Security Camera

Records and encodes video to HEVC/MP4 using hardware acceleration, uploads to Azure Blob Storage, detects faces and sends notifications.

Usage

Create Azure Storage account
Set in configuration
Run SecurityCamera.Console.exe.

Configuration

Configuration is stored in appsettings.json or can be added as CLI arguments.

Recording

Video
- Format: "HEVC" or "MP4" (default "HEVC")
- Quality: "HD720p", "HD1080p", "Uhd2160p" or any member of VideoEncodingQuality enum (default "HD1080p")
- Bitrate: the average bit rate of the video stream, in bits per second
- FrameRate: the number of frames per second
Audio
- Enabled: whether to enable audio recording (default false)
- Bitrate: the average bit rate of the audio stream, in bits per second
- BitsPerSample: the number of bits per audio sample
Recording
- ChunkSize: size of video chunks (default "00:10:00")
- MaximumRecordTime: maximum of total record time, shut down after that time passed (default null)
Blobs
- ConnectionString: the connection string for the Storage Account
- ContainerName: the name of the container to upload blobs (default "recordings")
- BufferSize: the size of the buffer to write the blob (default null, defaults to library default which is 4 MiB)
- BlobNameFormat: format used for naming blobs, with argument DateTimeOffset at index 0 (default "{0:yyyy/MM/dd/HH-mm-ss}.mp4")
- InitialSizeHint: initial size of Page Blob in bytes (default: 64 MiB)
- ResizeFactor: the factor to resize a Page Blob when its size limit is exceeded (default 2.0)
Wake
- Enabled: keep computer from sleep during recording (default: true)

Alerts

FaceDetection
- Enabled: whether to enable face detection (default true)
- DetectionMode: any member of FaceDetectionMode enum (default "HighPerformance")
- DesiredDetectionInterval: the time span for which face detection should be performed (default "00:00:00.500")
Sms
- ConnectionString: connection string for Azure Communication Services
- From: from number
- To: array of to numbers
WebHook
- FaceDetectionUrl: url to call
- FocusChangeUrl: url to call

Telemetry

ApplicationInsights: configure Application Insights

Web hooks

Face detection

The URL set in WebHook.FaceDetectionUrl is invoked with POST and content type application/json:

{
    "detectedFaces": [
        {
            "faceBox": {
                "X": 123,
                "Y": 456,
                "Width": 240,
                "Height": 240,
            }
        }
    ]
}

Focus change

The URL set in WebHook.FocusChangeUrl is invoked with POST and content type application/json:

{
    "state": "Focused"
}

Tutorial

How to record video

First, we need to create a MediaEncodingProfile:

var profile = MediaEncodingProfile.CreateHevc(VideoEncodingQuality.HD1080p);

Then, we have to initialize a MediaCapture.

using var capture = new MediaCapture();
await capture.InitializeAsync();

This is where we could select the camera or microphone we would like to use. If we don't specify any, it is going to use the default settings.

The next step is to start and stop recording:

await capture.StartRecordToStreamAsync(profile, stream);

// capture video ...

await capture.StopRecordWithResultAsync();

How to detect faces

For photos, there the Windows.Media.FaceAnalysis namespace contains a FaceDetector class:

var faces = await FaceDetector.DetectFacesAsync(bitmap);

For videos, the FaceTracker is the right choice, as it was optimized to work with VideoFrames:

var faces = await FaceTracker.ProcessNextFrameAsync(videoFrame);

To process a video in real-time and pass VideoFrames to the FaceTracker, we have multiple choices:

Grab a preview using GetPreviewFrameAsync periodically, and pass the VideoFrame for processing.
Implement a custom sink and process video using StartRecordToCustomSinkAsync.
Implement an IBasicVideoEffect and listen to VideoFrames in its ProcessFrame method.

Video processing and face detection is not easy at all, especially if we want to be efficient as well, even offload some of the work to the GPU. Fortunately, Microsoft has already implemented this functionaly as FaceDetectionEffect, which we can easily add to the MediaCapture:

var definition = new FaceDetectionEffectDefinition
{
    DetectionMode = FaceDetectionMode.HighPerformance,
};
var effect = await capture.AddVideoEffectAsync(definition, MediaStreamType.VideoRecord);
effect.FaceDetected += OnFaceDetected; // subscribe to event
effect.Enabled = true;

Stream to Azure Blob Storage

We can open a Page Blob or a Block Blob to write as a Stream, so this way we can transfer the video seamlessly to Azure without buffering to disk or memory:

var container = BlobServiceClient.GetBlobContainerClient(blobsOptions.ContainerName);
var blob = container.GetBlockBlobClient($"{now:yyyy/MM/dd/HH-mm-ss}.mp4");
await using var stream = await blob.OpenWriteAsync(overwrite: false);

In order to use a Stream in Windows Runtime, we have to it wrap into a IRandomAccessStream :

using var randomAccessStream = stream.AsRandomAccessStream();

Howewer, some media encodings are not streaming compatible, because they require seeking, which is not supported by the Azure Storage library, so the case of streaming is not that straight-forward unfortunately. An implementation which works with Page Blobs, can seek and replace pages of the blob on-demand, can be found in PageBlobRandomAccessStream.

How to keep computer from sleep

Normally, a device running a Windows Runtime app will dim the display (and eventually turn it off) to save battery life when the user is away, but video apps need to keep the screen on so the user can see the video. The DisplayRequest class lets you tell Windows to keep the display turned on so the recording can continue.

var displayRequest = new DisplayRequest();
displayRequest.RequestActive();

And then release:

displayRequest.RequestRelease();

How to send SMS

You can use Azure Communication Services to send SMS:

var client = new SmsClient();
await client.SendAsync(from, to, message);

Known issues

Wake: DisplayRequest requires STA
Recording: recording is paused while replacing blob chunk

Disclaimer

This project is only for learning purposes. The author doesn't take any responsibility for this software. Do not use this solution for real-life use cases.

Peter-Juhasz/security-cam