Efficient Scale Invariant Human Detection Using Histogram of Oriented Gradients for IoT Services

Abstract

Recent advancements in computer vision, multimedia and Internet of Things (IoT) have shown that human detection methods are useful for applications of intelligent transportation system in smart environment. However, detection of a human in real world remains a challenging problem. Histogram of oriented gradients (HOG) based human detection gives an emphasis towards finding an effective solution to the problems of significant changes in view point, fixed resolution human but are expensive to compute. The proposed algorithm aims to reduce the computations using approximation methods and adapts for varying scale. The features are modeled at different scales for training the classifier. Experiments have been conducted on human datasets to demonstrate the superior performance of the proposed approach in human detection and discussions are made to integrate and increase personalization for building smart environment using IoT.