Pattern Recognition 41 (2008) 432 – 444
A real-time object detecting and tracking system for outdoor night surveillance
Kaiqi Huang a , ∗ , Liangsheng Wang a , Tieniu Tan a , Steve Maybank b a National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing 100080, China b School of Computer Science and Information Systems, Birkbeck College, Malet Street, London WC1E 7HX, UK
Received 27 April 2006; received in revised form 31 March 2007; accepted 23 May 2007
Autonomous video surveillance and monitoring has a rich history. Many deployed systems are able to reliably track human motion in indoor and controlled outdoor environments. However, object detection and tracking at night remain very important problems for visual surveillance. The objects are often distant, small and their signatures have low contrast against the background. Traditional methods based on the analysis of the difference between successive frames and a background frame will do not work. In this paper, a novel real time object detection algorithm is proposed for night-time visual surveillance. The algorithm is based on contrast analysis. In the ﬁrst stage, the contrast in local change over time is used to detect potential moving objects. Then motion prediction and spatial nearest neighbor data association are used to suppress false alarms. Experiments on real scenes show that the algorithm is effective for night-time object detection and tracking. 2007 Published by Elsevier Ltd on behalf of Pattern Recognition Society. Keywords: Visual surveillance; Night; Contrast; Detection and tracking
Object detecting and tracking are important in any visionbased surveillance system. Various approaches to object detection have been proposed for surveillance, including
feature-based object detection [1–4], template-based object detection [8,9] and background subtraction or inter-frame
difference-based detection [5–7].
Background subtraction is the most popular detection method
used in object trackers. In Refs. [5–7], the values taken by individual pixels over time are statistically modeled. Detection is performed by ﬁnding those pixels with values that deviate from statistical model for the background. In Ref. , Wren used a single Gaussian kernel to model the YUV color space of each pixel. Stauffer and Grimson  generalized this scheme by modeling the RGB values of pixels using a Gaussian mixture Part of this paper has been announced in the ACCV06
∗ Corresponding author. Tel.: +86 10 62653768.
E-mail addresses: firstname.lastname@example.org (K. Huang),
email@example.com (L. Wang), firstname.lastname@example.org (T. Tan),
email@example.com (S. Maybank).
model. There are also many feature-based motion detection
methods, most of which rely on ﬁnding corresponding features in successive frames. Suitable features include Harris corners, color and contours [1,2,4].
Most algorithms for object detection and tracking are
designed for daytime visual surveillance. However, night-time visual surveillance has gradually attracted more and more attention. If scene is completely dark, then it is necessary to use a thermal infrared camera [10,11]. However, the cost of a thermal camera is too high for most surveillance applications. For this reason, thermal images are not considered in this paper. It is a great challenge to detect objects at night using ordinary CCTV cameras, because the images have low brightness, low
contrast, low signal to noise ratio (SNR) and nearly no color information. In this paper we focus on outdoor scenes with
low light levels, just sufﬁcient to allow a human observer to detect and track large moving objects.
Preprocessing is one way to improve image quality.
Narasimhan et al. study the visual effects of different weather conditions and remove weather effects due to fog and rain
[12,13]. There are also many video enhancement methods to...
Please join StudyMode to read the full document