计算机视觉教程

本示例主要实现利用Excel操作Access数据库功能，包含以下功能：一、向数据库中新添数据；二、修改数据库中数据；三、删除数据库中数据；四、高级查询数据库中数据；(包含精确、模糊查询，可选择显示字段)五、批量操作数据库；六、采用在Excel表的导航栏中增加管理菜单的方式来实现操作功能；使用方法：1、Info.xls文件与Info.mdb文件在同一文件夹下；2、打开Info.xls文件后选择“启用宏”，如果没有该提示，请查看Excel->工具->宏->安全性  里的设置；3、正确启用宏之后，在Excel表的顶部菜单上会增加一个“信息管理菜单”的新菜单，具体操作在该菜单中实现； 本示例由黄云原创，仅供大家学习研究使用，在编写VBA时黄云尽量采用注释的方法来描述各部分作用，以期大家能够看懂。转载请保留宏中的注释信息。

《DUILIB加载XML界面》对应源码，博客地址：http://blog.csdn.net/harvic880925/article/details/8920326，有兴趣的朋友可以去看看

基于matlab的纠错码源码实现。在编码端，根据BCH码的（n，k）产生生成矩阵G和校验矩阵H

【实例简介】雷达在下视情况下的地杂波的模拟和方针技术研究

MIPI Alliance Specification for D-PHY Version 1.00.00 – 14 May 2009配合“MIPI Alliance Specification for Camera Serial Interface 2 (CSI-2)“ 一起看。http://download.csdn.net/detail/micro_st/4242724Version1.00.0014-May-2009MIPI Alliance Specification for D-PHY2 The material contained herein is not a license, either expressly or impliedly, to any IPR owned or3 controlled by any of the authors or developers of this material or MIPl. The material contained herein is4 provided on an"as iS basis and to the maximum extent permitted by applicable law, this material is5 provided AS IS AND WITH ALL FAULTS, and the authors and developers of this material and MIP6 hereby disclaim all other warranties and conditions, either express, implied or statutory, including, but not7 limited to, any (ifany)inplied warranties, duties or conditions of merchantability, of fitness for a8 particular purpose, of accuracy or completeness of responses, of results, of workmanlike effort, of lack of9 viruses, and of lack of negligence10 All materials contained herein are protected by copyright laws, and may not be reproduced, republisheddistributed, transmitted, displayed, broadcast or otherwise exploited in any manner without the express12 prior written permission of MIPI Alliance. MIPl, MIPI Alliance and the dotted rainbow arch and all related3 trademarks, tradenames, and other intellectual property are the exclusive property of MIPI Alliance and14 cannot be used without its express prior written permission15 ALSO, THERE IS NO WARRANTY OF CONDITION OF TITLE, QUIET ENJOYMENT, QUIET16 POSSESSION. CORRESPONDENCE TO DESCRIPTION OR NON-INFRINGEMENT WITH17 REGARD TO THIS MATERIAL OR THE CONTENTS OF THIS DOCUMENT.IN NO EVENT WILLI8 ANY AUTHOR OR DEVELOPER OF THIS MATERIAL OR THE CONTENTS OF THIS DOCUMENT9 OR MIPI BE LIABLE TO ANY OTHER PARTY FOR THE COST OF PROCURING SUBSTITUTE20 GOODS OR SERVICES. LOST PROFITS. LOSS OF USE. LOSS OF DATA OR ANY INCIDENTAL.21 CONSEQUENTIAL, DIRECT, INDIRECT, OR SPECIAL DAMAGES WHETHER UNDER22 CONTRACT TORT WARRANTY OR OTHERWISE ARISING IN ANY WAY OUT OF THIS OR23 ANY OTHER AGREEMENT SPECIFICATION OR DOCUMENT RELATING TO THIS MATERIAL24 WHETHER OR NOT SUCH PARTY HAD ADVANCE NOTICE OF THE POSSIBILITY OF SUCH25 DAMAGES26 Without limiting the generality of this Disclaimer stated above, the user of the contents of this Document is27 further notified that MIPI: (a)does not evaluate, test or verify the accuracy, soundness or credibility of the28 contents of this Document;(b)does not monitor or enforce compliance with the contents of this Document29 and (c)does not certify, test, or in any manner investigate products or services or any claims of compliance30 with the contents of this Document. The use or implementation of the contents of this Document may31 involve or require the use of intellectual property rights ("IPR")including(but not limited to) patents32 patent applications, or copyrights owned by one or more parties, whether or not Members of MIPIMIPI33 does not make any search or investigation for IPR, nor does miPi require or request the disclosure of any34 IPR or claims of IPR as respects the contents of this document or otherwise35 Questions pertaining to this document, or the terms or conditions of its provision, should be addressed36 MIPI Alliance. Inc37 c/o IEEE-ISTO38 445 Hoes lane39 Piscataway, NJ0885440 Alin: Board SecretaryCopyright C 2007-2009 MIPl Alliance, Inc. All rights reservedMIPI Alliance Member ConfidentialVersion1.00.0014-May-2009MIPI Alliance Specification for D-PHY42 Contents43 Draft Version 1.00.00-14 May 2009441 Overview1451.2 Purpose.