Recognition Of Face Detection System Based On Video

Mohammed, Sari Abdo Ali (2015) Recognition Of Face Detection System Based On Video. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Recognition Of Face Detection System Based On Video.pdf - Submitted Version Download (492kB)

Abstract

Nowadays, face recognition is an important aspect for security issues and identification. A lot of efforts were done by researches in the field of face recognition and detection. In this project of face recognition, the project divided into several stages, face detection, method of recognition and the database. Face recognition is implemented using visual studio 2010, OpenCV library and Emgu library. In order to link visual studio with OpenCV library there are some configurations has to be set before proceed with the coding process. There are three main aspects in this project; the first aspect is to develop an algorithm to detect frontal face. The second aspect is to develop an algorithm that recognizes a person identity on a video. The last aspect is to analyze the effects of illumination changes on the recognition. Face detection is the process of identify faces on video or on image and differentiate it from the objects on the background. Face detection is done using Haarcascade algorithm. Haarcascade classifier is trained using five hundred positive and negative images from internet. In order to get high rate of recognition, the image of the detected face has to be processed where it is converted from colored image (with green, blue and red colors) to grayscale image to reduce the data during the process. The image after that cropped and save the face only to reduce the noise of the background. Speeded up robust feature (SURF) is used to extract the facial features in order to match the images for faster recognition. Emgu library with c# is used to do the recognition stage. After the finishing the programing process and mount the hardware, the system should detect frontal faces and find the identity of people that have been detected as well as recognize the identity of the detected face.

Actions (login required)

View Item