarticular 27.27% (3), polyarticular with rheumatoid factor negative 27.27% (3), polyarticular with rheumatoid factor positive 9.09 (1), arthritis with enthesitis 9.09 (1). JIA was discovered at steinbrocker functional stages II and III. Treatment included NSAIDs 72.72% (8), corticosteroid therapy combined with methotrexate and hydroxychloroquine 54.54% (6) and rehabilitation 63.63% (7). Conclusion: JIA is a little-known disease in the semi-rural environment of Côte d’Ivoire. It is characterised by a long delay in diagnosis in relation to poverty and the lack of recourse to alternative medicine in our populations;hence, it is important to raise awareness of the condition among the population and practitioners. Multidisciplinary intervention would improve the diagnosis and management of children with JIA.

基于事件触发的奇异semi-Markov跳变系统的有限时间控制

为减少不必要的网络资源传输,解决实际复杂系统的有限时间问题,提出一种基于事件触发方案的奇异semi-Markov跳变系统的有限时间控制器设计方法。在事件触发方案下,用系统的时滞表示网络诱导时延,闭环系统可以建模为一个时滞系统。首先,通过构造合适的Lyapunov-Krasovskii泛函,以严格矩阵不等式的形式给出充分条件,以使事件触发方案下的闭环系统正则、无脉冲、随机有限时间有界;然后,通过求解矩阵不等式,得到状态反馈控制器的参数;最后,给出一个数值示例,通过仿真证明所提方法的有效性。

Estimating the Drought-Induced Yield Loss for Winter Wheat in a Semi-Arid Region of the Southern United States Using a Drought Index

The economy of most rural locations in the semi-arid region of Llano Estacado in the southern United States is predominantly based on agriculture, primarily beef and wheat (Triticum aestivum L.) production. This region is prone to drought and is projected to experience a drier climate. Droughts that coincide with the critical phenological phases of a crop can be remarkably costly. Although drought cannot be prevented, its losses can be minimized through mitigation measures if it is predicted in advance. Predicting yield loss from an imminent drought is an important need of stakeholders. One way to fulfill this need is using an agricultural drought index, such as the Agricultural Reference Index for Drought (ARID). Being plant physiology-based, ARID can represent drought-yield relationships accurately. This study developed an ARID-based yield model for predicting the drought-induced yield loss for winter wheat in this region by accounting for its phenological phase-specific sensitivity to water stress. The reasonable values of the drought sensitivity coefficients of the yield model indicated that it could reflect the phenomenon of water stress decreasing the winter wheat yields in this region reasonably. The values of the various metrics used to evaluate the model, including Willmott Index (0.86), Nash-Sutcliffe Index (0.61), and percentage error (26), indicated that the yield model performed fairly well at predicting the drought-induced yield loss for winter wheat. The yield model may be useful for predicting the drought-induced yield loss for winter wheat in the study region and scheduling irrigation allocation based on phenological phase-specific drought sensitivity.

宁南黄土区中庄小流域土壤密度坡面变化及尺度效应

[目的]分析土壤密度在坡面上的变化特征及尺度效应,探讨土壤密度坡面尺度转换路径,为精准监测土壤密度、科学评价土壤功能、指导土壤资源管理提供理论依据。[方法]在宁夏南部半干旱黄土丘陵区中庄小流域,选择3个退耕还林还草后的典型坡面,按坡位从上至下设置了5个样地,调查各样地的植被类型、土地利用、立地条件等,采用环刀法分层测定了200 cm深度的土壤密度。[结果]在东坡、西坡、南坡,土壤密度随土层加深基本表现为线性增加,土层每加深10 cm土壤密度平均增加0.0062,0.0036,0.0075 g/cm^(3),其原因为植被根系、土壤有机质、压实作用的土层差异。在0—200 cm土层,东坡和南坡土壤密度坡面均值均为1.22 g/cm^(3),低于西坡(1.26 g/cm^(3))。3个坡面土壤密度的坡面变化存在差异,东坡、西坡、南坡沿坡变化表现为逐渐增加或先增后减、先升后降、先降后升,整体上为坡面下部大于上部,主要原因包括植被类型、土地利用、人类活动、恢复年限等坡面差异。以离坡顶水平距离增加100 m土壤密度坡面滑动平均值的变化量评价坡面尺度效应,0—200 cm土层的平均土壤密度坡面尺度效应在东坡为0.032(g/cm^(3))/100 m,在西坡为0.024(g/cm^(3))/100 m(离坡顶0—250 m)和-0.007(g/cm^(3))/100 m(离坡顶250—329 m),在南部为-0.011(g/cm^(3))/100 m(离坡顶0—140 m)和0.018(g/cm^(3))/100 m(离坡顶140—368 m)。构建3个坡面不同坡位样地各土层的土壤密度与坡面均值的比值随离坡顶水平距离增加而变化的关系式(R2>0.80,p<0.0001),且基于拟合关系式预测的土壤密度坡面均值的相对误差均小于2.5%,藉此可由任一坡位样地土壤密度测定值准确便捷地估算坡面均值。[结论]半干旱黄土区土壤密度坡面变化和坡位差异明显,且植被类型、人为干扰、土地利用等对土壤密度坡面分布格局影响较大,未来应加强主要作用因子对土壤密度坡面变化及尺度效应的贡献研究。

