Does someone need motion detection feature?

[ImageMiner]

Hi,

I'm really new to this forum and I would like to know, if there are more people interested in a motion detection feature for vlc.
The integrated video filter didn't fit my needs, so I coded a new motion detection video filter, which works fine for me.
I'm using it as a traffic counter and car tracking system to get traffic statistics.
Acually it is not well integrated into my vlc version and I'm thinking about to design more interface features like parameters and may be export of data.
I know, that vlc is a media player. So it can be dangerous to integrate such a feature. I know of some linux projects, which handle video survelliance in generell ... so they are more powerfull in visualization and data collection tasks.
I would like to aim a small, robust motion detection feature, which can trigger some simple events, that's all.

... and if there are more vlc users, who are interested in a motion detection feature .... I would keep developing this part of code and provide it to all vlc users.

Regards

ImageMiner

PS: May be I'm blind and there exists a projekt with these objectives already, so you can ignore my request!

[Jean-Baptiste Kempf]

Yes, I believe some people are interested.

But, did you try the vlc motion detection video filter?

[ImageMiner]

Hi j-b,

yes, I used the source code of motiondetect.c as training material to understand the mechanisms within vlc. As I have said I'm new to vlc development.
I read in another thread, that the motion detection video filter was just a quick implementation to emulate a so called "minority report effect ".
Thats why i called my idea to use vlc as motion detection software to be dangerous, because it is a player, not a survelliance station.
In the meantime I found another way to do video filtering -> opencv (opencv_wrapper.c), but I'm not sure, that this way will lead me to a robust, simple and fast implementation. I plan to use the motion detection feature in a small hardware environment.
And secondly I'm optimizing the video processing algorithms on two simple motion detection tasks. This strategy will have the benefit, that I don't have to consider about a wide set of generalization properties.

My next step will be the correct integration into vlc.

Regards

ImageMiner

[adler]

Hi, I'm also interested in getting opencv working with vlc. After reading about the sorry state of affairs when it comes to getting the motion detector video filter to work, I am pursuing a different solution.

I have just written a previous QT app (nothing to with vlc) that uses the QT framework's built-in plugin manager to dynamically load .so/.dll files that perform my video analysis. When paintEvent() is called for the video frame's widget, I pass a pointer for the image buffer to the plugin instance which does all of the opencv-related stuff (in my case, facedetect) before painting the frame.

Now for various reasons I need vlc to do the capturing from my webcam (as opposed to using opencv's cvQueryFrame()). After looking at the src for VLC I thought of bypassing vlc's plugin/filter system and using QT's plugin manager to load my face detect modules and running them just before the frame for the video widget gets painted. Just like my previous app.

Before I get started on this, I see from your post that you have had some success with getting an opencv video filter to work. Was this using opencv 2.0? Was the opencv wrapper a problem? Could you go into a little detail on your experiences in getting it working?

While my approach is one of expediency (no need for any wrappers, get to reuse previous code) it just doesn't feel right. Using QT gui code for calling my opencv related functions is bad architecture. Have you had any more success with your approach?

[ImageMiner]

Hi adler,

sorry, but I didn't have integrated opencv into vlc successfully. I wasn't able to incorperate opencv via wraper.
Btw actually in the last weeks, I wasn't able to continue the development. I hope I will do some work on it in the next
weeks.
As I have written in my post before, the next step wil be the correct integration into vlc code structure.

You wrote, that you are doing face detection. It's obviously a more power consuming task.
If the algorithm for feature extraction is not so complicated, I could integrate it into the plugin.
Am I right, that the algorithm have to extract a feature vector and then compares it against a feature vector database?
There are several algorithms in this sector. Did you use a special one for your purpose or is it a general mechanism?

Regards

ImageMiner

[perryjameswarren]

I am putting together a whole house automation project. I have the various audio video pieces put together and even worked out how to get Hulu, Netflix and Amazon pushed out to the house and remote devices via the internet. I have a few test security cameras and mics up but I would love this system to be single entry point if possible. Next steps (dare I say it) are to interface in some lighting controls and hvac systems. I have found some nice security packages for for motion detection (ping if you want more info) and would love to see some further development in this area.