shisanchuan/remortcamerconnect_arm-pc

this is a project about how to transmit picture from ARM's camer to Ubuntu'PC, and take opencv's tool detect picture.some ostacle will be recode and give a way to fix.

RemortCamerConnect_ARM-PC

this is a project about how to transmit picture from ARM's camer to Ubuntu'PC, and take opencv's tool detect picture.some obstacle will be recode and give a way to fix.

bulid

on different platform can use this command,file_name and outname is easy to understand,"pkg-config --cflags --libs opencv" is used to link libopencv

```shell
g++ -std=c++11 {file_name} -o {outname} `pkg-config --cflags --libs opencv`
```

cd RemortCamerConnect_ARM-PC/TCP/
cmake .
cd build/src
./camer -s 8080

How to use

you must first run **pc_**.cpp,and then run **pi_**.cpp

• Server : ./camer -s 8080
• Client : ./camer -c 127.0.0.1 8080

Some obstacle i have meet

• TCP&UDP send/recv or sendto/recvfrom buffer problem TCP&UDP buffer have 2^16 byte size(about 65536), so send a picture(640x480)is out this rang,so we have two way to solve this problem,using circulatory transmission or set buffer in linux kernel enviorment: kernel: /proc/sys/net/core/rmem_default,/proc/sys/net/core/wmem_default,/proc/sys/net/core/rmem_max,/proc/sys/net/core/wmem_max; recv buffer: cat /proc/sys/net/ipv4/tcp_rmem,three parameter is min,default,max(bit); send buffer: cat /proc/sys/net/ipv4/tcp_wmem,three parameter is min,dafault,max(bit); so you can adjust mem:

   sysctl -w net.core.rmem_max= 8388608   <-size you want set
   ...		<- wmem_max,wmem_min,wmem_default etc.
   sysctl -w net.ipv4.tcp_mem='8388608 8388608 8388608'  <-min,default,max

  you can get network flow by this command:

   ifconfig
   sudo tcpdump -ni eth0

• How to transmit opencv's Mat into *unsigned char? when i get a image with Mat's format cv::Mat image=imread("pic.jpg",0),i want to translate Mat's format into *unsigned char:

   unsigned char* image_buffer=image.data;
   memcpy(buffer,image_buffer,image_size);

  translate *unsigned char into Mat's format:

   cv::Mat image=cv::Mat(cv::Size(wide,high),CV_8UC1,buffer);
   cv::imshow("pic",image);

  it seem have a better choise in translate picture by use imencode and imdecode

   //imag encode
   std::vector<int> quality;
   quality.push_back(CV_IMWRITE_JPEG_QUALITY);
   quality.push_back(50);
   cv::imencode(".jpg",Mat imag,std::vector<uchar>& send_buffer,quality);
   //imag decode
   cv:: Mat image=cv::imdecode(vector<uchar> & recv_buffer,CV_LOAD_IMAGE_COLOR);
  

• How to use thread liberate in C++?

   void func(void*);
   thread threads;
   threads=thread(func,void*);
   threads.detach();//unblock or block:threads.join();
   ...

  if we want to creat a class to manage thread easily:

   int mythread::Add(void (*fun)(void)){
   threads.push_back(std::thread(fun));
   std::cout<<">>>"<<threads.size()<<std::endl;
   threads[threads.size()-1].detach();
   return 0;
   }
   ...
   mythread m;
   m.Add([]{std::cout<<"ffff"<<std::endl;});

  mutilthread have a comment parameter by used std::ref():

   std::thread(fun,std::ref(str));
  

  if we have better choise to use thread ,we can repackage thread lib into a easy used class.

• CMakeLists.txt enable opencvlib and threadlib

   FIND_PACKAGE( OpenCV REQUIRED ) 
   FIND_PACKAGE ( Threads REQUIRED )
   TARGET_LINK_LIBRARIES(camer ${OpenCV_LIBS} ${CMAKE_THREAD_LIBS_INIT})

• HoughCircles Function Promble C++: void HoughCircles(InputArray image, OutputArray circles, int method, double dp, double minDist, double param1=100, double param2=100, int minRadius=0, int maxRadius=0 ),if set minRsdius and MaxRadius ,when image have not a Circle in it,this function will faill by circles is empty,so it is better if set it ignore,and select circle after:

   cv::HoughCircles(grayblur, circles, CV_HOUGH_GRADIENT,2, grayblur.cols/4,CV_HOUGH_GRADIENT ,CV_HOUGH_GRADIENT);		
  for(int i=0;i<circles.size();i++)
  	if(circles[i][2]>40&&circles[i][2]<80)return circles[i];

• PC or USBcamer Problem USBcamer&PC'camer will have a selfdetect before run,so it is better used cap.read(image),and if it return false,we decide not send it.

• imdecode&imencode Problem if imdecode is fail,it will return a empty Mat,The function reads an image from the specified buffer in the memory. If the buffer is too short or contains invalid data, the empty matrix/image is returned,so if picture reciver failure ,ignore this recv is better.

   cv::Mat deimage=cv::imdecode(data_decode,CV_LOAD_IMAGE_COLOR);
   if(deimage.empty())return -1;
   else image=deimage;