当前位置:网站首页>Robomaster visual tutorial (0) Introduction

Robomaster visual tutorial (0) Introduction

2022-07-07 23:54:00 Raring_ Ringtail

background

Senior graduated , Thinking about the four years of college, I also need to summarize . From RoboMaster Let's write this competition . I was a sophomore in the competition , Do the visual recognition part . The idea was simple , Because I like Linux I didn't want to adjust the electric control of the welding board, so I participated in the visual part ( Although ultimately responsible for a large part of the electronic control ...). Doing vision can use clever calculations and develop some very interesting things , Learn a lot of cutting-edge knowledge .

When I was a sophomore, I really started from scratch . Although the team had graduate students doing vision , But there is no technology inheritance , There is no one to ask questions . Although there are nominally three people in the inspection group , A graduate student, a junior, and me , But they all ran away . So I can understand some new contacts RM The confusion in the heart of visual people . This year, after I finished my postgraduate entrance examination, I went to help this class do a visual , Found a lot of problems , Many of the pits we have stepped on before are still stepping on , So it is still a state of doing things in their own way, delaying the final rush every day .

I intend to RoboMaster The technical essentials of all aspects of vision are written into a series of tutorials , Help novices get started , Also hope to have some reference value for other teams . The code mainly refers to Southeast University SuperNova Open source code of the team ( Portal :GitHub), Their code is excellent , Built a good project framework . I admire their courage to share such good code .

Recently applied for a wechat official account , Name is Notes on Jiangda . As an engineering man, it's a headache to name , With reference to Yeon meow's Netease cloud music radio, Yeon meow remembers a Jiangda diary , It is convenient for me to help others by summarizing my own learning and research experience . Interested friends can pay attention to .
This series of tutorials will be constantly updated .

RoboMaster Visual tutorial directory

1. camera

  • Camera parameters
    Roller shutter exposure and global exposure
    Exposure
    Gamma correct
    Frame rate and camera selection
  • The lens
  • Filter
  • Linux Camera drive
    Camera debugging tool
    Use RMVideoCapture Class calls the camera
    Use OpenCV Of VideoCapture To call the camera

2. A clever plan (Nvidia Tegra K1) The system configuration

  • overview
  • Wonderful calculation data link summary
  • Smart computing system reset / clone / recovery
  • Miaoshu what to do after installing the system
    Miaozao connects directly to the computer through the network cable and shares the computer network
    Smart remote debugging
    Wonderful installation FTP
    Smart computing configuration software source
    The title bar of smart computing system jumps one by one, and the solution
    Wonderful installation OpenCV / CUDA
    Wonderful installation GCC5
    Wonderful calculation installation configuration Qt Creator

3. Visual recognition program framework

  • overview
  • Multithreading
  • Besides multithreading , You can also use multiple processes
  • Next, take the open source program of Southeast University as an example to talk about their overall architecture
    Let's get to the point
    Project profile overview
    ImgProdCons class
    The main function
    Wrap the algorithm with classes

4. Armor plate recognition algorithm

  • overview
    Here are some information links , Classics !
  • Armor plate identification
  • test_infantry.cpp
  • Analyze the armor plate
  • Identification function int ArmorDetector::detect()

5. Target location solution ( Get corners through pixels )

  • overview
  • The defect of using pixel coordinates directly
  • Camera calibration
  • Get the required corner according to the principle of small hole imaging
  • Angle measurement verification

6. Target location solution (PnP Solve the relative position between the target and the camera )

  • overview
  • Algorithm principle
  • solvePnP Use process of
  • experiment : Measure the position of the QR code relative to the camera
  • RoboMaster Position solution in visual program
  • Expand

7. Windmill big symbol recognition

  • Share some information links

8. A serial port communication

  • overview
  • DJI Open source code serial port
  • Southeast University open source code serial port part
  • Qt Write serial port assistant
  • A little digression

9. Windmill big symbol recognition 2

  • Image preprocessing
  • Armor plate contour recognition
  • Identify the armor plate to hit
  • Windmill positioning

10. Target forecast

  • Why should we make target prediction
  • Several situations and analysis of prediction problems
  • How to make predictions

11. summary

I applied for my own official account Notes on Jiangda , It is convenient for me to help others by summarizing my own learning and research experience . Interested friends can pay attention to .

原网站

版权声明
本文为[Raring_ Ringtail]所创,转载请带上原文链接,感谢
https://yzsam.com/2022/02/202202130554125633.html