…..,.,.,,.,..472 Terminology…2.1 Definitions162.2 Abbreviations…172.3 Acronyms51 3 D-PHY Introduction523.1 Summary of Phy functionality533.2 Mandatory Functionality················2054 4 Architecture21554.1 Lane modules…564.2 Master and slave2254.3 High Frequency Clock Generation22584.4 Clock lane data lanes and the phy-Protocol interface.224.5 Selectable Lane Options·;····················234.6 Lane Module Types4.6.1 Unidirectional Data Lane…264.6.2 Bi-directional data lanes26634.6.3 Clock lane.274.7 Configurations….7654.7.1 Unidirectional Configurations............664.7.2Bi-Dal Half-Duplex Configurations674.7.3 Mixed Data Lane configurations32Copyright C 2007-2009 MIPl Alliance, Inc. All rights reservedMIPI Alliance Member Confidential111Ⅴ ersion1.00.0014-May-2009MIPI Alliance Specification for D-PHY695.1Transmission Data Structure,………………………∴335.1.1Data unitsa勹5.1.2 Bit order Serialization and De-Serialization33725.1.3 Encoding and decoding735.1.4 Data Buffering,33745.2 Lane States and Line levels755.3 Operating Modes: Control, High-Speed, and Escape5. 4 High-Speed Data Transmission··········;·5. 41 Burst payload data785.4.2 Start-of-Transmission795.4.3End-of-transmission805.4.4 HS Data Transmission burst.365.5 Bi-directional data Lane turnaround5.6 Escape Mode41835.6.1Remote triggers42845.6.2 Low-Power data Transmission43855.6.3 Ultra-Low Power State865.6.4 Escape Mode State Machine43875.7 High-Speed Clock Transmission885. 8 Clock lane Ultra-Low Power State50959 Global Operation Timing Parameters.……5.10 System Power States56915.11 Initialization56925.12 Calibration5.13 Global Operation Flow Diagram57945.14 Data Rate Dependent Parameters(informative)955. 14.1 Parameters Containing Only UI values965. 14.2 Parameters Containing Time and Ul values59Copyright C 2007-2009 MIPl Alliance, Inc. All rights reservedMIPI Alliance Member ConfidentialVersion1.00.0014-May-2009MIPI Alliance Specification for D-PHY5.14.3 Parameters Containing Only Time Values…………5.14.4 Parameters Containing Only Time Values That Are Not Data Rate Dependent6 Fault detection611006.1 Contention detection1016.2 Sequence Error Detection.……611026.2.1 SoT Error621036.2.2 SOT Sync Error1046.2.3 EoT Sync Error1056.2. 4 Escape Mode Entry Command error.1066.2.5 LP Transmission Sync error621076.2.6 False Control error1086.3 Protocol Watchdog Timers(informative)62l096.3.1 HS RX Timeout6.3.2HS TX Timeout………………·················+···:··:·················∴62l116.3.3Escape mode timeout62l126.3. 4 Escape Mode Silence Timeout6.3.5 Turnaround errors114 7 Interconnect and Lane Configuration.641157.1 Lane configuration1167.2 Boundary Conditions.....…647.3 Definitions………64l187.4S- parameter Specifications………….651197.5 Characterization Conditions207.6 nterconnect Specifications………1217.6.1 Differential characteristics1227. 6.2 Common-mode characteristics671237.6.3 Intra-Lane Cross-Coupling1247. 6. 4 Mode-Conversion limitsCopyright C 2007-2009 MIPl Alliance, Inc. All rights reservedMIPI Alliance Member ConfidentialVersion1.00.0014-May-2009MIPI Alliance Specification for D-PHY1257.6.5 Inter-Lane Cross-Coupling671267. 6.6 Inter-Lane static skew1277.7 Driver and receiver Characteristics1287.7.1 Differential Characteristics1297. 7.2 Common-Mode characteristics1307.7.3 Mode-Conversion Limits1317.7.4 Inter-Lane Matching132 8 Electrical Characterislics701338.1 Driver characteristics1348.1.1 High-Speed Transmitter1358.1.2 Low-Power Transmitter1368.2 Receiver Characteristic·…············…·······…8301378.2.1 High-Speed Receiver801388.2.2Low- Power receiver.................