Citation: Oluyide, O.M.; Tapamo,
J.-R.; Walingo, T.M. Automatic
Dynamic Range Adjustment for
Pedestrian Detection in Thermal
(Infrared) Surveillance Videos.
Sensors 2022, 22, 1728. https://
doi.org/10.3390/s22051728
Academic Editors: Nunzio Cennamo,
Yangquan Chen, Subhas
Mukhopadhyay, M. Jamal Deen,
Junseop Lee and Simone Morais
Received: 3 February 2022
Accepted: 16 February 2022
Published: 23 February 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
Automatic Dynamic Range Adjustment for Pedestrian
Detection in Thermal (Infrared) Surveillance Videos
Oluwakorede Monica Oluyide, Jules-Raymond Tapamo * and Tom Mmbasu Walingo
School of Engineering, University of KwaZulu-Natal, Durban 4041, South Africa;
213554623@stu.ukzn.ac.za (O.M.O.); walingo@ukzn.ac.za (T.M.W.)
* Correspondence: tapamoj@ukzn.ac.za; Tel.: +27-031-260-2751
Abstract:
This paper presents a novel candidate generation algorithm for pedestrian detection in
infrared surveillance videos. The proposed method uses a combination of histogram specification and
iterative histogram partitioning to progressively adjust the dynamic range and efficiently suppress
the background of each video frame. This pairing eliminates the general-purpose nature associated
with histogram partitioning where chosen thresholds, although reasonable, are usually not suitable
for specific purposes. Moreover, as the initial threshold value chosen by histogram partitioning is
sensitive to the shape of the histogram, specifying a uniformly distributed histogram before initial
partitioning provides a stable histogram shape. This ensures that pedestrians are present in the
image at the convergence point of the algorithm. The performance of the method is tested using four
publicly available thermal datasets. Experiments were performed with images from four publicly
available databases. The results show the improvement of the proposed method over thresholding
with minimum-cross entropy, the robustness across images acquired under different conditions, and
the comparable results with other methods in the literature.
Keywords: infrared surveillance; pedestrian detection; video surveillance
1. Introduction
Video surveillance is gaining worldwide prevalence, and it is not uncommon to find
cameras mounted in airports, schools, office buildings, and many private residential areas
in addition to their traditional presence on public and government buildings and large
organisations. This significant increase in interest and utilization of Video Surveillance
Systems (VSSs) outside of the public and security sectors is propelled by user demand
for security due to increasing crime rates, global security threats, and advancement in
technology, which has significantly dropped the cost of acquiring and ease of installing
VSS. The consensus in demand is for the VSS to be proactive and persistent in monitoring.
The most prevalent VSSs employ visible light cameras that do not easily function in a
persistent—all day, all night, and every day of the week—manner and their performance
is hampered by over, under, and uneven illumination during the day and the use of
artificial light at night. Thermal infrared cameras can monitor persistently because they
are not affected by the problems of visible light cameras. All objects, whether hot, at room
temperature, or frozen, emit energy from the infrared part of the electromagnetic spectrum
called heat signature. Sensors in thermal cameras can detect the objects’ infrared radiation
and create an image based on that information.
Proactive VSS systems predict critical events before they happen and provide alerts
for them. Pedestrian detection is a significant task towards achieving proactive systems as
it provides primitive information for interpretation of the video footage. Pedestrians are
typically warmer than most objects in thermal video streams because human skin emits
(radiates) infrared energy almost perfectly with emissivity at 0.98 where 1 is the value of
the perfect radiator [
1
]. The amount of infrared energy detected from objects in a scene by
thermal sensors depends on the emissivity of the object, the reflectivity of other objects
in the immediate vicinity [
2
], and the prevailing weather conditions. Figure 1 presents a
Sensors 2022, 22, 1728. https://doi.org/10.3390/s22051728 https://www.mdpi.com/journal/sensors