Decentralized Semi-Supervised Learning for Stochastic Configuration Networks Based on the Mean Teacher Method

The aim of this paper is to broaden the application of Stochastic Configuration Network (SCN) in the semi-supervised domain by utilizing common unlabeled data in daily life. It can enhance the classification accuracy of decentralized SCN algorithms while effectively protecting user privacy. To this end, we propose a decentralized semi-supervised learning algorithm for SCN, called DMT-SCN, which introduces teacher and student models by combining the idea of consistency regularization to improve the response speed of model iterations. In order to reduce the possible negative impact of unsupervised data on the model, we purposely change the way of adding noise to the unlabeled data. Simulation results show that the algorithm can effectively utilize unlabeled data to improve the classification accuracy of SCN training and is robust under different ground simulation environments.

基于MVBCN-FLW的中文法律文书命名实体识别

中文法律文书命名实体识别是智慧司法的基础性任务。目前的中文法律文书命名实体识别研究中已经取得一些成果,但其中大部分方法依赖于已标注的法律语料而未有效利用未标注的法律语料,且不能深入获取法律文书的特征。针对上述问题,提出一种中文法律文书命名实体识别框架。该框架使用基于双向编码器的转换器模型来学习中文法律文书的向量表示,并使用能够融合法律术语特征的双向长短时记忆网络语言模型来捕捉法律文书序列的上下文特征向量。该框架将中文法律文书的向量表示和上下文特征向量进行融合,融合后的特征向量被输入到一个由双向门控循环单元、自注意力机制和条件随机场组成的模块中进行训练。此外,为了使框架在缺少已标注的法律语料时也能得到更加充分的训练,使用未标注的法律语料进行自训练,生成新标注的法律语料并将其与初始标注的法律语料合并,通过进行迭代训练来提升框架性能。实验结果表明,该框架优于其他基于主流神经网络的命名实体识别模型。

基于EMO-GAN的恶意URL检测框架

随着万维网的广泛应用和网络威胁的日益严峻,统一资源定位符(uniform resource locator,URL)的安全性成为了网络安全领域的研究热点,如何有效检测并防范恶意URL已经成为了业内非常关注的问题。针对恶意URL检测中存在的数据获取困难、特征表示不足以及模型概念漂移挑战,提出了一种基于EMO-GAN的恶意URL检测框架(EMO-GAN-based malicious URL detection framework,EMO-GANUDF)。该框架通过结合极度随机树(extremely randomized trees,ET)和边缘生成对抗网络(margin generative adversarial network,MarginGAN)进行半监督学习,有效解决了数据获取困难问题。在特征提取上,该框架提出了一种综合统计、字符和词汇特征的特征表示方法,实现了URL的高效特征表示。此外,为了应对模型概念漂移问题,该框架提出了一种支持在线学习(online learning)的分类器,增强了模型拓展性和适应性。在多个数据集和不同检测方法上进行对比实验,所提方法在Malicious URLs公开数据集上达到了99%的准确率和84%的F 1分数,较其他检测方法取得了更好的效果,证明了其有效性及优越性。

