基于多尺度的贝叶斯模型显著性检测

针对传统基于贝叶斯模型的显著性检测算法存在准确率不理想的问题,提出了一种基于多尺度的贝叶斯模型显著性检测算法。通过超像素分割算法(SLIC)将原图分割成不同尺度的超像素,根据超像素边界信息得到背景种子,进而通过距离计算和多尺度融合得到背景先验;对原图进行颜色增强,采用Harris算子对增强图进行检测角点求得凸包,融合不同尺度下的超像素得到凸包先验;融合背景先验和凸包先验得到最终先验;利用颜色直方图和凸包计算似然概率;将最终先验和似然概率通过贝叶斯模型计算显著图。在公开数据集MSRA1000、ECSSD上与多种传统算法进行准确率和召回率对比,该算法有更好的表现。     

基于主动学习的离群点集成挖掘方法研究

离群点检测任务通常缺少可用的标注数据，且离群数据只占整个数据集的很小一部分，相较于其他的数据挖掘任务，离群点检测的难度较大，尚没有单一的算法适合于所有的场景。因此，结合多样性模型集成和主动学习思想，提出了一种基于主动学习的离群点集成检测方法OMAL（Outlier Mining based on Active Learning）。在主动学习框架指导下，根据各种基学习器的对比分析，选择了基于统计的、基于相似性的、基于子空间划分的三个无监督模型作为基学习器。将各基学习器评判的处于离群和正常边界的数据整合后呈现给人类专家进行标注，以最大化人类专家反馈的信息量；从标注的数据集和各基学习器投票产生的数据集中抽样，基于GBM（Gradient BoostingMachine）训练一个有监督二元分类模型，并将该模型应用于全数据集，得出最终的挖掘结果。实验表明，提出方法的AUC有了较为明显的提升，且具有良好的运行效率，具备较好的实用价值。     

面向连续查询的敏感语义位置隐私保护方案

针对连续查询位置服务中构造匿名区域未考虑语义位置信息导致敏感隐私泄露问题，通过设计[(K,θ)]-隐私模型，提出一种路网环境下面向连续查询的敏感语义位置隐私保护方案。该方案利用Voronoi图将城市路网预先划分为独立的Voronoi单元，依据用户的移动路径和移动速度，选择具有相似特性的其他[K-1]个用户，构建匿名用户集；利用匿名用户集用户设定的敏感语义位置类型和语义安全阈值，以及用户所处语义位置的Voronoi单元，构建满足[(K,θ)]-隐私模型的语义安全匿名区域，可以同时防止连续查询追踪攻击和语义推断攻击。实验结果表明，与SCPA算法相比，该方案在隐私保护程度上提升约15%，系统开销上降低约20%。     

基于条件信息卷积生成对抗网络的图像识别

传统的图像识别方法需要大量有标签样本进行训练，且模型训练难以达到稳定。针对这些问题，结合条件生成网络和信息最大化生成网络的结构优势建立了条件信息卷积生成网络（C-Info-DCGAN）。模型增加图像的类别信息和潜在信息作为输入数据，然后利用Q网络去更好地发挥类别信息和潜在信息对训练的引导作用，并且利用深度卷积网络来加强对图像特征的提取能力。实验结果表明，该方法能够加快模型训练收敛速度，并有效提高图像识别的准确率。     

结合边缘特征先验引导的深度卷积显著性检测

针对当前基于深度学习的显著性检测算法缺少利用先验特征和边缘信息，且在复杂场景中难以检测出鲁棒性强的显著性区域的问题，提出了一种结合边缘特征，利用先验信息引导的全卷积神经网络显著性检测算法。该算法利用三种被经常用到的先验知识结合边缘信息形成先验图，通过注意力机制将提取的先验特征与深度特征有效融合，最终通过提出的循环卷积反馈优化策略迭代地学习改进显著性区域，从而产生更可靠的最终显著图预测。经过实验定性定量分析，对比证明了算法的可靠性。     

