使用H264编解码器和C+，通过FFMpeg将FLV流式传输到RTMP+；flv.js的API

我想使用H264编解码器使用OpenCV从网络摄像头流式传输实时视频，并将其转换为FLV，然后通过RTMP服务器流式传输，并使用FLV.js在浏览器中捕获流。基本上，除了不能在flv.js中读取流之外，我的一切都正常工作。我可以用

ffplay

打开流，所以我认为至少大多数设置都是正确的

我目前的执行情况：

#include <iostream>
#include <vector>

#include <opencv2/highgui.hpp>
#include <opencv2/video.hpp>

extern "C" {
#include <libavformat/avformat.h>
#include <libavcodec/avcodec.h>
#include <libavutil/avutil.h>
#include <libavutil/pixdesc.h>
#include <libavutil/imgutils.h>
#include <libswscale/swscale.h>
}

void stream_video(double width, double height, int fps, int camID)
{
  av_register_all();
  avformat_network_init();

  const char *output = "rtmp://localhost/live/stream";
  const AVRational dst_fps = {fps, 1};
  int ret;

  // initialize video capture device
  cv::VideoCapture cam(camID);
  if (!cam.isOpened())
  {
    std::cout << "Failed to open video capture device!" << std::endl;
    exit(1);
  }

  cam.set(cv::CAP_PROP_FRAME_WIDTH, width);
  cam.set(cv::CAP_PROP_FRAME_HEIGHT, height);

  // allocate cv::Mat with extra bytes (required by AVFrame::data)
  std::vector<uint8_t> imgbuf(height * width * 3 + 16);
  cv::Mat image(height, width, CV_8UC3, imgbuf.data(), width * 3);

  // open output format context
  AVFormatContext *outctx = nullptr;
  ret = avformat_alloc_output_context2(&outctx, nullptr, "flv", output);
  if (ret < 0)
  {
    std::cout << "Could not allocate output format context!" << std::endl;
    exit(1);
  }

  // open output IO context
  if (!(outctx->oformat->flags & AVFMT_NOFILE))
  {
    ret = avio_open2(&outctx->pb, output, AVIO_FLAG_WRITE, nullptr, nullptr);
    if (ret < 0)
    {
      std::cout << "Could not open output IO context!" << std::endl;
      exit(1);
    }
  }

  // create new video stream
  AVCodec *codec = avcodec_find_encoder(AV_CODEC_ID_H264);
  AVStream *strm = avformat_new_stream(outctx, codec);
  AVCodecContext *avctx = avcodec_alloc_context3(codec);

  avctx->codec_id = AV_CODEC_ID_H264;
  avctx->width = width;
  avctx->height = height;
  avctx->pix_fmt = AV_PIX_FMT_YUV420P;
  avctx->framerate = dst_fps;
  avctx->time_base = av_inv_q(dst_fps);

  ret = avcodec_parameters_from_context(strm->codecpar, avctx);
  if (ret < 0)
  {
    std::cout << "Could not initialize stream codec parameters!" << std::endl;
    exit(1);
  }

  AVDictionary *opts = nullptr;
  av_dict_set(&opts, "preset", "superfast", 0);
  av_dict_set(&opts, "tune", "zerolatency", 0);

  // open video encoder
  ret = avcodec_open2(avctx, codec, &opts);
  if (ret < 0)
  {
    std::cout << "Could not open video encoder!" << std::endl;
    exit(1);
  }

  // initialize sample scaler
  SwsContext *swsctx = sws_getContext(width, height, AV_PIX_FMT_BGR24, width, height, avctx->pix_fmt, SWS_BICUBIC, nullptr, nullptr, nullptr);
  if (!swsctx)
  {
    std::cout << "Could not initialize sample scaler!" << std::endl;
    exit(1);
  }

  // allocate frame buffer for encoding
  AVFrame *frame = av_frame_alloc();

  std::vector<uint8_t> framebuf(av_image_get_buffer_size(avctx->pix_fmt, width, height, 1));
  av_image_fill_arrays(frame->data, frame->linesize, framebuf.data(), avctx->pix_fmt, width, height, 1);
  frame->width = width;
  frame->height = height;
  frame->format = static_cast<int>(avctx->pix_fmt);

