Human Interaction Anticipation by Combining Deep Features and Transformed Optical Flow Components

Abstract

The anticipation of ongoing human interactions is not only highly dynamic and challenging problem but extremely crucial in applications such as remote monitoring, video surveillance, human-robot interaction, anti-terrorists and anti-crime securities. In this work, we address the problem of anticipating the interactions between people monitored by single as well as multiple camera views. To this end, we propose a novel approach that integrates Deep Features with novel hand-crafted features, namely Transformed Optical Flow Components (TOFCs). In order to validate the performance of the proposed approach, we have tested the proposed approach in real outdoor environments, captured using single as well as multiple cameras, having shadow and illumination variations as well as cluttered backgrounds. The results of the proposed approach are also compared with the state-of-the-art approaches. The experimental results show that the proposed approach is promising to anticipate real human interactions.

Keywords

• Human interaction anticipation
• video surveillance
• deep learning
• transformed optical flow