A Dual-Camera Surveillance Video Summarization
Generating Strategy for Multi-Target Capturing
Qingyun Shen
1
Key Laboratory of Jiangxi Province
for Image Processing and Pattern
Recognition, Nanchang Hangkong
University, Nanchang 330063,China;
2
School of Information Engineering,
Nanchang Hangkong University,
Nanchang 330063,
China
+86 18679122531
shinseiun@yahoo.com
Cihui Yang
1
Key Laboratory of Jiangxi Province
for Image Processing and Pattern
Recognition, Nanchang Hangkong
University, Nanchang 330063,China;
2
School of Information Engineering,
Nanchang Hangkong University,
Nanchang 330063,
China
+86 18779133310
yangcihui@nchu.edu.cn
Shipin Wen
University of Electronic Science and
Technology of China,
Chengdu 610054,
China
+86 18971124190
wenshiping226@126.com
ABSTRACT
Traditional surveillance video contains a large amount of
information which is too jumbled. Real-time video summarization
can solve this problem but also face much challenges. Different
from file summarization, real-time video summarization requires
higher efficiency. Meanwhile, the validity and quality of a
summarization should be ensured. To tackle these problems, we
propose a real-time video summarization strategy based on dual-
camera. In our strategy, a static camera and a PTZ camera are
necessary. The static camera is used to monitor the scene to detect
and track moving targets, and the PTZ camera is used to capture
the close-up information of moving targets as video
summarization, and the collaboration of these two cameras is
crucial. Specifically, in order to obtain multi-target summarization
efficiently and effectively, the priority of target capturing is
determined by its spatial information and historical representation
in the scene. Extensive experiments are performed on real-time
outdoor scene with our method. Experimental results show that
our proposed method is robust enough to capture multiple targets
in the same scene at the same time.
CCS Concepts
• Computing methodologies → Image processing.
Keywords
Real-time video summarization; Dual-camera coordination;
Multi-target priority judgment.
1. INTRODUCTION
It's very difficult for people to extract valuable information from a
long surveillance video in a very short time since it contains lots
of information. Video summarization is a kind of technology
which can extract key frames representing video content from
video automatically and save the time of finding valuable
information[1; 3]. It has been widely used, such as extracting
interesting events from sports games video, discovering specific
objects in video, and so on [6; 9-11; 15; 17]. Video summarization
is also used in surveillance video [14; 19], focuses on the target in
video, mainly on the appearance and disappearance of the target.
However, there are still many challenges in surveillance video
summarization. For instance, a surveillance camera with a wide
field of vision cannot provide high-quality target information. If a
camera is used to capture a single target for high-quality
information, the scene information will be lost. Thus, the
researchers proposed dual-camera system to solve this problem.
In recent years, many researchers have done lots of research on
the dual-camera system. Chen et al. [4] proposed a dual-camera
system, in which Gaussian mixture model and Kalman filter are
used to detect and track targets respectively, and a close-up
camera is used to obtain target-centric video. However, the close-
up camera is limited to capture one target, and the close-up
information of the other targets in the scene are lost. Ghidoni et al.
[8] describes a system based on an omnidirectional camera and a
PTZ camera that can record faces appearing in a scene. When
capturing a target, this system only considers the distance of the
target from the camera and the target’s moving speed, which
makes it lose the information of the targets farther away from the
camera. There are other dual-camera systems, such as [12; 18],
that can capture more information, but also have problems with
poor quality and easy to miss target. How to obtain high quality
summarization information without loss of target is still a key
problem of video summarization.
Aim at the above problem, we proposed an effective and robust
surveillance video summarization generating strategy. The
implementation of this strategy is based on two cameras. One is a
wide-angle static camera for global monitoring, and the other is a
high definition PTZ camera to capture the close-up information of
the moving target in a monitoring scenario and generate video
summarization. Summarization acquisition of multi-target will
determine the priority of multi-target by combining its spatial
location and historical performance in the scene. The problem of
low quality of video summarization is solved by using double
cameras. In general, our method solves the problem of low-quality
summarization, and the designed multi-target capture strategy can
effectively reduce target loss.
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. Copyrights
for components of this work owned by others than ACM must be honored.
Abstracting with credit is permitted. To copy otherwise, or republish, to
post on servers or to redistribute to lists, requires prior specific
permission and/or a fee. Request permissions from Permissions@acm.org.
ICVIP 2019, December 20–23, 2019, Shanghai, China
© 2019 Association for Computing Machinery.
ACM ISBN 978-1-4503-7682-2/19/12…$15.00
https://doi.org/10.1145/3376067.3376071