  // write header
  ret = avformat_write_header(outctx, nullptr);
  if (ret < 0)
  {
    std::cout << "Could not write header!" << std::endl;
    exit(1);
  }

  // encoding loop
  int64_t frame_pts = 0;
  unsigned nb_frames = 0;
  bool end_of_stream = false;

  do
  {
    nb_frames++;

    if (!end_of_stream)
    {
      cam >> image;
      // convert cv::Mat to AVFrame.
      const int stride[] = {static_cast<int>(image.step[0])};
      sws_scale(swsctx, &image.data, stride, 0, image.rows, frame->data, frame->linesize);
      frame->pts = frame_pts++;
    }
    // encode video frame.
    AVPacket pkt = {0};
    av_init_packet(&pkt);

    ret = avcodec_send_frame(avctx, frame);
    if (ret < 0)
    {
      std::cout << "Error sending frame to codec context!" << std::endl;
      exit(1);
    }

    ret = avcodec_receive_packet(avctx, &pkt);
    if (ret < 0)
    {
      std::cout << "Error receiving packet from codec context!" << std::endl;
      exit(1);
    }

    // rescale packet timestamp.
    av_packet_rescale_ts(&pkt, avctx->time_base, strm->time_base);
    // write packet.
    pkt.pts = AV_NOPTS_VALUE;
    pkt.dts = AV_NOPTS_VALUE;
    av_interleaved_write_frame(outctx, &pkt);

    std::cout << " Frames: " << nb_frames << '\r' << std::flush;

    av_packet_unref(&pkt);
  } while (!end_of_stream);

  av_write_trailer(outctx);
  std::cout << nb_frames << " frames encoded" << std::endl;

  av_frame_free(&frame);
  avcodec_close(avctx);
  avio_close(outctx->pb);
  avformat_free_context(outctx);
}

int main()
{
  double width = 1280, height = 720, fps = 30;
  int camID = 1;

  stream_video(width, height, fps, camID);

  return 0;
}

如果我像

ffmpeg-re-I input.mp4-c copy-f flv那样流视频，我将非常感谢任何帮助来修复流，使之能够与
flv.js
正常工作rtmp://localhost/live/stream

我可以在

flv.js

中打开流，没有任何问题，所以我想大致实现内部代码中的“此命令”。 如果有人想编译代码并检查代码，我也会将代码放在GitHub存储库中。

我自己解决这个问题。 这不能按预期工作的主要原因是AVStream extradata（sps和pps标头）为空。我需要从AVCodecContext手动复制

extradata

和

extradata\u size

（不确定为什么这不是自动完成的）。在我完成这项工作后，我第一次在

flv.js

中看到了这张图片。然后我只需要正确地计算

frame->pts

，就可以让视频流正常工作。我在下面附上完整的工作代码，以防其他人遇到同样的问题

#include <iostream>
#include <vector>

#include <opencv2/highgui.hpp>
#include <opencv2/video.hpp>

extern "C" {
#include <libavformat/avformat.h>
#include <libavcodec/avcodec.h>
#include <libavutil/imgutils.h>
#include <libswscale/swscale.h>
}

cv::VideoCapture get_device(int camID, double width, double height)
{
  cv::VideoCapture cam(camID);
  if (!cam.isOpened())
  {
    std::cout << "Failed to open video capture device!" << std::endl;
    exit(1);
  }

  cam.set(cv::CAP_PROP_FRAME_WIDTH, width);
  cam.set(cv::CAP_PROP_FRAME_HEIGHT, height);

  return cam;
}

void initialize_avformat_context(AVFormatContext *&fctx, const char *format_name)
{
  int ret = avformat_alloc_output_context2(&fctx, nullptr, format_name, nullptr);
  if (ret < 0)
  {
    std::cout << "Could not allocate output format context!" << std::endl;
    exit(1);
  }
}

void initialize_io_context(AVFormatContext *&fctx, const char *output)
{
  if (!(fctx->oformat->flags & AVFMT_NOFILE))
  {
    int ret = avio_open2(&fctx->pb, output, AVIO_FLAG_WRITE, nullptr, nullptr);
    if (ret < 0)
    {
      std::cout << "Could not open output IO context!" << std::endl;
      exit(1);
    }
  }
}

