We use the improved quantitative rescattering (QRS) model to simulate the experimentally measured correlated two-electron momentum distributions (CMDs) for nonsequential double ionization (NSDI) of Ne exposed to intense elliptically polarized laser pulses with a central wavelength of 788 nm at a peak intensity of 5 x1014 W cm-2 for the ellipticities ranging from 0 to 0.25 (Kang et al 2018 Phys. Rev. A 97 063403). Only the CMDs for recollision direct ionization are calculated and the contribution from recollision excitation with subsequent ionization is ignored. With the assumption that recollisions occur after the zero crossing of the electric field along the major polarization, the QRS model successfully reproduces the experimentally observed asymmetry pattern of electron pairs in the CMDs for NSDI of Ne under elliptical polarization. By reducing the peak intensity claimed in experiment by about 30%, good agreement is achieved between the present results and the experimental measurements.

文本检测领域的非极大值抑制(NMS)算法延迟过大,影响整体实时性,为此提出一种后处理加速方法.采用基于位置范围的候选框分类方法代替提前排序,减少计算复杂度;通过多次缩放优化交并比计算公式,补充完全覆盖的约束条件,减少由缩放导致的小尺寸候选框冗余的问题;将判断条件复用并设计三级流水线的计算单元,进一步减少计算延迟.实验结果表明,加速器在Zynq-XC7Z020上部署的功耗为3.28 W,相比CPU实现的NMS和LANMS,性能提高了67倍和38倍.

Object detection in remote sensing images is a challenging task in the field of computer vision. Existing remote sensing image detection methods ignore the speed with the aim of improving the accuracy; however, it is also essential to increase the detection speed in real-time detection scenes, such as in resource surveys and maritime rescue. This paper proposes a lightweight target detection network to realize a trade-off between detection accuracy and speed. The design replaces the original backbone network of YOLOv4 with the pruned MobileNetV2. In addition, the ordinary convolution calculation of the feature extraction method is replaced by deep separable convolution to considerably reduce the computational complexity of the model. Finally, the receptive field enhancement module and attention mechanism module are embedded to improve the detection accuracy of the model. The experimental results on the images in the dataset containing the remote sensing images show that the accuracy, as measured by the mean average precision, is 89. 80%; further, the model detection speed is 33. 4 frame/s. Compared with YOLOv4, the accuracy only decreases by 1. 48 percentage points, but the detection speed increases by nearly 1. 5 times. Compared with the YOLOv4-Tiny algorithm, the average accuracy is 9. 05 percentage points higher. The proposed model successfully meets the trade-off requirements of speed and accuracy. The weight of the model is only 44 MB, which makes it easy to deploy and indicates that it meets the requirements of real-time detection scenarios.

We simulate the two-dimensional high-energy photoelectron momentum distributions (PMDs) for high-order above threshold ionization (HATI) of hydrogen atoms in intense elliptically polarized laser fields by using the strong-field approximation and solving the time-dependent Schrodinger equation. It is found that the centers of the left-and right-side circular rescattered ridges in the PMDs due to elastic collision of the returning electron with the parent ion, that locate on the polarization axis for linearly polarized laser field, move slightly below and above the major polarization axis, respectively, for elliptically polarized laser fields. We suggest that, by analyzing the center shift of the measured PMDs for HATI in elliptically polarized fields, one can identify the recollision time.

如今面向合成孔径雷达(synthetic aperture radar,SAR)图像的目标检测技术已得到广泛的研究与应用,并通过深度学习的方法可有效应用于复杂背景特征下的舰船检测,检测精度和速度也具有一定的效果.为提高在SAR舰船目标场景中的检测速度和精度,提出了一种可用于SAR舰船图像的单类目标快速检测方法,主要通过应用深度可分离卷积修改YOLOv4-Tiny网络,并在此基础上添加经改进的通道和空间注意力机制来提升精度.以输入图像尺寸为608×608为基准,相比于YOLOv4-Tiny模型,实验结果表明,该方法的检测精度可达93.91%,检测速度可达42.8 frame/s,提高了11.7 frame/s,模型权重仅有2.17 MB大小,约为YOLOv4-Tiny的1/10.模型利于硬件上板部署,满足SAR图像舰船目标的快速检测场景需求.

