Adaptive Foreground Object Extraction in Real Time Videos Using Fuzzy C Means with Weber Principle

Abstract

Objectives: We propose a foreground extraction method for video surveillance system is to detect the objects in real time. Methods: The proposed foreground extraction technique models the background using cluster centroids and optimized using fuzzy-c-means technique. The foreground is extracted using background subtraction. The optical flow is used to eliminate the falsely extracted foreground pixels.Findings: Traditional techniques, cluster centroids are initialized using random values or histogram peaks, but in our proposed system the cluster centroids are initialized using weber principle. Improvement: This proposed real-time foreground extraction approach yields better results than the previous algorithms with respect to quality of extraction and memory consumption.