Embedded system for real-time person detecting in infrared images/videos using super-resolution and Haar-like feature techniques

Abstract

This paper describes a real time person detection system using near infrared images/videos. This novel system integrates person detection and super resolution algorithms performing person recognition. Additionally, we use a detector Haar-like features and for increasing resolution we use classical algorithms like Nearest neighbor interpolation, Bilineal interpolation and bicubic interpolation. The detector is trained using Adataboost and cascade classifiers and the implementation is performed in the embeded system Raspberry pi2 with the Noir Pi Camera. The implemented embedded system runs at about 20 frames/second.