evilpaul / canoecamera Goto Github PK

video and snapshot capabilities are not retrieved due to an exception generated when attempting to interrogate the device at

After adding the webcam node into a CANoe project the memory usage of RuntimeKernel.exe can be seen constantly increasing until some garbage collection is triggered, then repeats.

This does not occur if the NewFrame callback contains now functionality. Consider how the image data is used from the callback function onwards and ensure memory is cleared when no longer required.

Current implementation only allows decimal frame rates to be used.

Backport implementation from Accord.NET in order to allow setting fractional frame rates (e.g. 29.97)

Deprecated APIs are currently used that has now been removed in FFMPEG 5. these legacy API calls need to be replaced before FFMPEG versions later than 4.X can be used.

If your camera has auto-exposure, or a long manual exposure, then you will not get new frames at the specified rate.

Currently each new frame is written to the output video when received; fewer frames due to a longer exposure will mean the time in the output video will speed up.

The timestamp added to the video is assumed based upon

1. Measurement time of initial request
2. Incremented timestamp based upon frame rate

Modify the behaviour so that the timestamp reflects the actual measurement time when that frame was received. (may have relation to #3)

Update to support FFMPEG 4.0

Due to the CANoe concurrency architecture everything looks to be run on a single core.

What can be done to move the high image processing load onto a separate core?

The saved videos display normally when played in VLC (and windows media player, etc.), but when jumping to specific timestamps or dragging the playback bar the displayed image is garbage until the next keyframe.

Investigate if any settings can be changed when saving a video. (usage of FFMPEG)

This may be related to #2, but could be related to other factors (e.g. camera frame rate, frame size, storage speed).

When requesting the start of a video or to save an image there looks to be some delay before the operation actually occurs.

Due to the disconnect of using sysvars as triggers the operations are not synchronous so where is the timing difference occurring?

The NewFrame callback is run from a different thread so calls to retrieve the CANoe measurement time throw exceptions. Another means needs to be made to monitor the timings of requests and actions.

It would be nice to be able to set specific CameraControlProperties (i.e. fixed focus/exposure) in order to ensure consistent camera behaviour between uses/PCs.