PERFORMANCE ANALYSIS WITH DIFFERENTIAL EVOLUTION OPTIMIZATION FOR MOVING OBJECT DETECTION USING BACK GROUND SUBTRACTION

Detecting and recognizing objects automatically in digital videos is a major challenge in video analysis. As part of this paper, we faced the difficult problem of segmenting objects in videos with constantly moving backgrounds. These are situations that occur when, for example, you film rivers, the sky, or a scene containing smoke, rain, etc. This is a subject little studied in the literature because very often the scenes treated are rather static and only a few parts move, for example the leaves because of the wind. Another source of movement relates to changes in light. The main difficulty, in the case of scenes with a moving background, is to be able to differentiate the movements of the object from those of the background which can sometimes be very similar. Indeed, for example, an object in a river can move at the same rate as water. The algorithms of the literature extracting fields of displacement then fail and those based on background modelings generate very many errors. It is therefore within this complicated framework that we have tried to provide solutions.