彭波半细毛羊冷冻保存睾丸组织支持细胞分离、纯化及鉴定

[目的]睾丸支持细胞(Sertoli cells, SCs)位于生精小管内上皮,为不同分化阶段雄性生殖细胞发育提供物理构架和能量物质支持,对精子发生过程产生至关重要的作用。试验旨在研究新鲜绵羊睾丸组织样品获取不便的情况下,从冷冻保存的彭波半细毛羊睾丸组织分离支持细胞的技术方法。[方法]将冷冻保存液预处理后冻存的彭波半细毛羊睾丸组织进行复苏,采用苏木精-伊红(HE)染色检测复苏后睾丸组织形态,利用组合酶消化法、差速贴壁法进行睾丸组织支持细胞分离纯化培养,观察细胞生长规律,绘制生长曲线,采用RT-PCR和免疫荧光染色(IF)技术进行睾丸支持细胞特异性基因鉴定。[结果]冻存复苏睾丸组织曲细精索及睾丸间质保存完好。组合酶消化后能获得满足试验需要的生精上皮细胞混悬液,分离培养2~4 h后支持细胞贴壁,形态呈梭形或不规则多边形,培养3~4 d后汇合,其中培养1~2 d生长速度缓慢,培养3~6 d进入对数期,增殖速度加快,培养7~8 d后增殖速度下降。RT-PCR检测显示,GNDF、WT1、ABP、SOX9基因均在彭波半细毛羊睾丸支持细胞中特异性表达;免疫荧光染色显示,特异性抗体GATA4、Vimentin免疫阳性。[结论]经冻存液预处理的冷冻保存睾丸组织复苏后,可分离出满足试验条件的原代支持细胞,应用于科学试验研究。

Solution of the Matrix Second Semi-Tensor Product Equation A ∘ l X ∘ l B=C

In this paper, the solution of the matrix second semi-tensor product equation A∘lX∘lB=Cis studied. Firstly, the solvability of the matrix-vector second semi-tensor product equation is investigated. At the same time, the compatibility conditions, the sufficient and necessary conditions and the specific solution methods for the matrix solution are given. Secondly, we further consider the solvability of the second semi-tensor product equation of the matrix. For each part, several examples are given to illustrate the validity of the results.

Pseudo-Semi-Overlap Functions-Based Fuzzy Rough Sets Applied to Image Edge Extraction

As an extension of overlap functions, pseudo-semi-overlap functions are a crucial class of aggregation functions. Therefore, (I, PSO)-fuzzy rough sets are introduced, utilizing pseudo-semi-overlap functions, and further extended for applications in image edge extraction. Firstly, a new clustering function, the pseudo-semi-overlap function, is introduced by eliminating the symmetry and right continuity present in the overlap function. The relaxed nature of this function enhances its applicability in image edge extraction. Secondly, the definitions of (I, PSO)-fuzzy rough sets are provided, using (I, PSO)-fuzzy rough sets, a pair of new fuzzy mathematical morphological operators (IPSOFMM operators) is proposed. Finally, by combining the fuzzy C-means algorithm and IPSOFMM operators, a novel image edge extraction algorithm (FCM-IPSO algorithm) is proposed and implemented. Compared to existing algorithms, the FCM-IPSO algorithm exhibits more image edges and a 73.81% decrease in the noise introduction rate. The outstanding performance of (I, PSO)-fuzzy rough sets in image edge extraction demonstrates their practical application value.

基于对抗学习和一致性正则的半监督语义分割方法

为了降低语义分割任务对像素级标签的需求,提出一种基于对抗学习和Mean teachers模型的半监督语义分割方法。该方法训练过程分为两个阶段,第一阶段在分割网络之后连接判别网络,通过对抗学习使分割网络预测结果逐渐接近真实标签;第二阶利用第一阶段的网络参数做指数移动平均得到教师网络,与分割网络做一致性训练,使模型性能进一步提升。使用PASCAL VOC 2012数据集进行实验,结果表明在使用相同数量的标签训练下,该方法的分割图的质量和评价指标mIoU优于现有半监督语义分割方法。

