Application that streams video segment files to predefined remote location.

Project contains two main components: streamer application and capturer script.

1. Streamer is responsible for sending files to remote location and controlling congestion and segments backlog.
2. Capturer is responsible for grabbing H264 video via raspivid application and at the same time segmenting it using FFmpeg's hls muxer. Segment files are saved in a predefined location for pickup by the streamer.

Capturer is written specifically for Raspberry Pi (relies on raspivid), but segment files can be delivered by other means as long as valid configuration is provided for the streamer component (which is device-agnostic).

Usually segment files has fixed duration, so if application is configured to capture 10 seconds of video stream and put it to segment file, we have 10 seconds for sending that file. To control congestion while sending segments, streamer queues files from newest to oldest and creates a time window that it has for sending files until new segment arrives. Time window for sending batch of segments is 90% of single segment duration - given default 10 seconds for each segment, that's 9 seconds. After sending a segment, time spent on sending it is subtracted from time window. Segments are picked up from queue until there's time left in time window.

This behaviour is different that usual streaming via UPD sockets in a way that segments backlog is more resilient to short network disruptions. In that case stream won't be missing frames - unless streaming device doesn't have needed bandwidth capacity to clear the backlog.

Due to storage constraints backlog have a defined duration that, when exceeded, will force streamer to drop oldest segments from queue, permanently removing them from storage. This precaution must be taken in order to avoid running out of storage space that would disallow saving new segments.

Any decent HTTP server that supports PUT and DELETE method will be able to receive segment files. Using HTTP server allows access control (authentication) and transport security (HTTPS) out of the box. Due to privacy reasons using HTTPS is highly encouraged.

Example configuration for nginx:

http {
    # ...
    server {
        # ...

        location /camera {
            # Uncomment it if you want to have this location indexed by nginx.
            #autoindex on;

            # Uncomment it if you want to restrict access to this location.
            # Remember to set proper path and appropriate access rights to password file.
            #auth_basic "Restricted";
            #auth_basic_user_file /etc/nginx/.htpasswd;

            client_body_temp_path /var/www/camera/.client_tmp;
            client_max_body_size 10m;

            dav_methods PUT DELETE MKCOL COPY MOVE;
            dav_access group:rw  all:r;
            create_full_put_path on;
        }
    }
}

Configuration is done by environment variables - they must be set before launching either capturer or streamer components.

List of possible values:

1. CAMERA_BASE_URL - Base URL where segment files will be transferred to.
2. CAMERA_USERNAME - If set, then username - along with password - will be included in HTTP requests to server (currently only Basic authentication method is supported).
3. CAMERA_PASSWORD - Must be set if username is.
4. CAMERA_BASE_DIR - Base directory where capturer will save and streamer will look for video segments file.
5. CAMERA_WIDTH - video width in pixels. Defaults to 1280.
6. CAMERA_HEIGHT - video height in pixels. Defaults to 720.
7. CAMERA_FPS - video frame rate. Defaults to 5.
8. CAMERA_SEGMENT_DURATION - individual segment duration in seconds. Defaults to 10.
9. CAMERA_MAX_BACKLOG_TOTAL_SEGMENT_DURATION - maximal total duration of segments in the backlog. Defaults to 43200 (12 hours, approx. 1.5 GB of video files). Set to 0 keep all files in the backlog (dangerous, can run out of storage).
10. CAMERA_SEGMENT_EXTENSIONS - list of segment file extensions to look for in base directory. Defaults to '.ts .mp4'