改进OTSU算法和边缘检测的图像分割算法研究

最大类间方差法是一种最广泛使用的图像阈值自动选取方法之一,在分析理解该算法的基础上,提出了一种改进最大类间方差法和边缘检测相结合的分割方法,该算法在选取最佳阈值时不仅考虑目标和背景的均值还考虑了它们的方差。通过实验分析表明该算法能够达到较好的分割效果,有较好的使用价值。     

基于1.55μm激光雷达的晴空风切变结构研究北大核心CSCD

晴空风切变是公认的航空危险天气。为了研究155μm激光测风雷达对晴空风切变的探测效果,利用激光测风雷达数据,结合自观系统实时风场数据,对2022年2月13日影响银川机场的晴空风切变过程进行详细分析。激光雷达探测结果表明风切变发生前高空偏北风有下沉特征,动量下传导致地面北风增强;此次风切变由东北风与西南风交汇形成,东北大风呈楔形从近地层嵌入,形成水平风切变,随后向西南传播影响机场跑道。预报员利用激光测风雷达可以提前15min对此次风切变进行预警。     

电子设备电源系统的谐波特性研究

电子设备的电源系统或电力系统中,存在多种类型的谐波源,分为谐波电流源和谐波电压源,本文针对典型电源系统开关电源系统进行分类和建模,给出各类谐波源的特征,并对容性滤波整流电路的谐波特性进行了分析.     

IBM集群式高性能计算机体系结构及气象应用

本文详细描述了南京信息工程大学引进的IBM集群式高性能计算机的体系结构,并结合了WRF气象模式具体应用,分析了高性能计算机体系结构与气象领域之间的关系。     

Highly Ordered Mesoporous Silicon Carbide Ceramics with Large Surface Areas and High Stability

Highly ordered mesoporous silicon carbide ceramics have been successfully synthesized with yields higher than 75 % via a one‐step nanocasting process using commercial polycarbosilane (PCS) as a precursor and mesoporous silica as hard templates. Mesoporous SiC nanowires in two‐dimensional (2D) hexagonal arrays (p6m) can be easily replicated from a mesoporous silica SBA‐15 template. Small‐angle X‐ray diffraction (XRD) patterns and transmission electron microscopy (TEM) images show that the SiC nanowires have long‐range regularity over large areas because of the interwire pillar connections. A three‐dimensional (3D) bicontinuous cubic mesoporous SiC structure (Ia3d) can be fabricated using mesoporous silica KIT‐6 as the mother template. The structure shows higher thermal stability than the 2D hexagonal mesoporous SiC, mostly because of the 3D network connections. The major constituent of the products is SiC, with 12 % excess carbon and 14 % oxygen measured by elemental analysis. The obtained mesoporous SiC ceramics are amorphous below 1200 °C and are mainly composed of randomly oriented β‐SiC crystallites after treatment at 1400 °C. N₂‐sorption isotherms reveal that these ordered mesoporous SiC ceramics have high Brunauer–Emmett–Teller (BET) specific surface areas (up to 720 m² g^–1), large pore volumes (～ 0.8 cm³ g^–1), and narrow pore‐size distributions (mean values of 2.0–3.7 nm), even upon calcination at temperatures as high as 1400 °C. The rough surface and high order of the nanowire arrays result from the strong interconnections of the SiC products and are the main reasons for such high surface areas. XRD, N₂‐sorption, and TEM measurements show that the mesoporous SiC ceramics have ultrahigh stability even after re‐treatment at 1400 °C under a N₂ atmosphere. Compared with 2D hexagonal SiC nanowire arrays, 3D cubic mesoporous SiC shows superior thermal stability, as well as higher surface areas (590 m² g^–1) and larger pore volumes (～ 0.71 cm³ g^–1).     

Design techniques of low-power embedded EEPROM for passive RFID tag

This paper presents an optimized embedded EEPROM design approach which has reduced the power significantly in a short-range passive RFID tag. The proposed array control circuit employs an improved structure to minimize the leakage of memory bit cells. With the proposed array circuit design, the passive RFID tag can operate drawing a low quiescent current. The RFID tag with the proposed EEPROM was fabricated in a standard 0.35-μm four-metal two-poly CMOS process. Measurement results show that the erasing/writing current is 45 μA, and reading current consumption is 3 μA with a supply voltage of 3.3 V. The data read time is 300 ns/bit.     

A learning scheme for low-speed precision tracking control of hybrid stepping motors

Servo control of the hybrid stepping motor is complicated due to its highly nonlinear torque-current-position characteristics, especially under low operating speeds. This paper presents a simple and efficient control algorithm for the high-precision tracking control of hybrid stepping motors. The principles of learning control have been exploited to minimize the motor's torque ripple, which is periodic and nonlinear in the system states, with specific emphasis on low-speed situations. The proposed algorithm utilizes a fixed proportional-derivative (PD) feedback controller to stabilize the transient dynamics of the servomotor and the feedforward learning controller to compensate for the effect of the torque ripple and other disturbances for improved tracking accuracy. The stability and convergence performance of the learning control scheme is presented. It has been found that all error signals in the learning control system are bounded and the motion trajectory converges to the desired value asymptotically. The experimental results demonstrated the effectiveness and performance of the proposed algorithm.     

南海海边整层大气相干长度的测量与分析

大气相干长度是描述光波大气传输中湍流效应的重要参量,具有区域变化特性。海边大气相干长度的测量结果很少有报道,利用差分像运动法大气相干长度仪观测了茂名博贺海边的整层大气相干长度,获得了海边大气相干长度的全天数据。数据分析表明,海边整层大气相干长度日变化明显,夜晚起伏无规则,上半夜平均天文视宁度好于下半夜,白天最差;与陆地数据比较得知,海边整层大气相干长度日变化不如陆地上的剧烈,海边日变化成近"V"型,陆地上的为近"U"型;通过分析比较,提出了一些启发性的见解。     

莱斯信道下分布式大规模MIMO系统基站选择算法的研究

该文考虑一种分布式大规模MIMO系统,假设基站端与用户之间的信道为莱斯信道,研究了该系统中基站选择的算法。首先给出了系统采用匹配滤波和迫零预编码时,用户下行可达速率的闭式表达式,并分析了系统的功率效率性能。然后基于此闭式表达式,以最大化系统的频谱效率为目标,提出了基于增量选择和基于用户优先级的基站选择算法。这两种算法只需要系统获取基站端与用户之间的信道统计特征信息,从而有效降低了系统开销。仿真结果表明,所提出的两种基站选择算法性能仍能逼近最优算法。特别地,当采用匹配滤波预编码且基站端天线数趋于无穷时,基于用户优先级的基站选择算法优于基于增量选择的算法。     

视觉显著性检测：一种融合长期和短期特征的信息论算法

针对传统视觉显著性检测算法单纯使用当前观测图像的信息或是先验知识的不足,该文引入了长期特征和短期特征的概念,分别代表先验知识和当前观测图像的信息,并提出了一种基于信息论的算法将它们融合。首先,分别根据人眼跟踪数据和当前观测图像的内容来训练长期和短期稀疏词典并对图像进行稀疏编码,将得到的稀疏编码作为长期和短期特征。其次,针对现有算法只能在整幅图像上或是在一个固定大小的局部邻域内进行统计的缺陷,该文提出一种基于信息熵的特征概率分布估计方法,该方法可以根据当前观测图像的具体情况自适应地选择一个最佳的区域大小来计算长期和短期特征出现的概率。最后,利用香农自信息来输出图像的显著性检测结果。同8种流行算法在公开的人眼跟踪测试库上进行的主观和定量的实验对比证明了该文算法的有效性。