结合预训练模型和语言知识库的文本匹配方法

针对文本匹配任务,该文提出一种大规模预训练模型融合外部语言知识库的方法。该方法在大规模预训练模型的基础上,通过生成基于WordNet的同义—反义词汇知识学习任务和词组—搭配知识学习任务引入外部语言学知识。进而,与MT-DNN多任务学习模型进行联合训练,以进一步提高模型性能。最后利用文本匹配标注数据进行微调。在MRPC和QQP两个公开数据集的实验结果显示,该方法可以在大规模预训练模型和微调的框架基础上,通过引入外部语言知识进行联合训练有效提升文本匹配性能。     

基于语言特征自动获取的反问句识别方法

反问句是以疑问的形式表达强烈情感的修辞方式,对其有效识别可为自然语言处理中的情感分析任务提供技术支持。该文提出了一种基于语言特征自动获取的反问句识别方法。首先,利用标签注意机制,建立了一个数据驱动的特征抽取模型,用于获取与任务相关的词汇、句法结构、符号标记和话题等语言特征。其次,利用Bi-LSTM模型分别对句子和语言特征进行表示,两者的交互注意被用于获取句子的各个词和符号的注意力权重向量。该权重向量作用于句子的表示,用于构建一个强化语言特征的反问句识别模型。在中文微博数据集上的实验结果表明,提出的方法与之前的工作相比,反问句识别性能有显著提升。     

基于CDIO工程教育的应用型人才培养 理论与实践研究 — —以红外多路遥控系统设计为例

The construction of specialty groups in higher vocational colleges is in full swing. However, due to the cognitive bias and the lack of construction paradigm for reference, the construction of specialty groups is still in the period of "crossing the river by feeling the stones" with different opinions. There are some problems in the construction of specialty groups, such as unclear under- standing of the concept of specialty groups, unreasonable logic of groups, unclear construction of groups, and nonstandard manage- ment of groups. This paper describes the specific content of the construction of modern information technology specialty group based on the Sydney Agreement, hoping to form a reference paradigm for the construction of specialty group of similar colleges and universities.     

Smart work packaging-enabled constraint-free path re-planning for tower crane in prefabricated products assembly process

Lack of constraint-free crane path planning is one of the critical concerns in the dynamic on-site assembly process of prefabrication housing production (PHP). For decades, researchers and practitioners have endeavored to improve both the efficiency and safety of crane path planning from either static environment or re-planning the path when colliding with constraints or periodically updating the path in the dynamic environment. However, there is a lack of approach related to the in-depth exploration of the nature of dynamic constraints so as to assist the crane operators in making adaptive path re-planning decisions by categorizing and prioritizing constraints. To address this issue, this study develops the smart work packaging (SWP)-enabled constraints optimization service. This service embraces the core characteristics of SWP, including adaptivity, sociability, and autonomy to achieve autonomous initial path planning, networked constraints classification, and adaptive decisions on path re-planning. This service is simulated and verified in the BIM environment, and it is found that SWP-enabled constraints optimization service can generate the constraint-free path when it is necessary.     

Transfer learning between crop types for semantic segmentation of crops versus weeds in precision agriculture

Agricultural robots rely on semantic segmentation for distinguishing between crops and weeds to perform selective treatments and increase yield and crop health while reducing the amount of chemicals used. Deep‐learning approaches have recently achieved both excellent classification performance and real‐time execution. However, these techniques also rely on a large amount of training data, requiring a substantial labeling effort, both of which are scarce in precision agriculture. Additional design efforts are required to achieve commercially viable performance levels under varying environmental conditions and crop growth stages. In this paper, we explore the role of knowledge transfer between deep‐learning‐based classifiers for different crop types, with the goal of reducing the retraining time and labeling efforts required for a new crop. We examine the classification performance on three datasets with different crop types and containing a variety of weeds and compare the performance and retraining efforts required when using data labeled at pixel level with partially labeled data obtained through a less time‐consuming procedure of annotating the segmentation output. We show that transfer learning between different crop types is possible and reduces training times for up to 80%. Furthermore, we show that even when the data used for retraining are imperfectly annotated, the classification performance is within 2% of that of networks trained with laboriously annotated pixel‐precision data.