NOTES ON REFINABLE FUNCTIONS

1.IntroductionItiswell--knownthatrefinablefunctionsplayanimportantroleinthestudyingofwavelet.Usually,onehopesthatrefinablefunctionshavesomeparticularpropertiessuchassmoothnessandintegrability.Inthisnote,thezerosofanintegrablerefinablefunctionareobtained.Inparticalarbyexamplesoneshowsthatthelinearspaceassociatingthetranslatesoverthelatticepointsofarefinablefunctioncouldincludepolynomialspaceofdegreehigherthanitssmoothorder.LetsbeapositiveintegerandletR'(resp.C')bethes-dimensionalreal(qgmplex)…     

厚铜板阻焊起泡原因分析及改善

目前厚铜印制板的应用越来越多,使其在阻焊印刷过程中工艺参数管控与板面清洁工作就显得尤为重要.本文主要分析厚铜板油墨起泡的原因及探究改善措施;分析线路毛边长度、油墨粘度、阻焊印刷参数、静电对油墨起泡的影响后,通过新增二流体补充蚀刻、管控油墨粘度、静置加吸真空、调整印刷室湿度,针对性改善油墨起泡.     

基于Mask R-CNN的ORB去误匹配方法

为了提高多目标图像的ORB匹配的正确率,提出一种基于Mask R-CNN的图像ORB去除误匹配方法,该算法首先通过Faster R-CNN方法对图像进行识别,运用区域推荐网络得到矩形框标注的感兴趣区域和类别标签,该步骤可以得到感兴趣区域的预测类别和坐标信息,并且通过全卷积网络卷积层进行像素级别校正,得到像素级别的目标所属类别,然后进行目标分割。最后在原有ORB特征点匹配基础上,剔除两幅图像中相同目标分割区域以外的误匹配点。为了验证该方法的有效性,对传统ORB匹配与基于本文方法的ORB匹配进行了仿真实验。改进后的算法,使得在多目标环境下的目标的匹配精度提高了约18.6%,结果表明,本文算法较传统的ORB匹配算法的精度有一定提高。     

运用泽尼克多项式进行物面波前数据拟合

本文提出了一种运用泽尼克(Zernike)多项式的波前数据拟合方法.给出了离散点上正交多项式的构造并描述了具体计算步骤.可对待测物面低调制度点、散斑点、噪音及灰尘点区域进行波前数据拟合,通过一定阈值的设定,将误差点用掩模滤除并予以拟合.进一步提高了物体测量精度.     

Experimental investigation on: Laser shock micro-forming process using the mask and flexible pad

A forming process called the mask and flexible pad laser shock forming was proposed to fabricate the micro-features on the copper foil. In this process, the mask and laser beam were used as rigid punches. Shock waves induced by plasma were used as the source of loading and plasticine was used as a flexible pad. This was a micro scale and high strain rate forming process and the traditional forming method with micro-mold was changed. In the experiment, surface morphology of formed parts was represented and it was found that the mask played a significant role in the forming process. In order to understand the forming process in the experiment, process parameters, including laser pulse energy, numbers of laser pulse and grain size, were analyzed. The experimental results showed that different parameters had different effects on formed parts. The surface quality and the thickness distribution of formed parts were investigated. It was found that formed parts could keep good surface quality after laser shocking and the reasons were explored. The thickness distribution was measured and the thickness thinning rate was calculated. There was no local tightening or rupture in the forming area. In this paper, the micro-features could be obtained on metallic foils and the method of mold-free was proved to be feasible.     

结合区分性训练深度神经网络的歌声与伴奏分离方法

针对音乐信号中的歌声与伴奏相互关联难以分离的问题,提出了一种区分性训练深度神经网络(Deep Neural Network,DNN)的音乐分离方法。首先,在DNN模型的基础上同时考虑歌声与伴奏间的重建误差和区分性信息,提出了一种改进的目标函数进行区分性训练;其次,在DNN模型上额外添加一层,引入时频掩蔽对估计出的歌声伴奏进行联合优化,相应的时域信号由傅里叶逆变换获得;最后,验证不同参数设置对分离性能的影响,并与现有的音乐分离方法进行对比.实验结果表明,改进的目标函数和时频掩蔽的引入明显提高了DNN的分离性能,且与现有的音乐分离方法相比分离性能最高提高了4 dB从而证实所提方法是一种有效的音乐分离方法。     

基于改进Mask R-CNN的模糊图像实例分割的研究

Mask R-CNN是现阶段实例分割相对成熟的方法,针对Mask R-CNN算法当中还存在的分割边界精度以及对于模糊图片鲁棒性较差等问题,该文提出一种基于改进的Mask R-CNN实例分割方法。该方法首先提出在Mask分支上使用卷积化条件随机场(ConvCRF)来优化Mask分支对于候选区域进一步分割,并使用FCN-ConvCRF分支来代替原有分支;之后提出新锚点大小和IOU标准,使得RPN候选框能够涵盖所有实例区域;最后使用一种添加部分经过转换网络转换的数据进行训练的方法。总的mAP值与原算法相比提升了3%,并且分割边界精确度和鲁棒性都有一定提高。     

圆锥磁绝缘传输线的横向空间电荷流

根据电磁场基本理论及电子运动守恒方程，导出圆锥传输线横向空间电荷流的数值模型和磁绝缘临界条件。通过数值计算，讨论了电压及圆锥几何结构参数等对横向空间电荷流和磁绝缘性能的影响。电压较高时,无磁场的空间电荷流较大，而磁绝缘性能更好。在传输线的三个几何参数中，几何因子对传输性能影响最大。     

一种新型的底部对准光学系统设计

介绍了一种新型的底部对准系统,倍率的变化不是靠换不同物镜,而是靠变换辅助物镜焦距来变换倍率。x、y向工作台带动整支光路运动,不会给光路带来光轴偏离误差,z向调焦则通过前后移动物镜观察成像。     

掩模光刻机中找平控制研究

基片与掩模版(或找平版)找平技术是掩模光刻机中的一个关键技术,找平的结果直接决定了光刻机的分辨率精度。找平的目的就是要实现掩模版平面和基片平面的平行,实现基片与掩模版在移动到设定的一个微小间隙后,基片上的各点到掩模版的距离相同,从而保证基片上图形的一致性,提高设备的分辨率。找平过程的控制主要在对找平的判断、压力采集反馈、找平压力的控制等方面。