Foreground-Region-Selection Algorithm for Detecting Moving Objects in Dynamic Background

Detection of moving objects in video sequences is the first relevant step of information extraction in many computer vision applications. The undesired background image can be filtered out and the complete foreground image can be retained. By doing this, it provides a focus for tracking, recognition, classification, and activity analysis, making these later steps more efficient. In this paper, we build an adaptive background model using a self-organizing neural network; this model can handle scenes containing moving backgrounds, gradual illumination variations, and shadows cast by moving objects; further, this model has no bootstrapping limitation. However, background subtraction leads to a serious camouflage problem. Owing to this phenomenon, we propose a foreground-region-selection algorithm that combines the image space information and initial object mask generated from improved watershed algorithm and background subtraction respectively. The camouflage problem can be effectively solved using the proposed algorithm. The detection results of the proposed algorithm are better than the background subtraction results obtained from the experiments.