Face detection and face recognition in indoor surveillance system

Abstract

Face detection plays an important role in indoor surveillance systems. Although there have been many improvements since the 90s, many problems are still unsolved due to the complexity of background, complex computation, long time execution, illumination changes, etc. This paper presents a method to detect faces in a surveillance system under real-world indoor condition. The proposed method combines Background Subtraction Model Feature Analysis approach and Image base approach to localize faces. Many experiments have been conducted with several indoor datasets to evaluate the performance and accuracy of the proposed methods. The result achieves a detection rate of 97.072% and processing speed more than 30 frames per seconds in average. Keywords: face detection, indoor surveillance system, background subtraction, local binary pattern cascade classifier, single shot multi box detection (SSD) face detector, face recognition.