Feature Suggestion
  • Motion Detection settings are tricky. Trading off false positives against false negatives. And with three variables to control (mask, sensitivity, duration) its more of an art than a science.

    Here is a suggestion to give the user more control. First, there should be an index of "probability of important motion detected" that is some function of the intenstity and duration of motion. For example if there is a high intensity but short duration event then that gets some score and it might be the same score as a low-intensity but high duration event. Of course the higher the intensity and the higher duration the higher the score should be. Given that here are two things that could be done with it.

    1. Just as now, all events whose intenstity/duration exceed the user's specifications trigger a recorded motion event and are offered to the user as events to pay attention to. However, events that are close but below the threshold are also recorded but hidden. The user has the option to look at these "near misses" when he wants. This allows him to get a better sense of the false negatives his specifications have set and fine tune them. Plus sometimes when there is extra suspicion he will want to see those near misses anyway.

    2. In continuous captures, imagine a plot of the score along a timeline. Higher scores mean events that the user is more likely to want to look at. He can then jump to those points in the continuous capture that have high scores.
  • I like the idea of #2. With all the usual issues of outdoor cameras, there is no combination of masking/sensitivity that gives me confidence in motion capture, so I record it all. Even simply having the option to flag points of the continuous capture timeline where a threshold of motion is detected would be useful.

    With respect to #1, it would be nice to be able to select a few ranges of existing video and teach the software what type(s) of motion should rank the highest for a particular camera. Still, I would just use that as a ranking score for marking a continuous capture.

    Maybe when there are (affordable) options for security cameras that can accurately sense distance we'll see a jump in the effectiveness of software's ability to more usefully differentiate between different events.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!