An occlusion-robust feature selection framework in pedestrian detection

Zhixin Guo, Wenzhi Liao, Yifan Xiao, Peter Veelaert, Wilfried Philips

Research output: Contribution to journalArticle

2 Downloads (Pure)

Abstract

Better features have been driving the progress of pedestrian detection over the past years. However, as features become richer and higher dimensional, noise and redundancy in the feature sets become bigger problems. These problems slow down learning and can even reduce the performance of the learned model. Current solutions typically exploit dimension reduction techniques. In this paper, we propose a simple but effective feature selection framework for pedestrian detection. Moreover, we introduce occluded pedestrian samples into the training process and combine it with a new feature selection criterion, which enables improved performances for occlusion handling problems. Experimental results on the Caltech Pedestrian dataset demonstrate the efficiency of our method over the state-of-art methods, especially for the occluded pedestrians.
Original languageEnglish
Article number2272
Number of pages18
JournalSensors
Volume18
Issue number7
DOIs
Publication statusPublished - 13 Jul 2018

Fingerprint

occlusion
Feature extraction
redundancy
learning
Redundancy
education
Patient Selection
Noise
Pedestrians
Learning
Efficiency

Keywords

  • pedestrian detection
  • feature selection
  • occlusion handling
  • deep learning

Cite this

Guo, Zhixin ; Liao, Wenzhi ; Xiao, Yifan ; Veelaert, Peter ; Philips, Wilfried. / An occlusion-robust feature selection framework in pedestrian detection. In: Sensors. 2018 ; Vol. 18, No. 7.
@article{bff90cb6a4294349b7015c3051b32906,
title = "An occlusion-robust feature selection framework in pedestrian detection",
abstract = "Better features have been driving the progress of pedestrian detection over the past years. However, as features become richer and higher dimensional, noise and redundancy in the feature sets become bigger problems. These problems slow down learning and can even reduce the performance of the learned model. Current solutions typically exploit dimension reduction techniques. In this paper, we propose a simple but effective feature selection framework for pedestrian detection. Moreover, we introduce occluded pedestrian samples into the training process and combine it with a new feature selection criterion, which enables improved performances for occlusion handling problems. Experimental results on the Caltech Pedestrian dataset demonstrate the efficiency of our method over the state-of-art methods, especially for the occluded pedestrians.",
keywords = "pedestrian detection, feature selection, occlusion handling, deep learning",
author = "Zhixin Guo and Wenzhi Liao and Yifan Xiao and Peter Veelaert and Wilfried Philips",
year = "2018",
month = "7",
day = "13",
doi = "10.3390/s18072272",
language = "English",
volume = "18",
journal = "Sensors",
issn = "1424-8220",
number = "7",

}

An occlusion-robust feature selection framework in pedestrian detection. / Guo, Zhixin; Liao, Wenzhi; Xiao, Yifan; Veelaert, Peter; Philips, Wilfried.

In: Sensors, Vol. 18, No. 7, 2272, 13.07.2018.

Research output: Contribution to journalArticle

TY - JOUR

T1 - An occlusion-robust feature selection framework in pedestrian detection

AU - Guo, Zhixin

AU - Liao, Wenzhi

AU - Xiao, Yifan

AU - Veelaert, Peter

AU - Philips, Wilfried

PY - 2018/7/13

Y1 - 2018/7/13

N2 - Better features have been driving the progress of pedestrian detection over the past years. However, as features become richer and higher dimensional, noise and redundancy in the feature sets become bigger problems. These problems slow down learning and can even reduce the performance of the learned model. Current solutions typically exploit dimension reduction techniques. In this paper, we propose a simple but effective feature selection framework for pedestrian detection. Moreover, we introduce occluded pedestrian samples into the training process and combine it with a new feature selection criterion, which enables improved performances for occlusion handling problems. Experimental results on the Caltech Pedestrian dataset demonstrate the efficiency of our method over the state-of-art methods, especially for the occluded pedestrians.

AB - Better features have been driving the progress of pedestrian detection over the past years. However, as features become richer and higher dimensional, noise and redundancy in the feature sets become bigger problems. These problems slow down learning and can even reduce the performance of the learned model. Current solutions typically exploit dimension reduction techniques. In this paper, we propose a simple but effective feature selection framework for pedestrian detection. Moreover, we introduce occluded pedestrian samples into the training process and combine it with a new feature selection criterion, which enables improved performances for occlusion handling problems. Experimental results on the Caltech Pedestrian dataset demonstrate the efficiency of our method over the state-of-art methods, especially for the occluded pedestrians.

KW - pedestrian detection

KW - feature selection

KW - occlusion handling

KW - deep learning

U2 - 10.3390/s18072272

DO - 10.3390/s18072272

M3 - Article

VL - 18

JO - Sensors

JF - Sensors

SN - 1424-8220

IS - 7

M1 - 2272

ER -