Opencv: Raspberry pi rtsp server, ffmpeg decoding

[kwstanths]

System information (version)

• OpenCV => 3.1
• Opencv-python => 3.1.0-dev
• Operating System / Platform => Ubuntu 16.04 LTS

Detailed description

I want to process an rtsp stream created from a raspberry camera frame by frame with python or C++.

Steps to reproduce

.py
     import cv2
     rtsp_url = "rtsp://192.168.55.151:8554/stream.sdp"
     video = cv2.VideoCapture(rtsp_url)
     while(1):
            ret, frame = video.read()
            cv2.imshow('VIDEO', frame)
            cv2.waitKey(1)

The stream is created from the raspberry machine using vlc: raspivid -o - -t 0 -hf -w 640 -h 480 -fps 30 | cvlc stream:///dev/stdin --sout "#rtp{sdp=rtsp://192.168.55.151:8554/stream.sdp}" :demux=h264

The stream is successfully opened from vlc player with: rtsp://192.168.55.151:8554/stream.sdp

From the python code though, i get: method SETUP failed: 461 Client error (an ffmpeg error)

If i run the command line ffmpeg: ffmpeg -rtsp_transport tcp -i rtsp://192.168.55.151:8554/stream.sdp
i get the same exact error: method SETUP failed: 461 Client error as well as Protocol not supported
If i run the command line ffmpeg: ffmpeg -rtsp_transport udp -i rtsp://192.168.55.151:8554/stream.sdp
the stream decodes successfully.

Is this transport protocol a bug, or am i doing something wrong? Can i specify the transport protocol in the OpenCV code?

[alalek]

To prevent network glitches / video stream disconnection / etc OpenCV forces TCP transport protocol. You should reconfigure your rstp server for TCP support or use non-FFmpeg OpenCV backend (Gstreamer?).

BTW, Usage questions should go to Users OpenCV Q/A forum: http://answers.opencv.org

[masterchop]

You may want to install the UV4L driver for the raspberry camera, it will allow you to have a url with the video stream which is compatible iwth opencv and works just fine. i tested already.

[alfandinorasyid]

You may want to install the UV4L driver for the raspberry camera, it will allow you to have a url with the video stream which is compatible iwth opencv and works just fine. i tested already.

do you have a tutorial for this?