Biometric Security System for Watchlist Surveillance

Video-Surveillance is a wide area of research and various researchers doing research in this field from the last decade. In our approach we used one of the reliable biometric of a human i.e. face. Here we are using well-known PCA (Principal Component Analysis) Algorithm for the training process of the system because it is one of the best dimensionality reduction and one of the reliable algorithms. The weakness of current algorithm is that when this algorithm is used alone for a surveillance purpose it does not give best results and when it combines with some other techniques which provide good results with good speed and accuracy. Now a day most researchers are doing research with single image only. In our approach we are having only one image per person in a dataset PCA alone does not work well with that single image. Our approach is to use PCA with NN (Nearest Neighbors) and having confidence value that is favorable to use with NN replacing threshold value concept of PCA. For a single image we need to set a specific threshold value depending upon the environment for recognition. And it is itself a very challenging task.