We show through simulation that the improved quantitative rescattering model (QRS) can successfully predict the nonsequential double ionization (NSDI) process by intense elliptically polarized laser pulses. Using the QRS model, we calculate the correlated two-electron and ion momentum distributions of NSDI in Ne exposed to intense elliptically polarized laser pulses with a wavelength of 788 nm at a peak intensity of 5.0 x 10(14) W/cm(2). We analyze the asymmetry in the doubly charged ion momentum spectra observed by Kang et al. [Phys. Rev. Lett. 120, 223204 (2018)] in going from linearly to elliptically polarized laser pulses. Our model reproduces the experimental data well. Furthermore, we find that the ellipticity-dependent asymmetry arises from the drift velocity along the minor axis of the elliptic polarization. We explain how the correlated electron momentum distributions along the minor axis provide access to the subcycle dynamics of recollision. From these findings, we expect that we can extend the QRS model for NSDI toward more complicated laser fields in the future.

Speech emotion recognition is essential in human-computer interaction.In this study, a Chinese speech emotion recognition method based on the Trumpt-6 convolutional neural network model was developed to solve the problem of low efficiency and accuracy of Chinese speech emotion recognition.During the process of extracting the Mel Frequency Cepstral Coefficient (MFCC) feature, the pixel size of the MFCC feature map was reduced to improve the processing efficiency of single recognition.This was achieved by increasing the number of sampling points in the frame windowing operation, adding the features in each Hamming window, and reducing the number of Hamming windows.Gaussian white noise was used to enhance the data set to minimize overfitting during the training process.The experimental results for the CASIA speech emotion data set show that the test accuracy of this method is 95.7%, which is better than those of traditional methods, such as Lenet-5, Recurrent Neural Network(RNN), and Long Short-Term Memory(LSTM).The Trump-6 convolutional neural network model uses 2 048 sampling points and only 176 550 parameters for training.This method has fewer parameters, a simpler structure, and faster processing than ResNet34 and the cyclic neural network model using deep convolutional neural networks.

Within the framework of the improved quantitative rescattering (QRS) model, we simulate the correlated two-electron momentum distributions (CMDs) for nonsequential double ionization (NSDI) of Ar by elliptically polarized laser pulses with a wavelength of 788 nm at an intensity of 0.7 × 1014 W/cm2 for the ellipticities ranging from 0 to 0.3. Only the CMDs for recollision excitation with subsequent ionization (RESI) are calculated and the contribution from recollision direct ionization is neglected. According to the QRS model, the CMD for RESI can be factorized as a product of the parallel momentum distribution (PMD) for the first released electron after recollision and the PMD for the second electron ionized from an excited state of the parent ion. The PMD for the first electron is obtained from the laser-free differential cross sections for electron impact excitation of Ar+ calculated using state-of-the-art many-electron R-matrix theory while that for the second electron is evaluated by solving the time-dependent Schrödinger equation. The results show that the CMDs for all the ellipticities considered here exhibit distinct anticorrelated back-to-back emission of the electrons along the major polarization direction, and the anticorrelation is more pronounced with increasing ellipticity. It is found that anticorrelation is attributed to the pattern of the PMD for the second electron ionized from the excited state that, in turn, is caused by the delayed recollision time with respect to the instant of the external field crossing. Our work shows that both the ionization potential of the excited parent ion and the laser intensity play important roles in the process.

针对大多数基于FPGA的加速器受限于运算资源与访存带宽,很难部署大型的神经网络这一问题,通过研究神经网络的计算特点,运用通用计算模型的思想,提出了一种指令集架构的神经网络协处理器的设计方案.该方案中的协处理器具有一套专用的运算指令集,支持多种神经网络结构的运算,搭配相应的指令生成程序,能够灵活快速的进行神经网络在FPGA平台上的部署.考虑神经网络运算的相似性,对电路进行复用,降低资源的占用;设计内存多端口读写控制模块,通过协调片上缓存与片外存储,降低对访存带宽的需求.使用python与UVM验证方法学搭建验证平台进行验证,并在PYNQ-Z2开发板上进行实验.结果表明:方案中的协处理器对激活函数的运算误差在0.05以下,对其他类型指令的运算误差在10-4级别,工作在100MHz下的运算性能达到41.73GOPS,达到同类设计的主流水平,消耗的资源比同类设计平均降低80％.

Using an improved quantitative rescattering model, we calculate the correlated two-electron momentum distributions (CMDs) for nonsequential double ionization of Ar exposed to intense laser pulses with a wavelength of 790 nm at a peak intensity of 1.0 x 10(14) W/cm(2). We analyze the drastic variations in the CMDs that were observed by Kubel et al. [New 3. Phys. 16, 033008 (2014)] in the transition from near-single-cycle to multicycle driving laser pulses. Our model reproduces their experimental data well. We also find that the transition from near-single-cycle to multicycle driving laser pulses depends strongly on the details of the pulse envelope. Special attention is paid to the mechanisms responsible for the cross-shaped structure observed experimentally with 4 fs pulses. Our analysis reveals that the cross-shaped structure in the carrier-envelope phase-averaged CMD for near-single-cycle pulses can be attributed to strong backward scattering of the recolliding electron as well as the narrow momentum distributions of the tunnel-ionized electrons compared to those for long pulses. This also explains why the cross-shaped distributions collapse to a rather structureless distribution when the pulse duration is increased to 8 fs.

