Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models
Human detection in videos plays an important role in various real life applications. Most of traditional approaches depend on utilizing handcrafted features which are problem-dependent and optimal for specific tasks. Moreover, they are highly susceptible to dynamical events such as illumination chan...
Saved in:
Main Authors: | , , |
---|---|
Format: | Article |
Published: |
Hindawi
2018
|
Subjects: | |
Online Access: | http://eprints.um.edu.my/22582/ https://doi.org/10.1155/2018/1639561 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
id |
my.um.eprints.22582 |
---|---|
record_format |
eprints |
spelling |
my.um.eprints.225822019-09-26T06:31:16Z http://eprints.um.edu.my/22582/ Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models AlDahoul, Nouar Sabri, Aznul Qalid Md Mansoor, Ali Mohammed QA75 Electronic computers. Computer science Human detection in videos plays an important role in various real life applications. Most of traditional approaches depend on utilizing handcrafted features which are problem-dependent and optimal for specific tasks. Moreover, they are highly susceptible to dynamical events such as illumination changes, camera jitter, and variations in object sizes. On the other hand, the proposed feature learning approaches are cheaper and easier because highly abstract and discriminative features can be produced automatically without the need of expert knowledge. In this paper, we utilize automatic feature learning methods which combine optical flow and three different deep models (i.e., supervised convolutional neural network (S-CNN), pretrained CNN feature extractor, and hierarchical extreme learning machine) for human detection in videos captured using a nonstatic camera on an aerial platform with varying altitudes. The models are trained and tested on the publicly available and highly challenging UCF-ARG aerial dataset. The comparison between these models in terms of training, testing accuracy, and learning speed is analyzed. The performance evaluation considers five human actions (digging, waving, throwing, walking, and running). Experimental results demonstrated that the proposed methods are successful for human detection task. Pretrained CNN produces an average accuracy of 98.09%. S-CNN produces an average accuracy of 95.6% with soft-max and 91.7% with Support Vector Machines (SVM). H-ELM has an average accuracy of 95.9%. Using a normal Central Processing Unit (CPU), H-ELM's training time takes 445 seconds. Learning in S-CNN takes 770 seconds with a high performance Graphical Processing Unit (GPU). Hindawi 2018 Article PeerReviewed AlDahoul, Nouar and Sabri, Aznul Qalid Md and Mansoor, Ali Mohammed (2018) Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models. Computational Intelligence and Neuroscience, 2018. pp. 1-14. ISSN 1687-5265 https://doi.org/10.1155/2018/1639561 doi:10.1155/2018/1639561 |
institution |
Universiti Malaya |
building |
UM Library |
collection |
Institutional Repository |
continent |
Asia |
country |
Malaysia |
content_provider |
Universiti Malaya |
content_source |
UM Research Repository |
url_provider |
http://eprints.um.edu.my/ |
topic |
QA75 Electronic computers. Computer science |
spellingShingle |
QA75 Electronic computers. Computer science AlDahoul, Nouar Sabri, Aznul Qalid Md Mansoor, Ali Mohammed Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models |
description |
Human detection in videos plays an important role in various real life applications. Most of traditional approaches depend on utilizing handcrafted features which are problem-dependent and optimal for specific tasks. Moreover, they are highly susceptible to dynamical events such as illumination changes, camera jitter, and variations in object sizes. On the other hand, the proposed feature learning approaches are cheaper and easier because highly abstract and discriminative features can be produced automatically without the need of expert knowledge. In this paper, we utilize automatic feature learning methods which combine optical flow and three different deep models (i.e., supervised convolutional neural network (S-CNN), pretrained CNN feature extractor, and hierarchical extreme learning machine) for human detection in videos captured using a nonstatic camera on an aerial platform with varying altitudes. The models are trained and tested on the publicly available and highly challenging UCF-ARG aerial dataset. The comparison between these models in terms of training, testing accuracy, and learning speed is analyzed. The performance evaluation considers five human actions (digging, waving, throwing, walking, and running). Experimental results demonstrated that the proposed methods are successful for human detection task. Pretrained CNN produces an average accuracy of 98.09%. S-CNN produces an average accuracy of 95.6% with soft-max and 91.7% with Support Vector Machines (SVM). H-ELM has an average accuracy of 95.9%. Using a normal Central Processing Unit (CPU), H-ELM's training time takes 445 seconds. Learning in S-CNN takes 770 seconds with a high performance Graphical Processing Unit (GPU). |
format |
Article |
author |
AlDahoul, Nouar Sabri, Aznul Qalid Md Mansoor, Ali Mohammed |
author_facet |
AlDahoul, Nouar Sabri, Aznul Qalid Md Mansoor, Ali Mohammed |
author_sort |
AlDahoul, Nouar |
title |
Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models |
title_short |
Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models |
title_full |
Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models |
title_fullStr |
Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models |
title_full_unstemmed |
Real-Time Human Detection for Aerial Captured Video Sequences via Deep Models |
title_sort |
real-time human detection for aerial captured video sequences via deep models |
publisher |
Hindawi |
publishDate |
2018 |
url |
http://eprints.um.edu.my/22582/ https://doi.org/10.1155/2018/1639561 |
_version_ |
1646210282586898432 |
score |
13.18916 |