void set_codec_params(AVFormatContext *&fctx, AVCodecContext *&codec_ctx, double width, double height, int fps)
{
  const AVRational dst_fps = {fps, 1};

  codec_ctx->codec_tag = 0;
  codec_ctx->codec_id = AV_CODEC_ID_H264;
  codec_ctx->codec_type = AVMEDIA_TYPE_VIDEO;
  codec_ctx->width = width;
  codec_ctx->height = height;
  codec_ctx->gop_size = 12;
  codec_ctx->pix_fmt = AV_PIX_FMT_YUV420P;
  codec_ctx->framerate = dst_fps;
  codec_ctx->time_base = av_inv_q(dst_fps);
  if (fctx->oformat->flags & AVFMT_GLOBALHEADER)
  {
    codec_ctx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
  }
}

void initialize_codec_stream(AVStream *&stream, AVCodecContext *&codec_ctx, AVCodec *&codec)
{
  int ret = avcodec_parameters_from_context(stream->codecpar, codec_ctx);
  if (ret < 0)
  {
    std::cout << "Could not initialize stream codec parameters!" << std::endl;
    exit(1);
  }

  AVDictionary *codec_options = nullptr;
  av_dict_set(&codec_options, "profile", "high", 0);
  av_dict_set(&codec_options, "preset", "superfast", 0);
  av_dict_set(&codec_options, "tune", "zerolatency", 0);

  // open video encoder
  ret = avcodec_open2(codec_ctx, codec, &codec_options);
  if (ret < 0)
  {
    std::cout << "Could not open video encoder!" << std::endl;
    exit(1);
  }
}

SwsContext *initialize_sample_scaler(AVCodecContext *codec_ctx, double width, double height)
{
  SwsContext *swsctx = sws_getContext(width, height, AV_PIX_FMT_BGR24, width, height, codec_ctx->pix_fmt, SWS_BICUBIC, nullptr, nullptr, nullptr);
  if (!swsctx)
  {
    std::cout << "Could not initialize sample scaler!" << std::endl;
    exit(1);
  }

  return swsctx;
}

AVFrame *allocate_frame_buffer(AVCodecContext *codec_ctx, double width, double height)
{
  AVFrame *frame = av_frame_alloc();

  std::vector<uint8_t> framebuf(av_image_get_buffer_size(codec_ctx->pix_fmt, width, height, 1));
  av_image_fill_arrays(frame->data, frame->linesize, framebuf.data(), codec_ctx->pix_fmt, width, height, 1);
  frame->width = width;
  frame->height = height;
  frame->format = static_cast<int>(codec_ctx->pix_fmt);

  return frame;
}

void write_frame(AVCodecContext *codec_ctx, AVFormatContext *fmt_ctx, AVFrame *frame)
{
  AVPacket pkt = {0};
  av_init_packet(&pkt);

  int ret = avcodec_send_frame(codec_ctx, frame);
  if (ret < 0)
  {
    std::cout << "Error sending frame to codec context!" << std::endl;
    exit(1);
  }

  ret = avcodec_receive_packet(codec_ctx, &pkt);
  if (ret < 0)
  {
    std::cout << "Error receiving packet from codec context!" << std::endl;
    exit(1);
  }

  av_interleaved_write_frame(fmt_ctx, &pkt);
  av_packet_unref(&pkt);
}

void stream_video(double width, double height, int fps, int camID)
{
  av_register_all();
  avformat_network_init();

  const char *output = "rtmp://localhost/live/stream";
  int ret;
  auto cam = get_device(camID, width, height);
  std::vector<uint8_t> imgbuf(height * width * 3 + 16);
  cv::Mat image(height, width, CV_8UC3, imgbuf.data(), width * 3);
  AVFormatContext *ofmt_ctx = nullptr;
  AVCodec *out_codec = nullptr;
  AVStream *out_stream = nullptr;
  AVCodecContext *out_codec_ctx = nullptr;

  initialize_avformat_context(ofmt_ctx, "flv");
  initialize_io_context(ofmt_ctx, output);