Recently, the quantitative rescattering model (QRS) for nonsequential double ionization (NSDI) is modified by taking into account the potential change (PC) due to the presence of electric field at the time of recollision. Using the improved QRS model, we simulate the longitudinal momentum distributions of doubly charged ions He2+ by projecting the correlated two-electron momentum distributions for NSDI of He onto the main diagonal. The obtained results are compared directly with the experimental data at different intensities. It is found that when the PC is considered, the width of momentum distributions reduces and the agreement between theory and experiment is improved.

程序设计课程以教为主,创新实践课以实操为主.文章以"搭建智能小车和控制系统"为创新实践需求,充分利用课上和课下资源,促使本科生将程序设计、机械自动化、通信等多学科知识点进行融合,强化基础课程和实现培养独立创新能力;将程序设计课程通过软件开发形式与硬件相结合,从问题中寻找解决方案,从应用解决方案中凝聚问题,提高学生的学习主动性和积极性.文章介绍了以软硬件协同开发为向导的本科创新实践,探讨计算机基础课程的创新人才培养经验.

We calculate the correlated two-electron momentum distributions for nonsequential double ionization of helium in a 400-nm laser pulse with a peak intensity of 6.5 x 10(14) W/cm(2), which is below the recollision threshold, by using the quantitative rescattering model in which the lowering of the threshold due to the presence of the electric field at the instant of recollision is taken into account. While distinct correlated back-to-back emission of the electrons along the polarization direction is predicted in accord with other existing theoretical simulations, we suggest an alternative mechanism that is responsible for the anticorrelation. At intensities below the recollision threshold, recollision excitation can only take place when the barrier of the Coulomb potential is sufficiently suppressed by the electric field. The excited electron begins to ionize at a "delayed" recollision time after a crossing of the field and hence the probability of being tunnel-ionized after the subsequent peak of the field is increased. It is demonstrated that the "delayed" recollision time predominantly determines the parallel momentum distribution of the tunneling electron and plays a decisive role in forming anticorrelation.

单向点对点传输线是目前应用最广泛也是最基本的传输线类型.传统单一建模分析方法已不能完全满足对高速信号的分析.对传输线中高速信号反射现象结合集总模型与分布式模型进行建模,分析了线长、过孔对信号反射的影响;对解决信号反射的阻抗匹配方案进行建模仿真,分析匹配电阻、分支线长、过孔以及端接电压对信号反射的影响,通过理论分析与仿真验证证明设计方案的可行性与实用性.所提出的分析设计方案已应用到多个产品的设计中.

由于传统列驱动芯片在采用串行方式逐行逐列进行数据传输时存在传输效率低的问题,本文设计一种基于控制串行传输的列驱动芯片.该芯片结合串行传输以及并行传输方式的特点,在JXI5020列驱动芯片的基础上,增加控制数据移位寄存器以及控制数据处理器将用于子空间寻址的控制信号串行化传输,以串行方式实现基于并行传输的子空间扫描;该芯片打破了数据在串行方式下逐行逐列传输的限制,降低时间冗余率,提高了数据传输的效率.通过搭建LED扫描驱动系统及其仿真验证平台,验证了该驱动芯片的正确性.在显示256级灰度、分辨率为960* 720的图像时,与JXI5020进行对比,得出本文设计的驱动芯片能够在没有亮度损失的条件下,将冗余率降低至5.9％,有效减少了单元面板信号接口的数量.

无论是使用传统的方法进行人脸识别,还是使用神经网络进行人脸识别,都存在运算量大、运算时间长等问题,很难对视频中的人脸进行实时检测与匹配.针对上述问题,使用轻量化神经网络进行人脸检测,使用运算简单的哈希算法计算人脸图像相似度,并对多个哈希相似度值加权进行人脸匹配,是减少运算时间、实现快速人脸识别的可行方案.使用轻量化神经网络Mobilenet作为人脸特征提取网络,使用剪枝的SSD模型作为检测网络,通过级联Mobilenet与SSD实现人脸的检测,之后对检测到的人脸图像进行识别.首先,分别计算人脸图像的均值哈希相似度与感知哈希相似度.然后,分别使用α和β 作为均值哈希与感知哈希的加权系数对图像的均值哈希与感知哈希相似度值进行加权,并将结果作为图像的最终相似度.当加权后的相似度值大于设定的阈值I时,则认为两张图像中的人脸是同一个人;当加权后的相似度值小于设定的阈值K时,则认为两张图像中的人脸是不同的人.对于相似度处于阈值I和阈值K之间的图像,将它们按照相似度值从高到低的顺序择优匹配.所提方法在WiderFace和FDDB上的人脸检测准确率分别达到92.5％和94.2％,每张图片的平均处理时间为56 ms;在ORL标准人脸库进行人脸匹配的准确率达到96.2％.使用摄像头进行实时人脸识别测试时,所提方法的人脸识别准确率为95％,平均人脸识别速度为80 ms.实验证明,所提方法在保证较高准确率的前提下,能够实现实时的人脸检测与匹配.