面向类不平衡和重叠的工控数据异常检测的半监督欠采样方法

工业控制系统异常检测面临着数据缺乏标签信息、类不平衡和类重叠的耦合问题,导致现有的分类器难以精准检测异常数据。现有的数据级采样方法在打伪标签、数据平衡或检测重叠区域时存在着打伪标签结果不准确、采样效果稳定性差以及重叠识别率低等问题。为此,提出一种基于半监督学习的欠采样方法(SSLU-LP)。该方法通过异构集成将标签传播机制和单类分类器结合,补充数据伪标签;利用最小生成树策略构建重叠区域检测模型;采用欠采样策略,通过最近邻搜索有选择性地去除部分多数类样本。最后该方法与四种经典分类器结合,在九个工控数据集上与九种混合算法进行比较。实验结果表明,所提方法可以精准地为无标签数据打伪标签,高效且有效检测出不平衡数据集中的重叠数据,改善了分类器的训练效果,提高了分类器的异常检测性能。

基于群论的频率图在旅行商问题中的应用

针对最小生成树(minimum spanning tree,MST)和旅行商问题(travelling salesman problem,TSP),介绍了完全图上的两类特殊图并定义了这些图上的交运算,每类特殊图和交运算构成一个半群。根据半群性质计算出频率图,分析了最优哈密顿圈(optimal Hamiltonian cycle,OHC)和MST中边的频率性质,证明了频率图上OHC中边的频率下界,该频率下界用于缩小OHC的搜索空间,降低了TSP的求解难度。此外,采用一些TSP算例验证了频率图上OHC中边的频率性质。

基于CNN和Transformer交叉教学的半监督医学图像分割

由于医学图像分割领域缺乏高质量的标注数据,半监督学习方法在医学图像语义分割任务中受到高度重视.为了充分利用卷积神经网络(Convolutional Neural Network,CNN)和Transformer在半监督学习中的优势,本文提出一种基于CNN与Transformer交叉教学的半监督医学图像分割方法.该方法将经典的深度协同训练从一致性正则化简化为交叉教学,利用循环伪标签方案使两个网络的预测差异转换为无监督损失,以鼓励两个网络具有一致的低熵预测.所提方法在ISIC 2018数据集上进行实验,在采用20%的标注比例时,Dice系数和Jaccard系数分别达到87.25%和79.17%,相比于监督U-Net++的训练结果分别提升了2.89%和3.53%,并且优于目前主流的半监督学习方法,验证了所提方法在半监督医学图像分割上的有效性和泛化性.

基于潜在有价值样本挖掘的半监督三维目标检测

为了解决当前半监督三维目标检测算法中,传统的固定阈值方法在过滤伪标签时不够灵活,舍弃了大量有价值的伪标签,没有充分利用潜在有价值样本的问题,提出了一种基于潜在有价值样本挖掘的半监督三维目标检测方法。首先,不再使用固定阈值过滤伪标签,采用基于得分聚类的自适应阈值生成方法,分别为不同的类别生成过滤伪标签时需要的阈值,保留更多有价值的伪标签;其次,由于标签由类别和边界框信息组成,提出了一种联合置信度过滤伪标签的方法,使用对象置信度、分类置信度和IoU置信度的乘积来过滤伪标签,改善伪标签的质量;最后,对样本数量较少的类别生成稠密伪标签,筛选未通过联合置信度过滤的部分数据,以软伪标签的形式保留伪标签,更充分地利用潜在有价值的样本。在KITTI数据集上,与PV-RCNN方法相比,所提方法在仅1%标记数据的情况下,汽车类提高了6.5百分点,行人类提高了9百分点,自行车类提高了25百分点,实验结果证明了所提方法的有效性。

B/N掺杂单栅GFET的自洽模型研究

建立了B/N掺杂单栅石墨烯场效应管(GFET)全工作区域下的漏极电流模型。利用自洽方法获取精确的电势-电荷关系,解决了建模过程中的多重对数问题。该模型明确捕捉到产生非零带隙和狄拉克(Dirac)点偏移对载流子片电荷密度和迁移率等参数的影响。此外,通过考虑相互作用参数和杂质浓度建立了一种半经典扩散迁移率模型。所提出的漏极电流模型和扩散迁移率模型的模拟数据与7.5%B-掺杂单栅GFET的实验数据具有良好的一致性,验证了方法的有效性。在不同掺杂浓度下,B/N掺杂单栅GFET漏极电流模型所预测的转移特性曲线与输出特性曲线表明,通过掺杂能够完全抑制单栅GFET的双极性行为,增强了输出饱和性,并提高了开关比。因此,B/N掺杂单栅GFET能够同时满足模拟/射频电路和数字电路的应用条件。