  out_codec = avcodec_find_encoder(AV_CODEC_ID_H264);
  out_stream = avformat_new_stream(ofmt_ctx, out_codec);
  out_codec_ctx = avcodec_alloc_context3(out_codec);

  set_codec_params(ofmt_ctx, out_codec_ctx, width, height, fps);
  initialize_codec_stream(out_stream, out_codec_ctx, out_codec);

  out_stream->codecpar->extradata = out_codec_ctx->extradata;
  out_stream->codecpar->extradata_size = out_codec_ctx->extradata_size;

  av_dump_format(ofmt_ctx, 0, output, 1);

  auto *swsctx = initialize_sample_scaler(out_codec_ctx, width, height);
  auto *frame = allocate_frame_buffer(out_codec_ctx, width, height);

  int cur_size;
  uint8_t *cur_ptr;

  ret = avformat_write_header(ofmt_ctx, nullptr);
  if (ret < 0)
  {
    std::cout << "Could not write header!" << std::endl;
    exit(1);
  }

  bool end_of_stream = false;
  do
  {
    cam >> image;
    const int stride[] = {static_cast<int>(image.step[0])};
    sws_scale(swsctx, &image.data, stride, 0, image.rows, frame->data, frame->linesize);
    frame->pts += av_rescale_q(1, out_codec_ctx->time_base, out_stream->time_base);
    write_frame(out_codec_ctx, ofmt_ctx, frame);
  } while (!end_of_stream);

  av_write_trailer(ofmt_ctx);

  av_frame_free(&frame);
  avcodec_close(out_codec_ctx);
  avio_close(ofmt_ctx->pb);
  avformat_free_context(ofmt_ctx);
}

int main()
{
  // av_log_set_level(AV_LOG_DEBUG);
  double width = 1280, height = 720;
  int camID = 1, fps = 25;

  stream_video(width, height, fps, camID);

  return 0;
}

#包括
#包括
#包括
#包括
外部“C”{
#包括
#包括
#包括
#包括
}
cv：：视频捕获获取设备（int camID，双宽，双高）
{
cv：：视频捕获cam（camID）；
如果（！cam.isOpened（））
{
std:：cout codec_type=AVMEDIA_type_VIDEO；
编解码器\u ctx->宽度=宽度；
编解码器\u ctx->高度=高度；
codec_ctx->gop_size=12；
编解码器\u ctx->pix\u fmt=AV\u pix\u fmt\u YUV420P；
codec_ctx->帧率=dst_fps；
编解码器ctx->time_base=av_inv_q（dst_fps）；
如果（fctx->oformat->flags和AVFMT_GLOBALHEADER）
{
codec_ctx->flags |=AV_codec_FLAG_GLOBAL_头；
}
}
void initialize_codec_stream（AVStream*&stream，AVCodecContext*&codec_ctx，AVCodec*&codec）
{
int-ret=来自上下文的avcodec\u参数（流->codecpar，codec\u ctx）；
如果（ret<0）
{
标准：：cout width=宽度；
框架->高度=高度；
帧->格式=静态广播（编解码器ctx->pix\U fmt）；
返回框；
}
无效写入帧（AVCODECONTEXT*codec\U ctx、AVFormatContext*fmt\U ctx、AVFrame*frame）
{
AVPacket pkt={0}；
av_初始_数据包（&pkt）；
int-ret=avcodec\u send\u frame（codec\u ctx，frame）；
如果（ret<0）
{
std:：cout codepar->extradata\u size=out\u codec\u ctx->extradata\u size；
av_转储_格式（ofmt_ctx，0，输出，1）；
auto*swsctx=初始化样本缩放器（输出编解码器ctx、宽度、高度）；
auto*frame=分配\u帧\u缓冲区（out\u编解码器\u ctx、宽度、高度）；
内部电流大小；
uint8_t*cur_ptr；
ret=avformat\u write\u头（ofmt\u ctx，nullptr）；
如果（ret<0）
{
std：：cout图像；
const int stride[]={static_cast（image.step[0]）}；
sws_比例（swsctx，&image.data，跨距，0，image.rows，frame->data，frame->linesize）；
帧->pts+=av\U重新缩放\U q（1，输出编解码器\U ctx->时基，输出流->时基）；
写入帧（输出编解码器ctx，ofmt ctx，帧）；
}而（！_流的结束_）；
av_写入_预告片（ofmt_ctx）；
av_无帧（&帧）；
avcodec_关闭（out_codec_ctx）；
avio_关闭（ofmt_ctx->pb）；
avformat_free_上下文（ofmt_ctx）；
}
int main（）
{
//av_日志_设置_级别（av_日志_调试）；
双宽=1280，高=720；
int camID=1，fps=25；
流视频（宽度、高度、fps、camID）；
返回0；
}

