将帧从网络摄像头（RTSP）保存到mp4文件 [英] Saving frames from a network camera (RTSP) to a mp4 file

问题描述

我对如何将视频流保存到mp4文件感到困惑。我使用ffmpeg。让我解释一下问题：

1. 我使用avformat_open_input（），avformat_find_stream_info（）连接到网络摄像机， ，av_read_play（）和我获得的帧与av_read_frame（）。


2. 每次我得到一个帧av_read_frame（），我存储相应的AVPacket在循环缓冲区。
   
3. 在我的应用程序的某些点上，选择了这个循环缓冲区的范围。我发现一个用来从一个关键帧开始的关键帧。


4. 一旦我有一个从关键帧开始的AVPacket的列表，我写头，帧和尾，如下面的代码所述。

问题是，如果我尝试使用VLC，Windows Media Player或其他视频观看它， 。

我也意识到这些包的pts是不连续的，而dts是连续的。我知道B帧，但这是我的情况下的一个问题吗？

  //准备输出
 AVFormatContext * oc = avformat_alloc_context（）; 
 oc-> oformat = av_guess_format（NULL，video.mp4，NULL）; 
 
 //必须写头，包和尾
 avio_open2（& oc-> pb，video.mp4，AVIO_FLAG_WRITE，NULL，NULL）; 
 
 //写标题
 AVStream * stream = avformat_new_stream（oc，（AVCodec *）context-> streams [video_stream_index]  - > codec-> codec）; 
 avcodec_copy_context（stream-> codec，context-> streams [video_stream_index]  - > codec）; 
 stream-> sample_aspect_ratio = context-> streams [video_stream_index]  - > codec-> sample_aspect_ratio; 
 avformat_write_header（oc，NULL）; 
 
 //对于每个帧... 
 ... av_write_frame（oc，circular [k]）; ... 
 
 //写预告片并关闭文件
 av_write_trailer（oc）; 
 avcodec_close（stream-> codec）; 
 avio_close（oc-> pb）; 
 avformat_free_context（oc）;非常感谢，
解决方案
首先：当你使用相机时，最好是通过TCP over TCP（TCP作为传输协议）。 
要启用此功能：
  AVDictionary * ifmtdict; 
 av_dict_set（& ifmtdict，rtsp_transport，tcp，0）; 
 ... 
 avformat_open_input（...，& ifmtdict）; 
 
第二个：
在数据包开始传输后，等待第一个关键帧并开始写入到这一刻的文件。
 
I am quite confused about how to save a video stream into a mp4 file. I am using ffmpeg. Let me explain the problem:

I connect to a network camera via RTSP (H.264 stream) with avformat_open_input(), avformat_find_stream_info(), av_read_play(), and I get frames with av_read_frame().
Each time I get a frame with av_read_frame(), I store the corresponding AVPacket in a circular buffer.
In some points in my application, a range of this circular buffer is selected. I find a key frame used to start from.
Once I have a list of AVPacket's starting from a key frame, I write header, frames, and tail, as I described below in the code.
The problem is that the resulting mp4 video has artifacts if I try to watch it using VLC, Windows Media Player, or another one.

I have also realized that the pts of that packets are not continuous, while dts are continuous. I know about B frames, but is this a problem in my case?
// Prepare the output
AVFormatContext* oc = avformat_alloc_context();
oc->oformat = av_guess_format(NULL, "video.mp4", NULL);

// Must write header, packets, and trailing
avio_open2(&oc->pb, "video.mp4", AVIO_FLAG_WRITE, NULL, NULL);

// Write header
AVStream* stream = avformat_new_stream(oc, (AVCodec*) context->streams[video_stream_index]->codec->codec);
avcodec_copy_context(stream->codec, context->streams[video_stream_index]->codec);
stream->sample_aspect_ratio = context->streams[video_stream_index]->codec->sample_aspect_ratio;
avformat_write_header(oc, NULL);

// FOR EACH FRAME...
... av_write_frame(oc, circular[k]); ...

// Write trailer and close the file
av_write_trailer(oc);
avcodec_close(stream->codec);
avio_close(oc->pb);
avformat_free_context(oc);
Thank you so much,
解决方案
First: when you work with the camera, it is better to work through an RTP over TCP (TCP as transport protocol).
To enable this feature:
AVDictionary *ifmtdict;
av_dict_set(&ifmtdict, "rtsp_transport", "tcp", 0);
...
avformat_open_input (..., &ifmtdict);
Second:
After the packets start coming, wait for the first keyframe and start to write to the file from this moment.

                        
这篇关于将帧从网络摄像头（RTSP）保存到mp4文件的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！