基于开放式知识表示的智能化产品设计

基于知识的智能化产品设计是CAD技术的发展方向之一,但现有知识表示方式缺乏对产品设计人员的开放性,不利于设计人员理解和维护设计系统中的知识。针对这一不足,研究机械产品设计领域常见的公式、表格、二维映射图、过程、规则等类型知识的开放式表示方法,给出了其BNF(Backus Naur Form)范式描述,并研究了各类知识的推理方法。所研究的开放式知识表示及相应推理方法具有以下优点:1)知识不再固化于设计系统;2)知识便于设计人员自行录入和维护。最后以基于知识的航空发动机涡轮叶片气膜孔设计为例,验证了该方法应用于工程实践的可行性。     

Interception Algorithm of S-cubed Signal Model in Stealth Radar Equipment

Radar equipment of stealth platforms such as aircraft have adopted the newest modern technology to design the signal waveforms. One of the important and effective methods is the hybrid waveform called spread spectrum stretch (S-cubed) which combines linear frequency modulation (LFM) and discrete phase code. In order to investigate the function of enemy’s stealth radar equipment, the interception algorithm of S-cubed is needed. In this paper, a novel detection and parameter estimation approach for the reconnaissance S-cubed radar signal is presented. First, the generalized time-frequency representation of Zhao, Atlas, and Marks (ZAM-GTFR) and Hough transforms (HT) are applied to detecting the signal, and then the initial frequency and modulation slope of LFM are estimated from the ZAM-GTFR. On the basis of LFM information, the reconstructing signal is generated. Finally, the code rate of discrete phase code is extracted from the negative peaks of the ZAM-GTFR. Simulation results show that the proposed algorithm has higher estimation accuracy when the signal to noise ratio (SNR) is above 3 dB.     

基于迁移学习的SAR目标超分辨重建

针对合成孔径雷达(SAR)目标超分辨重建问题,提出了一种基于迁移学习的超分辨方法。在光学图像梯度域中联合训练超完备字典与稀疏编码映射,利用半耦合字典联系SAR图像与光学图像,寻找SAR图像在半耦合字典下的稀疏编码,并在高分辨率字典下完成重建。结合SAR图像的先验信息,使用正则化方法对SAR目标进行特征增强。所提方法在TerraSAR-X数据和MSTAR数据上进行了仿真实验,重建结果表明,相比目前的插值方法和稀疏表示方法,所提方法空间分辨率可提高0.5~1.5个像素。正则化增强结果表明,引入稀疏先验的正则化增强能够进一步提高空间分辨率并抑制杂波比,最后分析了正则化参数的选取对图像质量的影响。     

导弹垂直装填中图像采集系统设计

在分析了某型舰空导弹垂直装填过程中“抗摇摆”系统需求的基础上,提出了基于 CCD成像的导弹贮运发射箱空间位置图像采集系统结构,给出了完整的电路设计。     

一种模糊决策专家系统的模型

目前决策型专家系统主要适合于解决精确的或不确定性决策问题,还存在模糊决策问题,如最佳方案选择问题。统一考虑决策问题中的不确定性和模糊性,提出了一种能统一处理带有不确定性和模糊性决策问题的单目标决策型专家系统模型。针对构造决策专家系统所涉及的具体问题,本文从知识表示、知识获取、模糊谓词计算、规则匹配、系统求解方式等几个方面介绍所提出的模型。     

局部空间-时间显著性的表达过程蕴涵对空间显著性的表达

建立的基于独立成分分析进行特征提取器的学习,进而模拟人类对视觉信息的局部空间-时间显著性的表达过程的计算模型,蕴涵了对局部空间显著性的有效表达;从而当据其设计为机器人主动视觉的注意机制的视线控制而用的算法,可以同一个处理过程适用于动态场景或静态场景的情形。对获自动态场景的特征提取器的频率特性的分析,揭示了此种内在的包含性的成因;算法的实验结果显示了局部空间-时间显著性表达过程对静态场景的局部空间显著性可有效表达。     

基于对偶四元数的编队飞行卫星相对位姿描述及算法研究

针对编队飞行微小卫星,突破轨道姿态分而治之的传统模式,利用对偶四元数,提出了主从星间的相对位置和姿态统一描述方法,建立了对偶四元数卫星编队运动学模型。该方法区别于描述刚体纯旋转变换的四元数方法,将主从星的本体系之间坐标变换的旋转和平移过程进行融合、统一,并具有非常简洁的形式。同时,给出了求解模型的算法,解算出主从星间的相对位置和姿态。仿真结果表明该模型及算法科学合理,能够满足编队飞行的测控要求。     

Young Children's Perception of Diagrammatic Representations

Diagrams and pictorial representations are common in children's lives and require abstraction away from visual perception. In three experiments, we investigated 4- to 8-year-olds’ comprehension of such representations. In Experiment 1 (N = 80), children were shown photographs of geometric objects and asked to choose the corresponding line drawing from among sets of four, or vice versa. Results showed considerable developmental progression, especially around age 6. Experiment 2 (N = 16) ruled out that 4-year-olds’ difficulties were due to problems with the visual matching task itself. Experiment 3 (N = 32) showed comparable performance for matching diagrams to 3D objects rather than to photographs. Findings suggest increasing understanding of diagrammatic representations around the time of school entry.     

飞机框肋类零件基础特征自动识别与提取算法

飞机框肋类零件是组成飞机骨架的重要零件,具有数量大、形状各异等特点,其生产制造所耗费的时间在飞机研制过程中占有较大比重。然而,通过现有CAD软件所提供的功能进行相关制造操作,无论是效率、质量等均已不能满足现代飞机设计和制造要求,围绕飞机框肋类零件研究和开发相关的自动化制造系统已迫在眉睫。基于框肋类零件边界表示模型对零件基础特征进行自动识别与提取,是实现后续相关工艺规划与加工的基础与前提。针对该问题,提出零件基础特征模型,并建立一种基于同侧面的特征识别算法,即:以零件STEP数据作为输入,选取两侧腹板面,应用属性邻接图（AAG）构建、有效邻面识别、关联面完整识别等方法,逐级识别各级关联面以构建两侧同侧面,通过同侧面单元匹配最终实现基础特征构造和特征邻接图构建。其中,针对零件三维模型中的碎面缺陷提出其定义与识别方法,以保证特征面识别的完整性。经由实例测试,验证所提算法的可行性与有效性。      

Trajectory Estimation with Multi-range-rate System Based on Sparse Representation and Spline Model Optimization

The classic state methods for trajectory estimation in boost phase with multi-range-rate system include method of point-by-point manner and that of spline-model-based manner. Both are deficient in terms of model-approximation accuracy and systematic error determination thus resulting in the estimation errors well beyond the requirements, especially, concerning the maneuvering trajectory. This article proposes a new high-precision estimation approach based on the residual error analysis. The residual error comprises three components, i. e. systematic error, model truncation error and random error. The approach realizes self-adaptive estimation of systematic errors in measurements following the theory of sparse representation of signals to minimize the low-frequency components of residual errors. By taking median- and high-frequency components as indexes, the spline model-approximation is improved by optimizing node sequence of the spline function and the weight selection for data fusion through iteration. Simulation has validated the performances of the proposed method.