….….821398.3 Line contention detection1408.4 Input Characteristics8441 9 High-Speed Data-Clock Timing1429.1 High-Speed Clock Timing861439.2 Forward High-Speed Data Transmission Timing871449.2.1 Data-Clock Timing Specifications1459.3 Reverse High-Speed Data Transmission Timing89146 10 Regulatory Requirements91147 Annex A Logical PHY-Protocol Inter face Description(informative)92148A 1 Signal Description149A 2 High-Speed Transmit from the Master Side150A3 High-Speed receive at the slave Sidel00151A 4 High-Speed Transmit from the Slave side152A.5 High-Speed Receive at the Master SideIOICopyright C 2007-2009 MIPl Alliance, Inc. All rights reservedMIPI Alliance Member ConfidentialVersion1.00.0014-May-2009MIPI Alliance Specification for D-PHY153A6 Low-Power Data Transmission102154A7 Low-Power Data Reception.103155A 8 Turn-around156 Annex B Interconnect Design Guidelines (informative)105157B. 1 Practical distances105158B 2 RF Frequency Bands: Interference.105B3 Transmission Line design160B4 Reference Layer.106161B 5 Printed-Circuit board106162B6 Flex-foils106163B 7 Series resistance106164B 8 Connectors106165 Annex C 8b9b Line Coding for D-PHY(normative)107166C 1 Line Coding Features...·············108167C.1.1Enabled Features for the Protocol108l68C 1. 2 Enabled Features for the Phy108169C2 Coding scheme170C 2.1 8b9b Coding Properties.....108171C 2.2 Data Codes: Basic Code Set……….109C.2.3 Comma Codes: Unique Exception Codes110173C 2.4 Control Codes: Regular Exception Codes…10174C.2.5 Complete Coding Scheme………175C 3 Operation with the D-PhY…11117yload: Data and Control177C.3.2 Details for Hs transmission………112178C.3.3 Details for LP Transmissionl12179C 4 Error Signal180C5 Extended PplCopyright C 2007-2009 MIPl Alliance, Inc. All rights reservedMIPI Alliance Member ConfidentialⅤ ersion1.00.0014-May-2009MIPI Alliance Specification for D-PHYl81C.6 Complete Code Set.….….l15182Copyright C 2007-2009 MIPl Alliance, Inc. All rights reservedMIPI Alliance Member Confidentialv111Version1.00.0014-May-2009MIPI Alliance Specification for D-PHYl83Figures184 Figure 1 Universal Lane Module functions21185 Figure2 Two Data Lane PHY Configuration.…………23186 Figure 3 Option Selection Flow Graph4187 Figure 4 Universal Lane Module Architecture25188 Figure 5 Lane Symbol Macros and Symbols Legend189 Figure 6 All Possible Data Lane Types and a basic Unidirectional Clock lane190 Figure 7 Unidirectional Single Data Lane Configuration30191 Figure 8 Unidirectional Multiple Data Lane Configuration without LPDT∴.30192 Figure 9 Two Directions Using Two Independent Unidirectional PHYs without LPDT.........31193 Figure 10 Bidirectional Single Data Lane Configuration31194 Figure 1l Bi-directional Multiple Data Lane Configuration......32195 Figure 12 Mixed Type multiple data Lane Configuration32196 Figure 13 Line level34197 Figure 14 High-Speed Data Transmission in Bursts36198 Figure 15 TX and rX State Machines for High-Speed Data Transmission37Figure16 Turnaround Procedure.……39200 Figure 17 Turnaround State Machine40201 Figure 18 Trigger-Reset Command in Escape Mode202 Figure 19 Two Data Byte Low-Power Data Transmission Example203 Figure 20 Escape Mode State Machine204 Figure2 I Switching the Clock Lane between Clock Transmission and low- Power mode………….47205 Figure 22 High-Speed Clock Transmission State Machine49206 Figure 23 Clock Lane Ultra-Low Power State State Machine········+·+···+·4···207 Figure 24 Data Lane Module State Diagram57208 Figure 25 Clock Lane Module state diagram58209 Figure 26 Point-to-point InterconnectCopyright C 2007-2009 MIPl Alliance, Inc. All rights reservedMIPI Alliance Member Confidential

