FACE DETECTION UNDER OCCLUSION

Abstract

In this thesis we are concentrating on the surveillance application which can be used to identify no of faces in a surveillance area. One of the main area is finding the attendance in the class and for marking the attendance. In this paper where the faces are occluded under varied occlusions can also be detected using the algorithm used in attaining the intended result.

This research represents a robust approach in obtaining the in- tended result. Here for this research I use the Speed up Robust Feature (SURF) algorithm to obtain the intended result. For the specific surveillance application we should have the database of the individuals in a screen. Then the images of persons stored in the database is evaluated against the current surveillance image for personal identification. A thorough search is done by the proposed SURF algorithm to obtain the output.

Features of both the images in the database and the surveillance image is extracted using the said algorithm and the features are matched to identify the matched features, from this matched features depending on the threshold value the output can be obtained. The result shows higher accuracy compared to other methods such as Voila Jones, SIFT, etc.