Where W and H stand for the height and the width
of the image respectively. The Movt stands for the size
of the moving object in the video, if the Movt is
bigger, the size of the moving object in the video is
bigger at the point of time t.
We use the gray difference between consecutive
frames in the shot to detect the moving object. What is
more, the sum of the pixels ( Movt ), at which place the
gray difference between consecutive frames is greater
than the threshold of difference (Diff), determines the
size of moving object in the video. Now we should
exactly determine the threshold (Diff), if the threshold
we select is too great, we will miss some really moving
object in our detection. On the contrary, we will
mistake much still object for moving object. It is
proved in our experiment that very accurate detection
can be achieved when the threshold is 30.
When we obtain the sum of difference pixels in
every frame, we can use a curve to describe the change
of the size of moving object according to the time in all
the video. In order to suppress noise influence, we
choose the average value filter to smooth curve
appropriately, here the step of the filter is 3, and then
we can ensure that the missing detection and the wrong
detection can be eliminated efficiently. The filter is
defined as