---

就这样！

恭喜你参加了比赛。我正在尝试使用上述代码将MPEG1流发送到http服务器。我在avcodec_open2中遇到错误，无法打开它。在删除配置文件、预设、调谐和添加AV_dict集（&CODEC_选项，“b”、“2.5M”、0）后，我正在尝试在“mpeg”和AV_CODEC_ID_MPEG1视频中转换；现在可以工作了。我在avcodec_receive_packet call中遇到错误。我不确定MPEG1视频是否可以通过RTMP成功流式传输。请查看RTMP规范文档。

#include <iostream>
#include <vector>

#include <opencv2/highgui.hpp>
#include <opencv2/video.hpp>

extern "C" {
#include <libavformat/avformat.h>
#include <libavcodec/avcodec.h>
#include <libavutil/imgutils.h>
#include <libswscale/swscale.h>
}

cv::VideoCapture get_device(int camID, double width, double height)
{
  cv::VideoCapture cam(camID);
  if (!cam.isOpened())
  {
    std::cout << "Failed to open video capture device!" << std::endl;
    exit(1);
  }

  cam.set(cv::CAP_PROP_FRAME_WIDTH, width);
  cam.set(cv::CAP_PROP_FRAME_HEIGHT, height);

  return cam;
}

void initialize_avformat_context(AVFormatContext *&fctx, const char *format_name)
{
  int ret = avformat_alloc_output_context2(&fctx, nullptr, format_name, nullptr);
  if (ret < 0)
  {
    std::cout << "Could not allocate output format context!" << std::endl;
    exit(1);
  }
}

void initialize_io_context(AVFormatContext *&fctx, const char *output)
{
  if (!(fctx->oformat->flags & AVFMT_NOFILE))
  {
    int ret = avio_open2(&fctx->pb, output, AVIO_FLAG_WRITE, nullptr, nullptr);
    if (ret < 0)
    {
      std::cout << "Could not open output IO context!" << std::endl;
      exit(1);
    }
  }
}

void set_codec_params(AVFormatContext *&fctx, AVCodecContext *&codec_ctx, double width, double height, int fps)
{
  const AVRational dst_fps = {fps, 1};

  codec_ctx->codec_tag = 0;
  codec_ctx->codec_id = AV_CODEC_ID_H264;
  codec_ctx->codec_type = AVMEDIA_TYPE_VIDEO;
  codec_ctx->width = width;
  codec_ctx->height = height;
  codec_ctx->gop_size = 12;
  codec_ctx->pix_fmt = AV_PIX_FMT_YUV420P;
  codec_ctx->framerate = dst_fps;
  codec_ctx->time_base = av_inv_q(dst_fps);
  if (fctx->oformat->flags & AVFMT_GLOBALHEADER)
  {
    codec_ctx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
  }
}

void initialize_codec_stream(AVStream *&stream, AVCodecContext *&codec_ctx, AVCodec *&codec)
{
  int ret = avcodec_parameters_from_context(stream->codecpar, codec_ctx);
  if (ret < 0)
  {
    std::cout << "Could not initialize stream codec parameters!" << std::endl;
    exit(1);
  }

  AVDictionary *codec_options = nullptr;
  av_dict_set(&codec_options, "profile", "high", 0);
  av_dict_set(&codec_options, "preset", "superfast", 0);
  av_dict_set(&codec_options, "tune", "zerolatency", 0);

  // open video encoder
  ret = avcodec_open2(codec_ctx, codec, &codec_options);
  if (ret < 0)
  {
    std::cout << "Could not open video encoder!" << std::endl;
    exit(1);
  }
}

