## RL反卷积都干了些啥

• ，损失函数，为最小化的目标
• ，不受椒盐噪声和点分布函数模糊效应影响的理想原始图像
• ，实际拍摄到的图像
• ，二维卷积
• ，按元素相乘

## 数码噪点的泊松分布

由于光线具有波粒二象性，在进行数字成像时，在像素点上的光子数量在概率上呈泊松分布。

## 问题的转化：极大似然估计

• ：整个图像每个像素的联合概率
• ：实际拍摄到的图像数据，包含噪声，M行N列
• ：无泊松噪声的理想图像数据，M行N列

# 生活意义

1. 可衡量的进步：职称、薪资、销售额等等
2. 创造

## 1. 简介

1. 如何为树莓派编译实时内核
2. 测试实时内核的进程调度延时性能

1. 树莓派3B
2. 8G TF卡

1.获取代码和工具：

2.编译内核

3.替换内核 进入树梅派终端：

1. 伯德图表示什么
2. 开环伯德图表示什么
3. 伯德积分定理意味着什么

## 水床效应(waterbed effect)

1. 没有不稳定的极点(确保上式被减数为0)
2. 极点比零点多至少2个(确保上式被减数为0)

## 1.测试环境

• 操作系统：Windows 10, 64bit
• 编译器：MinGW64
• 32位系统请下载对应的32位版本
• MATLAB2017a
• 2017a之后会略有不同

## 2. 配置环境与流程

1. 下载编译器MinGW64，对于64位系统而言，需要下载个安装器，可以安装相应版本的gcc编译器，我选择的是gcc-4.9.4，在测试环境中可以正常运行

1. 安装器是mingw-w64-install，安装时注意选择64位，同时安装完了还有一个坑：空格。
2. 下载器默认的安装路径"C:Files-w64_64-4.9.4-win32-seh-rt_v5-rev0"是包含有空格的，MATLAB并不能识别该路径，所以可以把mingw64移动到不含空格的路径下，比如"C:"。
2. 验证一下：

## 1. Introduction to Proportional-Resonant Controller

First, let's see the transfer function of PR controller and its bode diagram:

where the is the resonant frequency.

As can be seen in the bode diagram, the gain at 50Hz (314 rad/s) is infinite, so PR controller can be used to track reference of specific frequency. However, the reference frequency is not always a constant, e.g. the frequency of electricity grid. In practice, Quasi-Proportional-Resonant(QPR) can be used to solve this problem.

The transfer function of QPR and its bode diagram:

where the addition of reduces the gain at resonant frequency but increase the band width around resonant frequency.

You can change the parameter and run the demo code below to see how it changes the shape of bode diagram.

## 2. Implementation in C

The code is in the repository Controller, feel free to use it for your own application.

The structure of QPR controller implemented is as below:

As shown in this paper, the structure of QPR controller can be implemented using 2 integrators.

The implemented structure above has 2 advantages:

• Easy to implemented, only 2 integrators used.
• Autonomous resonant frequency adaption: can be fed into the controller and real-time resonant frequency adaption is achieved.

You can run this demo to see the effectiveness of this structure

• is set to be 0 and to be 10
• is set to be

The output response is as below:

As you can see in the result above, the input signal is a sine wave with an amplitude of 1 and the output with an amplitude of 10, corresponding to the parameter-- is 0 and is 10.

# 离散化方法的笔记

s域的传递函数为：

## 6. 其他方法

• 冲激响应不变法
• 阶跃响应不变法（ZOH法，零阶保持器法）
• 斜坡响应不变法（一阶保持器法）
• 幅频响应不变法
• 相频响应不变法