Recherche avancée

Sur d’autres sites (6525)

• Unable to correctly format opencv raw video data to pipe to ffmpeg to transcode and stream over network

  8 février 2014, par AnthonyAlatorre

  I am trying to pipe the output of a simple OpenCV program to ffmpeg using the following code.
  `

  int main(){
  
  
  
  
  
     VideoCapture cap(-1);
  
     if(!cap.isOpened()){
  
          cout << "Unable to capture webcam."<< endl;
  
          return -1;
  
     }
  
     cap.set( cv::CAP_PROP_FRAME_WIDTH, 640 );
  
     cap.set( cv::CAP_PROP_FRAME_HEIGHT, 480 );
  
  
  
     bool done = false;
  
     while(!done){
  
  
  
          Mat frame;
  
          cap >> frame;
  
          if(frame.empty()){
  
              cout << "Unable to read frame."<< endl;
  
              return -1;
  
          }
  
  
  
          //imshow("frame",frame);
  
          cout << frame.data;
  
          if(cv::waitKey(25) >= 0){ done=true; }
  
  
  
     }
  
  
  
     destroyAllWindows();
  
     return 0;
  
  }

  `
  I have tried using the commands from the following posts : Pipe raw OpenCV images to FFmpeg and How to avoid a growing delay with ffmpeg between sound and raw video data ?. But I still seem to be getting the incorrect output. I have also tried converting the BGR image to YCrCb and then splitting the YCrCb image to merge the same image back together however just switching the Cr & Cb channels. Here's the code for that

     bool done = false;
  
     while(!done){
  
  
  
          Mat frame;
  
          cap >> frame;
  
          if(frame.empty()){
  
              cout << "Unable to read frame."<< endl;
  
              return -1;
  
          }
  
  
  
          //imshow("frame",frame);
  
  //      cout << frame.data;
  
          cvtColor( frame, frame, cv::COLOR_BGR2YCrCb);
  
          vector<mat> ycrcb(3);
  
          split(frame,ycrcb);
  
          vector<mat> newChannels;
  
          newChannels.push_back(ycrcb[0]);
  
          newChannels.push_back(ycrcb[2]);
  
          newChannels.push_back(ycrcb[1]);
  
          Mat res;
  
          merge(newChannels,res);
  
          //cout &lt;&lt; res.data;
  
          //imshow("res",res);
  
          cout &lt;&lt; res.data;
  
  
  
          if(cv::waitKey(25) >= 0){ done=true; }
  
  
  
  
  
  
  
  
  
     }
  
  
  
  
  
  
  
  
  
     destroyAllWindows();
  
     return 0;
  
  }
  
  </mat></mat>

  `

  This seems to stop the video from stuttering and what not however the color is incorrect. However I am not sure if I have to do a 4:2:0 sub sampling on the newly converted YCrCb image. I have also gone through such post for help : Converting YUV into BGR or RGB in OpenCV and How to convert YUV422 (sub sampling) to YUV ?. I also have screenshots of the problems however I do not have enough rep points to post images. Any help would be appreciated and thank you very much for your time.

• FFmpeg Video Library Quick Review

  3 juillet 2019, par Rich_F

  I’m in need of reducing 18 videos so I can review them daily. I’d like to script some kind of compression or sampling using FFmpeg to not have to sit through 6 hours of video every day.

  What is the best way to review 6 hours of videos captured every day on SD cards ? The videos come in two cards, a 32 GB card ( 4 hours) and a 16 GB card ( 2 hours).

  Currently, I’m concatenating all videos and creating a screenshot every 3 seconds, which seems the longest I can go without missing something in the field of view (a vehicle will enter/leave). Still, this is thousands of images per card, and I select all, then preview and scroll down using page up/down. Any vehicle quickly appears in the thumbnails. This system isn’t bad, but I’m always inquisitive. I’m looking for more ways of sampling or seeing if any motion detection or frame analysis might register where I should be looking in my library of original captured videos.

  Options :

  1. Motion capture by analysis of frame by frame through trial and error on sensitivity to not measure wind on the bushes. This could be output to filename and time within the video (videos are captured 10 minutes long).

  2. Take my exported list of png files and string them together somehow into a slide show.

  3. Version 1 here, plus some retroactive capture of clips inside master clips.

  4. Some kind of compression. Again, I have to respect the 3-second max jump.

  Any other options that would be relatively easy to achieve without moving into any other software ? I’m trying to put the heavy lifting on the computer, as staring at videos is not something I’m finding I’ll be able to do for many days.

• Improve whileloading() vs. canplaythrough / onload under HTML5 (with support for HTTP range/partial requests). onload() may fire early, followed by numerous whileloading() requests during playback as the HTML5 audio object requests and buffers more audio ranges (eg., 0-10 seconds), eventually matching the total sound duration. (There is no bytes loaded / total, that I’m aware of, to report here, so it’s 0 / 1 and then 0.375 / 1, for example. I suppose I could use integer values, to be "safe", in the future.)

  21 juin 2012, par Scott Schiller

  m script/soundmanager2-jsmin.js m script/soundmanager2-nodebug-jsmin.js m script/soundmanager2-nodebug.js m script/soundmanager2.js Improve whileloading() vs. canplaythrough / onload under HTML5 (with support for HTTP range/partial requests). onload() may fire early, followed by numerous (...)