基于子波分解的多通道神经网络纹理分割方法

描述了一种体现多通道滤波技术的神经网络纹理分割方法，决策神经网络（ＤＢＮ）可提高纹理分类的精度，同时纹理的子波变换降低了图像数据间的相关性，提高了网络的学习效率，实验表明本文提出孤方法分类误差较低，获得了令人满意的纹理分割效果。     

Structural, electrical and optical properties of zinc oxide produced by oxidation of zinc thin films

We have investigated the effects of oxidation temperature on the physical properties of polycrystalline zinc oxide thin films. Zinc thin films are oxidized at different temperatures in air. We have found that increasing the oxidation temperature deteriorates the preferred c-axis orientation. Also, increasing the oxidation temperature enlarges the crystal size and increases the number of needle-shaped crystals on the surface of the ZnO samples. By increasing the oxidation temperature, more than zinc melting point, tensile stresses start to build up in the films. Also by increasing temperature, sheet resistance of the films decreases, while photoluminescence intensity ratio (green to orange) increases. Increasing the oxidation temperature reduces the transparency of the films, too. It is proposed that either an increase in the number of oxygen vacancies or a decrease in the volume of grain boundaries, is responsible for the observed behavior of the films at higher oxidation temperatures.     

Grain nucleation and texture analysis of electroless copper deposition on a palladium seed layer

Electroless copper grains were deposited on a Pd seed layer under varying bath conditions. The seed layer was determined to have a (111) texture using grazing incident x-ray (GIX) diffraction. Multiple nucleation sites in the grain boundaries were imaged using a scanning tunneling microscope. Continual copper growth produced row-like structures. The texture of the electrolessly deposited copper (ED-Cu) grains were determined to be (111). No radial grain orientation for the Pd seed layer or the ED-Cu thin film was detected using GIX diffraction. Atomic force microscope images indicated continual Cu nucleation throughout the deposition process. PdH was formed as a by-product of the electroless deposition process, and detected by x-ray diffraction.     

Kappa-Carrageenan, Sodium Chloride and Temperature Affect Yield and Texture of Structured Beef Rolls

Structured lean beef rolls (4–5% fat), formulated with 33% added water, 1% sodium chloride and 0.35% sodium tripolyphosphate and cooked to 63°, 73° and 83°, had low cook yields and poor texture and bind. Kappa-carrageen an (KC) added at 0.5–1% and NaCl at 2–3% increased cook yield and improved textural properties (bind, force to fracture, hardness). Rolls with 1.0% KC and 3% NaCl had the highest cook yield (154% meat weight basis) and the highest values for force to fracture and hardness. Cook yield decreased and hardness increased with increased cooking temperature. Effects of KC on yield and texture were most pronounced at the lowest NaCl level (1%) and the highest temperature (83°). KC also reduced purge of vacuum-packaged slices during refrigerated storage.     

大型厚板半球形工件热拉深模的设计

通过分析比较以往封头压模的设计参数与现有封头模具的差异 ,提出新的冲压件的压制工艺和模具设计方案。经过生产验证 ,使用效果很好 ,可作为球形封头热拉模的设计依据     

Production technology of large-area multicrystalline silicon solar cells

In 1996 a conversion efficiency of 17.1% had been obtained on 15 cm×15 cm mc-Si solar cell. In this paper, large-scale production technology of the high-efficiency processing will be discussed. Enlarging reactive ion etching (RIE) equipment size, technology of passivation, and fine contact grid with low resistance by screenprinted metallization, which is firing through PECVD SiN, have been investigated.     

基于CTS 的大地景仿真的参数优化设计

为解决大地景仿真时显示逼真度与画面流畅度相互制约的问题,对基于地形生成系统(creator terrain system,CTS)的大地景仿真进行参数优化设计.结合某型飞机飞行仿真系统研制实例,分析地形纹理和地形模型的生成原理,给出其参数设置,重点研究重要参数对纹理和地形生成的影响.仿真结果表明,该设计能有效解决地景逼真度与流畅度互相制约的难点.     

基于灰度共生矩阵和神经网络的Si3N4陶瓷推挤加工表面纹理分析

基于边缘破碎效应驱动裂纹软推挤加工是一项新颖的加工技术。通过采集Si₃N₄陶瓷的软推挤加工表面形貌,运用灰度共生矩阵(GLCM)分析了对比度、熵、相关性3个特征参数与加工表面纹理分布的内在关系。通过径向基网络和竞争层网络两类神经网络的分工协作,对不同加工参数下已加工表面的纹理特征进行预测和分类,其预测结果的相对误差能控制在5%之内。随着对比度和熵越大,相关性越小;分类等级越大,表面平整程度越差。通过系统实验探讨了各加工参数对纹理特征的影响,可靠地评估了加工质量的优劣。随着车刀进给速度或槽深的增大,加工表面质量变差;随着凸缘厚度的增大,加工表面质量先逐渐变差,但经过凸缘厚度2.5 mm分界点后却又有所改善。     

基于积谱矩阵局部二值模式的欺骗干扰识别

针对跟踪雷达的三种常见欺骗干扰,提出了一种基于积谱矩阵( SPM)局部二值模式( LBP)的识别算法。首先,计算雷达接收信号不同脉冲重复周期( PRI )的积谱,并将其按照频域和慢时域排列成一个二维积谱矩阵。然后,将图像处理的纹理特征引入到欺骗干扰识别的算法中,采用局部二值模式提取二维积谱矩阵的灰度图像特征。最后,通过仿真实验验证了算法的正确性,当雷达接收信号的信噪比( SNR)大于5 dB时,该欺骗干扰识别算法的平均识别概率优于90%。     

特征结合和相关反馈技术在医学图像检索中的应用

本文在单一检索的基础上,对基于灰度直方图的颜色特征,基于小波变换的纹理特征和基于不变矩的形状特征进行融合.为了使用户能够参与检索过程,又引入了相关反馈机制,通过调整权值使得检索的结果最终满足用户的检索要求.最后分别给出基于单一特征,特征融合和相关反馈方法的查准率和查全率,并对试验结果进行分析.