Face retrieval system based on elastic web crawler over cloud computing

Abstract

The web pages are considered as the main source of the available and provided information that is characterized by variation in its content. The facial recognition plays a key role in knowledge management and identity authentication systems. Although the rapid advance of the web technologies and face recognition systems, the improvement of real-time performance is still the bottleneck. The main objective of this study is to propose a real-time face retrieval system as a service over cloud computing based on a web face crawler. The proposed architecture ensures that the total response time is reduced and the resource utilization is optimized. The web crawlers fetch web pages and extract images in elastic storage over the cloud. Then the collected images are used to extract human faces and to prepare the faces images by succeeding phases to be ready for recognition and identifying the matched face of the collection. The proposed service depends on Principle Component Analysis (PCA) algorithm for feature extraction and dimensionality reduction. Furthermore, K-Nearest Neighbors (KNN) is used to classify the crawled facial images over cloud resources. The experimental results investigated that an enhancement of crawling speed is achieved by increasing the crawler instances. Moreover, the accuracy is enhanced in the face recognition based on the Euclidean over other metrics such as Manhattan and Cosine dissimilarity.