SwsContext *initialize_sample_scaler(AVCodecContext *codec_ctx, double width, double height)
{
  SwsContext *swsctx = sws_getContext(width, height, AV_PIX_FMT_BGR24, width, height, codec_ctx->pix_fmt, SWS_BICUBIC, nullptr, nullptr, nullptr);
  if (!swsctx)
  {
    std::cout << "Could not initialize sample scaler!" << std::endl;
    exit(1);
  }

  return swsctx;
}

AVFrame *allocate_frame_buffer(AVCodecContext *codec_ctx, double width, double height)
{
  AVFrame *frame = av_frame_alloc();

  std::vector<uint8_t> framebuf(av_image_get_buffer_size(codec_ctx->pix_fmt, width, height, 1));
  av_image_fill_arrays(frame->data, frame->linesize, framebuf.data(), codec_ctx->pix_fmt, width, height, 1);
  frame->width = width;
  frame->height = height;
  frame->format = static_cast<int>(codec_ctx->pix_fmt);

  return frame;
}

void write_frame(AVCodecContext *codec_ctx, AVFormatContext *fmt_ctx, AVFrame *frame)
{
  AVPacket pkt = {0};
  av_init_packet(&pkt);

  int ret = avcodec_send_frame(codec_ctx, frame);
  if (ret < 0)
  {
    std::cout << "Error sending frame to codec context!" << std::endl;
    exit(1);
  }

  ret = avcodec_receive_packet(codec_ctx, &pkt);
  if (ret < 0)
  {
    std::cout << "Error receiving packet from codec context!" << std::endl;
    exit(1);
  }

  av_interleaved_write_frame(fmt_ctx, &pkt);
  av_packet_unref(&pkt);
}

void stream_video(double width, double height, int fps, int camID)
{
  av_register_all();
  avformat_network_init();

  const char *output = "rtmp://localhost/live/stream";
  int ret;
  auto cam = get_device(camID, width, height);
  std::vector<uint8_t> imgbuf(height * width * 3 + 16);
  cv::Mat image(height, width, CV_8UC3, imgbuf.data(), width * 3);
  AVFormatContext *ofmt_ctx = nullptr;
  AVCodec *out_codec = nullptr;
  AVStream *out_stream = nullptr;
  AVCodecContext *out_codec_ctx = nullptr;

  initialize_avformat_context(ofmt_ctx, "flv");
  initialize_io_context(ofmt_ctx, output);

  out_codec = avcodec_find_encoder(AV_CODEC_ID_H264);
  out_stream = avformat_new_stream(ofmt_ctx, out_codec);
  out_codec_ctx = avcodec_alloc_context3(out_codec);

  set_codec_params(ofmt_ctx, out_codec_ctx, width, height, fps);
  initialize_codec_stream(out_stream, out_codec_ctx, out_codec);

  out_stream->codecpar->extradata = out_codec_ctx->extradata;
  out_stream->codecpar->extradata_size = out_codec_ctx->extradata_size;

  av_dump_format(ofmt_ctx, 0, output, 1);

  auto *swsctx = initialize_sample_scaler(out_codec_ctx, width, height);
  auto *frame = allocate_frame_buffer(out_codec_ctx, width, height);

  int cur_size;
  uint8_t *cur_ptr;

  ret = avformat_write_header(ofmt_ctx, nullptr);
  if (ret < 0)
  {
    std::cout << "Could not write header!" << std::endl;
    exit(1);
  }

  bool end_of_stream = false;
  do
  {
    cam >> image;
    const int stride[] = {static_cast<int>(image.step[0])};
    sws_scale(swsctx, &image.data, stride, 0, image.rows, frame->data, frame->linesize);
    frame->pts += av_rescale_q(1, out_codec_ctx->time_base, out_stream->time_base);
    write_frame(out_codec_ctx, ofmt_ctx, frame);
  } while (!end_of_stream);

  av_write_trailer(ofmt_ctx);

  av_frame_free(&frame);
  avcodec_close(out_codec_ctx);
  avio_close(ofmt_ctx->pb);
  avformat_free_context(ofmt_ctx);
}

int main()
{
  // av_log_set_level(AV_LOG_DEBUG);
  double width = 1280, height = 720;
  int camID = 1, fps = 25;

  stream_video(width, height, fps, camID);

  return 0;
}