Using the improved quantitative rescattering (QRS) model, we simulate the correlated two-electron momentum distributions (CMD) for nonsequential double ionization (NSDI) of Ar by near-single-cycle laser pulses with a wavelength of 750 nm at an intensity of 2.8 x 10(14) W/cm(2). With the accurate cross sections obtained from fully quantum mechanical calculations for both electron impact excitation and electron impact ionization of Ar+, we unambiguously identify the contributions from recollision direct ionization (RDI) and recollision excitation with subsequent ionization (RESI). Our analysis reveals that RESI constitutes the main contribution to NSDI of Ar under the conditions considered here. The simulated results are directly compared with experimental measurements [Bergues et al., Nature Commun. 3, 813 (2012)] in which each NSDI event is tagged with the carrier-envelope phase (CEP). It is found that the overall pattern of both the CEP-resolved and the CEP-averaged CMDs measured in experiment are well reproduced by the QRS model, and the cross-shaped structure in the CEP-averaged CMD is attributed to the strong forward scattering of the recolliding electron as well as the depletion effect in tunneling ionization of the electron from an excited state of the parent ion. (C) 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

Using the quantitative rescattering model, we simulate the correlated two-electron momentum distributions for nonsequential double ionization of helium by 800 nm laser pulses at intensities in the range of (2 - 15) × 1014 W/cm2. The experimentally observed V-shaped structure at high intensities [A. Rudenko et al., Phys. Rev. Lett. 99, 263003 (2007)] is attributed to the strong forward scattering in laser-induced recollision excitation and the asymmetric momentum distribution of electrons that are tunneling-ionized from the excited states. The final-state electron repulsion also plays an important role in forming the V-shaped structure.

Action recognition is an important area in computer vision. An action recognition system based on SoC FPGA and CNN model is presented. A pipelined LK optical flow module and a pedestrian detection module based on HOG and SVM are realized. The results of detection of dynamic pedestrian and static pedestrian are merged to recognize the area where the person is. The image and optical flow in the area are both input into the CNN model, to recognize what action it is. An ISA-NPU is proposed to realize the computation of the CNN model. The system is realized in Altera DE10-Nano Kit. With the co-operation of hardware and software, the system can recognize actions such as stand, walk, wave and squat. The system is accurate in recognition and flexible for expansion and modification of the CNN model.

We present numerical simulations on the correlated two-electron momentum distributions for the recollisional ($e, 2e$) process in nonsequential double ionization (NSDI) of helium by an 800-nm laser field at an intensity around $4.5\ifmmode\times\else\texttimes\fi{}{10}^{14}\phantom{\rule{0.16em}{0ex}}\mathrm{W}/{\mathrm{cm}}^{2}$, based on the improved quantitative rescattering model, in which the lowering of the kinetic energy required for the electrons to escape from the parent ion due to the presence of electric field at the time of recollision has been taken into account. According to the QRS model, the correlated two-electron momentum distributions for laser-induced ($e, 2e$) collision in NSDI can be factorized as a product of the returning-electron wave packet (RWP) and the field-free differential cross section (DCS) for ionization of the parent ion by the impact of the laser-induced returning electron. The RWPs which describe the momentum distribution of the returning electrons are obtained within the strong-field approximation for high-order above threshold ionization. In the calculations of the DCSs for electron impact ionization of ${\mathrm{He}}^{+}$, the precollision Coulomb interaction between the incident electron and the parent ion and the postcollision Coulomb interaction between the two outgoing electrons are considered. In addition, the dynamic screening (DS) of the three-body Coulomb interactions in the final state, due to the fact that the strength of the interaction of any two particles is affected by the presence of the third one, has also been taken into account. It has been found that while the postcollision Coulomb interaction is responsible for the observed fingerlike structure, the precollision Coulomb interaction changes the orientation of the two fingers from $\mathsf{V}$-type to parallel, the final state DS reduces the separation of the two fingers, and the lowering of threshold energy shifts the two fingers towards smaller momenta. All these effects improve the agreement of the simulated results with the experimental results.