One of security systems that can be implemented in public and private facilities is through video surveillance. However, conventional surveillance video recordings so far have no ability to recognize the person in the video recording. In this research, a face recognition system is designed as the value-added of the video recordings from IP cameras. The system of face recognition was designed to video surveillance recordings implemented in outdoor area. To detect the presence of human objects in the video, HOG (Histogram of Oriented Gradient) was used. Meanwhile, ICA (Independent Component Analysis) was used for the face recognition feature extraction algorithm. The output of the system was an identity of the object. Tests were conducted on 10 videos with a ratio of recognized and unrecognized face of 50% -50%. In this study, an ideal state of the test results was to detect human presence as a first step, and then the limited frame of the face detection process was carried out. The highest accuracy for detecting the presence of humans was found at 97% for the state of the object within 3-4.5 meters with a camera.

Keywords and phrases: face recognition, video, HOG, ICA.