|
|
| |
| Multi-Face Recognition System in Surveillance Video
Download PDF
|
Ibtisam Mohammed Humaid Al Bahri, Arockiasamy Soosaimanickam , Sallam Osman Fageeri
|
Abstract:
An automatic multi-face recognition system is one of the artificial intelligent systems. It helps to manage enormous number of tasks such as Access control & surveillance systems, monitoring management, and so other detection applications. Multi-face recognition system in surveillance video (MFRS) or automatic multi-face recognition. It is a system that have a capability to detect and read human face from surveillance video automatically and immediately and then sends them to database, afterwards result readable by machines. These kinds of systems have been widely used to recognize human face by using several algorithms and methodologies, including optical dimensions' recognition, convolutional neural network, artificial neural network or deep neural network, morphological operations, and facial features edge detection. This paper aims to discuss the multi-face recognition system using surveillance video. It aims at understanding and analyzing the concept of the human face recognition system, essentially those systems that don't required any human resources support to accomplish their functions and trying to evaluate those human face recognition systems by using real-time generated human faces dataset. For this purpose, this research uses an analytical and theoretical comparison between various previous research works in face recognition field to understand which deep learning algorithm is providing much accurate results. Also, there is a practical evaluation by using real human face images. In this study, there are three different stages of the human face recognition system. Initial operation starts from images aggregation, face detection, human face recognition, and finally identity recognition. In each processing phase, there is a sub preprocessing procedures and deep learning algorithms used. For instance, morphological operations were used to detect and extract face region and using convolutional neural network in face feature recognition level. The preprocessing levels are from the insert human image step until output the recognized data, however deep learning algorithms are distinct from one system framework to another. The performance of recognition operation is evaluated based on different metrics such as: classification accuracy, and precision, and recall.
|
|
|
| | |