压缩包里有两个文件，一个是课程设计时调试过得可以直接用用的，另一个是我进行了一定修改，但由于手头没有板子，没有调试过得，但注释比较全，便于理解。具体的程序解释可以看我的博客。初来乍到，多多关照~

个人整理的CNN学习笔记，所有素材均来自于互联网。把经典的摘之以作备案，待整理好这份笔记后，发现对CNN也有了深入理解，希望对你也有帮助！LOCALLY CONNECTED NEURAL NETCONVOLUTIONAL NETSTATIONARI? Statisties issimilar at dif ferent locationsLearn multiple filters.Example: 1000x1000 image1M hidden unitsFilter size: 10x10E.g. 1000x1000 image100M parameters100 FiltersFilter size: 10x10u鴻網互聯EEIDEE下面的分析来源于:htp/ log csdn. net/zouxy09/ article/details,/8781543我们知道,隐含层的每一个神经元都连接10x10个图像区域,也就是说每一个神经元存在10x10=100个连接权值参数。那如果我们每个神经元这100个参教是相同的呢?也就是说每个神经元用的是同一个卷积核去卷积图像。这样我们就只有多少个参数??只有100个参数啊!!!亲!不管你隐层的神经元个数有多少,两层间的连接我只有100个参数啊!亲!这就是权值共享啊!亲!这就是卷积神经网络的主打卖点啊!亲!(有点烦了,呵呵)也许你会问,这样做靠谱吗?为什么可行呢?好了,你就会想,这样提取特征也忒不靠谱吧,这样你只提取了一种特征啊?对了,頁聪明,我们需要提取多种特征对不?假如一种滤波器,也就是一种卷积核就是提出图像的一种特征,例如某个方向的边缘。那么我们需要提取不同的特征,怎么办,加多几种滤波器不就行了吗?对∫。所以假设我们加到100种滤波器,每种滤波器的参数不一样,表示它提出输入图像的不同特征,例如不同的边缘。这样每种滤波器去卷积图像就得到对图像的不同特征的放映,我们称之为 Feature Map。所以100种卷积核就有100个 Feature Map。这100个 Feature Map就组成了一层神经元。到这个时候明了了吧。我们这一层有多少个参数了?100种卷积核x每种卷积核共享100个参数=100×100=10K,也就是1万个参数。才1万个参数啊!亲!(又来了,受不了了!)见上图右:不同的颜色表达不同的滤波器嘿哟,遗漏一个问题了。刚才说隐层的参数个数和隐层的神经元个数无关,只和滤波器的大小和滤波器种类的多少有关。那么隐层的神经元个数怎么确定呢?它和原图像,也就是输入的大小(神经元个数)、滤波器的大小和滤波器在图像中的滑动步长都有关!例如,我的图像是1000×1000像素,而滤波器大小是10×10,假设滤波器没有重叠,也就是步长为10,这样隐层的神经元个数就是(1000×1000y(10×10}=100×100个神经元了,假设步长是8,也就是卷积核会重叠两个像素,那么……我就不算了,思想懂了就好。注意了,这只是一种滤波器,也就是一个 Feature Map的神经元个数哦,如果100个 Feature Map就是100倍了。由此可见,图像越大,神经元个数和需要训练的权值参数个数的贫富差距就越大。CONV NETS: EXTENSIONSBypool ing"(e. g. max or average)filterresponses at different locations we gain Over the years. some new medes have proven to be veryrubustness to the exact spatial location effective when plugged inte corv-netsof featuresL2 PoolingU, kEN(x,r)Local Contrast normalizationhN(x y)iN(xvIRenato总之,卷积网络的核心思想是将:局部感受野、权值共享(或者权值复制)以及时间或空间亚采样这三种结构思想结合起来获得了某种程度的位移、尺度、形变不变性。23实例视觉皮层存在两类相关的细胞,S细胞( Simple Cel)和C( Complex Cell)细胞。S细胞在自身的感受野内最大限度地对图像中类似边缘模式的剌激做出响应,而C细胞具有更大的感受野,它可以对图像中产生刺激的模式的空间位置进行精准地定位C是卷积层,S是下采样层。参看这个两个网址,对理解CNN非常有帮助http://www.68idc.cn/help/buildlang/ask/20150705417730.htmlhttp://www.68idc.cn/help/buildlang/ask/20150705419299.html涵pC1s2c384江蘇鴻網互刷三 -TOLERc是卷积层,S是下采样层。