Please use this identifier to cite or link to this item:
Title: Extreme learning machine (ELM) methods for pedestrian detection
Authors: Song, Qiaozhi
Keywords: DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems
DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems
Issue Date: 2016
Abstract: As a significant part of computer vision, pedestrian detection is a popular filed of research. The application of surveillance videos also becomes attractive due to the increasing safety concerns. In this project, an extreme learning machine based pedestrian detector is built for surveillance purpose. The project comprises of two main parts, which are feature extraction algorithm, and implementation of machine learning methods. The feature extraction process provides the description of the pedestrians for further processing, including background subtraction and Histograms of Oriented Gradients (HOG) features extraction. HOG as one of the most well-known and effective object detection, outperforms many existing features for pedestrian detection as well. Background subtraction serves as an add-on to implementation of HOG in videos, and it is proven that the background subtraction largely improves the efficiency of the system. It is still a very new idea to apply the state-of-art extreme learning machine (ELM) method to pedestrian detection. For better evaluation of the system, the learning methods employed is not only the ELM, but also two versions of support vector machine (SVM). It is shown that ELM achieves best testing accuracy and training time among the machine learning techniques. In this project, the overall pedestrian detection system is successfully designed and implemented in Matlab. To achieve the maximum operation speed and verify the applicability of the system, a modified system is also successfully realized on C++ platform.
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
  Restricted Access
Final Year Report1.84 MBAdobe PDFView/Open

Page view(s) 50

checked on Sep 28, 2020

Download(s) 50

checked on Sep 28, 2020

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.