A Consensus Based Method for Tracking Modelling Background S(15)

Modelling of the background (“uninteresting parts of the scene”), and of the foreground, play important roles in the tasks of visual detection and tracking of objects. This paper presents an effective and adaptive background modelling method for detectin

example. In the image sequence “Moved Object” (MO) in the Wallflower dataset, a person entered into a room, moved the chair and sat in the chair. While he made a phone call, he turned around in the chair. There is an overlapped region around the center of his body that remains about the same brightness and color despite the underlying movement. The TOM values of the pixels of that region keep increasing because the pixels keep being marked as foreground pixels at each frame. Thus, when the TOM values are increased to be larger than TTM (i.e., after a long time), these pixels are added to the background model. To tackle this issue, we must update the background samples at both the pixel level and the blob level (see next section).

2.3.5.2 Updating Background at Blob Level

Connected groups of foreground pixels below a certain size have their TOM updated "at the pixel level" as described in the previous section. Groups of pixels of size above this threshold are updated "at Blob Level" as we now describe.

First we decide whether a blob is "static". We judge if an object is moving or static by two criteria: the center of the object and the number of the pixels of the object. If either changes (compared with the values of the nearest blob in the previous frame) by a large amount, we judge that the object is moving (otherwise it is static - “static foreground”).

Note: an object which belongs to the background (such as the telephone, and the chairs in Figure

4) that then moves (“moved background”) is not distinguished from a foreground moving object (both the "hole" it leaves and the new region it occupies will typically be treated as foreground objects that have "moved"). For a blob (a large connected region) the TOM values of the pixels in the blob are updated as follows:

TOMt(m')=TOMt 1(m')+1if isstatic m'∈ m'∈ (8) otherwise TOMt(m')=0 m'∈

15