Background Removal Using RGB-D Data for Fright Recognition

Abstract

This research paper evaluates the automatic recognition of various human activities while moving inside a lobby using depth data for background removal and using the surrounding walls as reference lines of frame. The study examines detection of panic and fear as the subject of focus in the process of activity recognition. Many times, humans escape from danger and threat, try to evade, run to a secure spot because of fear caused by ambush, violence, presence of life threatening incidents such as shooting or because of natural calamity such as cyclone, hurricane, flash floods, earthquake and tornado. The research first uses a sequence of image to extract the human blob, shape form using image filtering. A background to foreground subtraction approach is taken to eliminate non-interesting regions. The blob, shape form is then normalized using a reference frame of lobby, room, followed by edge enhancements and then features are extracted by repeated application of a mesh with varying thresholds. Finally, a support vector machine (SVM) classifier was used to detect activity that represented fright. The results showed recognition accuracy of 74.8% for continuous, automatic, real time unconstrained video image series.

Authors and Affiliations

Amol S Patwardhan

Keywords

Related Articles

Performance Metrics and Analysis for Node Clone Detection Schemes in Wireless Sensor Networks

Wireless sensor networks are now-a-days rapidly developing in the field of science and technology. And it promises to have a large number of applications in next generation networks (NGN). The wireless sensor networks c...

slugComparison between the best online suites

The battle to be the best online suite in the cloud has never been more hotly contested. It’s not as straightforward as it seems either. This paper gives the introduction to Office Software Suite, core feature, cloud i...

Literature Survey on Various Software Cost Estimation Models

The main challenge in the development of large and complex projects is the cost estimation with more accuracy. Many estimation models are introduced in the course of time, which concludes that software cost estimation i...

Optimization and Vibration Analysis of Compressor Blade with Different Blade Angle

The effective uses of a Compressor blade are limited at its maximum operational junction frequency. The study was conducted by using the Finite element method. The compressor blades are used with flow of with rotation s...

A Review on Secure Routing in Wireless Sensor Networks

Sensor nodes are tiny, low-cost devices equipped with environment sensors and radio for wireless communication. These sensor nodes may constitute the network for monitoring physical phenomena. Such network is called Wir...

Download PDF file
  • EP ID EP22601
  • DOI -
  • Views 260
  • Downloads 4

How To Cite

Amol S Patwardhan (2016). Background Removal Using RGB-D Data for Fright Recognition. International Journal for Research in Applied Science and Engineering Technology (IJRASET), 4(9), -. https://europub.co.uk/articles/-A-22601