将QTGStreamer与QT和Raspberry Pi一起使用

Question

我在将视频流从我的覆盆子Pi解码到带有QT GUI的笔记本电脑时遇到了问题。

我的PI管道是（使用Adafruit Raspberry Pi摄像头）：

raspivid -t 999999 -h 480 -w 640 -fps 25 -hf -b 2000000 -o - | gst-launch-1.0 -v fdsrc ! h264parse !  rtph264pay config-interval=1 pt=96 ! gdppay ! tcpserversink host=10.0.0.128 port=5000 

只需使用管道上的笔记本电脑上的查看器：

gst-launch-1.0 -v tcpclientsrc host=10.0.0.128 port=5000  ! gdpdepay !  rtph264depay ! avdec_h264 ! videoconvert ! autovideosink sync=false 

以相当不错的速度给出了非常不错的颜色视频，尽管我没有衡量帧速率。

当我在GUI应用中使用QTGStreamer（源宽度= 640，高度= 480，我假设一个8位RGB映像）我正在以下460800代码中获得缓冲尺寸，我希望它能是921600。如果我使用qimage :: format_rgb888程序将崩溃，因为图像缓冲区太小。如果我使用qimage :: format_index8它将运行良好，请在我的GUI中显示视频，而所有的都是黑色和白色。有人有主意吗？这是我的相关代码：

bool CameraStreamer::initStreamer()
{
    gst_init (NULL, NULL);
    //gst-launch-1.0 -v tcpclientsrc host=10.0.0.128 port=5000  ! gdpdepay !  rtph264depay ! avdec_h264 ! videoconvert ! autovideosink sync=false
    pipeline = gst_pipeline_new("Camera");
    source                  = gst_element_factory_make ("tcpclientsrc",           "cam-source");
    depay                   = gst_element_factory_make("gdpdepay",      "depay");
    rtpdepay                = gst_element_factory_make("rtph264depay","rtp-depay");
    decoder                 = gst_element_factory_make ("avdec_h264",          "videodecoder");
    videoconvert            = gst_element_factory_make("videoconvert","video-convert");
    sink                    = gst_element_factory_make ("appsink",          "video-output");
    if (!pipeline || !source  || !depay || !rtpdepay || !decoder || !videoconvert || !sink ) {
      qDebug() << "One element could not be created. Exiting.n";
      return false;
    }
    callbacks.eos = NULL;
    callbacks.new_sample = newBufferCallback;
    callbacks.new_preroll = NULL;
    gst_app_sink_set_callbacks((GstAppSink *) sink, &callbacks, this, NULL);
    g_object_set (G_OBJECT(source), "port", 5001, NULL);
    g_object_set (G_OBJECT(source),"host","10.0.0.128",NULL);
    gst_bin_add_many (GST_BIN (pipeline),
                      source, depay,rtpdepay,decoder, videoconvert,sink, NULL);
    if (!gst_element_link_many (source, depay,rtpdepay,decoder, videoconvert,sink, NULL))
        g_warning ("Main pipeline link Fail...");
    ret = gst_element_set_state (pipeline, GST_STATE_PLAYING);
    if (ret == GST_STATE_CHANGE_FAILURE)
    {
        g_printerr ("Unable to set the pipeline to the playing state.");
        gst_object_unref (pipeline);
        return false;
    }
    return true;
}
GstFlowReturn CameraStreamer::newBufferCallback(GstAppSink *app_sink, void *obj)
{
    if(app_sink == NULL)
    {
        qDebug() << "app_sink is NULL";
        return GST_FLOW_ERROR;
    }
    GstSample* sample = gst_app_sink_pull_sample(app_sink);
    if(!sample)
    {
        qDebug() << "Error retreiving buffer...";
        return GST_FLOW_ERROR;
    }
    GstCaps* caps = gst_sample_get_caps (sample);
    if (!caps) {
        qDebug() << "could not get snapshot formatn";
        exit (-1);
    }
    gint width, height;
    GstStructure* s = gst_caps_get_structure (caps, 0);
    int res = gst_structure_get_int (s, "width", &width)
        | gst_structure_get_int (s, "height", &height);
    if (!res) {
        qDebug() << "could not get snapshot dimensionn";
        exit (-1);
    }
    GstMapInfo map;
    GstBuffer *buffer = gst_sample_get_buffer (sample);
    qDebug() << "size: " << gst_buffer_get_size(buffer);
    gst_buffer_map (buffer, &map, GST_MAP_READ);
    QImage img(map.data,width,height, QImage::Format_RGB888);
    img = img.copy();
    ((CameraStreamer*)obj)->emitNewImage(img);
    gst_buffer_unmap (buffer, &map);
    gst_sample_unref (sample);
    return GST_FLOW_OK;
}

Answer 1

如果是i420，则布局为：

460800 = 640 * 480 + 320 * 240 + 320 * 240

luma plain y是640 * 480，色平原u和v均为320 *240。因此，紫外线平原的分辨率较小，在循环循环时考虑了这些阵列。

wikipedia的颜色转换公式：

R = Y + 1.140 * V
G = Y - 0.395 * U - 0.581 * V
B = Y + 2.032 * U

Answer 2

，所以经过荒谬的时间和谷歌搜索后，我找到了答案。我最终使用OpenCV进行了实际的颜色转换。这是我的方法（从上方继续）：

GstBuffer *buffer = gst_sample_get_buffer (sample);
gst_buffer_map (buffer, &map, GST_MAP_READ);
cv::Mat temp_mat = cv::Mat(cv::Size(width, height+height/2), CV_8UC1, (char*)map.data);
cv::Mat result(height,width,3);
cv::cvtColor(temp_mat,result,CV_YUV2RGB_I420,3);
QImage rgb(result.size().width,result.size().height,QImage::Format_RGB888);
memcpy(rgb.scanLine(0), (unsigned char*)result.data, rgb.width() * rgb.height() * result.channels());
((CameraStreamer*)obj)->emitNewImage(rgb);
gst_buffer_unmap (buffer, &map);
gst_sample_unref (sample);

我将发布有关我的应用程序Git Repo的更多信息，但我认为这可能会对其他人有所帮助。

这是链接：相机流示例