输入的一幅图像,在C1层,通过和3个卷积模板做卷积运算,然后加上偏置值,再经过 sigmoid激活函数,得到3幅输出图像,在S2层,对C1层输出的3幅图像做下采样,假设采样因子是2,也就是图中每2*2的 patch中的4个像素进行求和,再加偏置,再通过激活函数,得到3张尺寸减小了的输出图像。同样的,再经过C3S4。将S4的输出拉成一个向量,输入传统的神经网络中,并得到输出24综合实例G1: feature maps8@28X28C3: f. mapInp ut20@10x1032×32S1: f. mapsS4: f. maps@14x1420@5x5C5:120Output: 9ConvolutionsSubsamplingSubsamplingconnectionConvolutions og. Csdn. n Convolutions 688图中的卷积网终Ⅰ作流程如下,输入层由32×32个感知节点组成,接收原始图像。然后,计算流程在卷积和子抽样之间交替进行,如下所述·第一隐藏层进行卷积,它由8个特征映射组成,每个特征映射由28×28个神经元组成,每个神经元指定一个5×5的接受域,这28×28个神经元共享5×5个权值参数,即卷积核第二隐藏层实现子抽样和局部平均,它同样由8个特征映射组成,但其每个特征映射由14×14个神经元组成。每个神经元具有一个2×2的接受域,一个可训练系数,一个可训练偏置和一个 sigmoid激活函数。可训练系数和偏置控制神经元的操作点;第三隐藏层进行第二次卷积,它由20个特征映射组成,每个特征映射由10×10个神经元组成。该隐藏层中的每个神经元可能具有和下一个隐藏层几个特征映射相连的突触连接,它以与第一个卷积层相似的方式操作。第四个隐藏层进行第二次子抽样和局部平均汁算。它由20个特征映射组成,但每个特征映射由5×5个神经元组成,它以与第一次抽样相似的方式操作。●第五个隐藏层实现卷积的最后阶段,它由120个神经元组成,每个神经元指定一个5×5的接受域。最后是个全连接层,得到输出向量。综合起来,如下图山辆入员/閩张38的圈像辆入|种特图到C28×28KD卷,:对应位遇相森再未和⊙如偏置每种图有个偏置a老积层f同治减函数6↑5的卷铝kC6个偏置值b输邶钟培钲图24X0mno:22域的值平均坐smle=2辆入6种征图列12|2k2个55的着核kKk22个偏置值b辆出12钟特钲图8×88×g年怦居D mean pooL斜样9=2s输出口钟特征图4x叶把S园輪出的特征图垣成一年向量入,维,:平4x×12=12F房入节点有2个192资接的神经刚络W1,09W参教W:20×192矩阵爹教bho0x1向量oh辆出节点有个,国为baby的度是o·上图中28如何变成24的?12如何变成8的?详情见这个网址:hto:/ud. stanfordedu/wki/ index. php/ Featureextractionusing convolution这个过程中,有这么几个参数:a.深度 depth:神经元个数,决定输出的dept厚度。同时代表滤波器个数。*b.步长 stride:决定滑动多少步可以到边缘。C.填充值zero- padding:在外围边缘补充若干圈0,方便从初始位置以步长为单位可以刚好滑倒末尾位置,通俗地讲就是为了总长能被步长整除。最左边一列是输入层,第二列是第一个滤波器(W0),第三列是第二个滤波器(W1),第四列是输出层。*该图有两个神经元,即 depth=2,意味着有两个滤波器数据窗口每次移动两个步长取3*3的局部数据,即stde=2。zero-padding=13cNN的激励层与池化层ReLU激励层不要用 sigmoid,因为它容易饱和、造成终止梯度传递,且没有0中心化ReLU的优点是收敛快,求梯度简单。·池化poo层池化,简言之,即取区域平均或最大,如下图所示Single depth slice1124Xmax pool with 2X2 filters5678 and stride 26832234304y参考文献:·htp:/ yann lecun. com/exdb/lenet/index. html( Yann securη实现的CNN演示,以动画的形式演示了位移、加噪、旋转、压缩等识别,最有价值的是把隐层用图像显示出来了,很生动形象)

本文针对我国人口增长中出现的新特点，建立了两个符合实际情况的预测模型。模型一：基于Logistic 模型，建立了含市、镇、乡人口相互流动关系的微分方程模型，求得全国总人口数在短期内将持续增长，到2010 年、2020 年分别为13.59 亿和14.44亿，具有较好的中短期预测效果。模型二：从年龄转移与总和生育率出发，建立了离散型人口发展模型。1. 针对性别比例，引入女性比例转移矩阵，利用计算机进行随机模拟，建立起动态的女性比例转移矩阵；2. 针对人口迁移，引入人口迁移率矩阵，将迁移率标准化后利用平均迁移率实现了对迁移人数的预测；3. 针对死亡率，利用分段加权法估计其随时间的变

EXCEL函数大全、实例、教程

包含数字水印的matlab源代码和图像数字水印的方案毕业论文MATLAB(DCT+DWT)