Object detection in dynamic background for visual surveillance applications

In computer vision applications, main goal is to detect object of interest that is often moving object in video frames. The proposed model is performed in two stages: training and testing. In the training phase, a background model is developed with initial few frames. In the testing phase, foreground is detected with improved thresholding scheme. In this work, dependency of using fixed threshold as used in considered literature has been avoided and main contribution towards a standard deviation based threshold is automatically evaluated during run-time and misclassification has been handled by using morphological filters in order to improve detection quality. The major strength of this work is that it is robust to the environmental changes and motion in the background. The proposed model reduces false detections and enhances pixel classification accuracy as depicted in experimental analysis.