motion detection during encoding, not capturing

[cpc1225]

Hi, I am having some problem require assistance and hope some expert here will help me out.
I have to scan through 9 hour of video recorded from my surveillance camera every night, I wonder is there any way to filter out the portion without motion with encoding software.
I know there are some software will only records base on motion detection, but my capture card only works with it’s own software and the function is very limited.
Can any one help?

[Mug Funky]

a few of the VfW codecs have a frame-drop threshold in them. there are also programs that can strip out the dropped frames, so you'll end up with a short video with only the bits with large enough differences.

also, RV and WMV have dropdupe style filters for the same thing

[cpc1225]

Thanks for the tips, but I could not understand those terms (RV, WMV, VFM).
Can you provide explanation in layman's term ( I thought I am good in this, but never come across those names ).

[Mug Funky]

arg, sorry... i always scold other people for talking in acronyms...

VfW = Video for Windows. this means basically anything that goes inside an avi file. virtualdub uses this interface. it has problems with things like b-frames, used by the more advanced codecs, but there are hacks to work around this (and dropped frames + bframes usually don't mix too well, so you should probably just avoid b-frames).

WMV = Windows Media Video. i'm not a fan of it (i'm a left-wing pinko, so i don't like Microsoft), but it is a pretty good codec all the same.

RV = RealVideo. i'd probably rate this codec higher than WMV, both technically and politically, but not by much . it's made by RealNetworks, which if i recall correctly was founded by a guy who broke away from Microsoft and decided to compete with them.


now, the layman's terms:

basically, some codecs provide the ability to not encode frames that are insufficiently different from their predecessor. this is a pretty simple way to save space. codecs capable of this will have something like a "frame-drop ratio" or threshold that as you increase it, it will more agressively throw out frames.

codecs like RealVideo have a "dropdupe" filter that is run before the encoder. i've never used RV for anything but playback, so i couldn't give you a good explanation of how to do this (try Helix Producer, or one of the frontends for it like Sirber's RealAnime, and it should be easy).

though now i think of it, the dupe calculations might be thrown off if there's a timecode displayed on the footage (i'm guessing there is one), in that everytime the timecode changes the codec sees a new frame, and keeps it.

there might be a way to avoid this using Matroska. you'll have to read up on this, but it uses VFR (Variable Frame Rate), like Real and Windows Media do. the advantage on matroska is you can use any codec you like, because the frame drops are done in the program stream, not the actual video stream.

i could imagine you generating the matroska VFR timecode file with a video that has the timestamps cropped out, then doing your final encode using the timecode file, but leaving the timestamps in this time. that way, they wont be considered for frame drop decisions, but they will be left in so you can still use them.

hehe.. without a sample video i couldn't give you more information than that. i hope it all makes sense, as there's frustratingly 2 types of timecodes i'm talking about in the same paragraph