Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
ORUSSI: Optimal Road sUrveillance based on Scalable vIdeo
1. ORUSSI: Optimal Road sUrveillance based on Scalable vIdeo
ERA–SME 2009 (6th call)
Bertini M., Del Bimbo A., Dini F., Nunziati W., Seidenari L.
Magenta s.r.l., Florence, Italy,
Media Integration and Communication Center, University of Florence, Italy
Integrasys inc., Madrid, Spain
{fabrizio.dini,walter.nunziati}@magentalab.it - http://www.magentalab.it
{bertini,delbimbo,seidenari}@dsi.unifi.it - http://www.micc.unifi.it
Project Goals
The growing mobility of people and goods has a very high societal cost in terms of traffic congestion and of fatalities and injured people every year. The management
of a road network needs efficient ways for assessment at minimal costs. Road monitoring is a relevant part of road management, especially for safety, optimal traffic
flow and for investigating new sustainable transport patterns. On the road side, there are several technologies used for collecting detection and surveillance information:
sophisticated automated systems such as in-roadway or over-roadway sensors, closed circuit television (CCTV) system for viewing real-time video images of the roadway
or road weather information systems for monitoring pavement and weather. Current monitoring systems based on video lack of optimal usage of networks and are
difficult to be extended efficiently. Our project focuses on road monitoring through a network of roadside sensors (mainly cameras) that can be dynamically deployed
and added to the surveillance systems in an efficient way. The main objective of the project is to develop an optimized platform offering innovative real-time media
(video and data) applications for road monitoring in real scenarios. The project will develop a novel platform based on the synergetic bundling of current research
results in the field of semantic transcoding, the recently approved standard Scalable Video Coding standard (SVC), wireless communication and roadside equipment.
On-board Video Analysis
Demo 1: Vehicle counting and speed estimation on Axis Demo 2: Fast Feature Detection
ADP platform We show how it is possible to extract low-level features directly on the camera. Low-
We developed an embedded traffic monitoring application on the Axis ADP level features are the starting point of several possible image processing applications.
platform. The application runs directly on camera and provides vehicle counting Our application* extracts FAST corners that can be used for:
and statistics on the velocity of observed vehicles. In a nutshell:
Virtual sensors extract a signal from the pixels’ value, based on edges intensity; Camera calibration.
the signal is used to infer the presence of a vehicle onto the sensor, making it (PTZ) Camera tracking.
possible to realize vehicle counting;
Text localization.
by using two properly arranged sensor, speed estimation is possible;
allows very efficient implementation: only the pixels corresponding to the Image complexity
sensors area have to be processed; characterization.
robust to a wide range of lighting and evironmental conditions. Camera tampering
detection
* we thank Leonardo Galteri who participated in the development of the on-board application.
Off-line Video Analysis
Demo 3: Anomaly Detection Demo 4: Selective Transcoding
To capture scene dynamic statistics together with appearance in video surveillance Our approach [2] is based on adaptive smoothing of individual video frames so that
application, we propose a method [1] based on dense spatio-temporal features. image features highly correlated to semantically interesting objects are preserved.
These features are exploited in a real-time anomaly detection system. Anomaly This adaptive smoothing can be seamlessly inserted into a video coding pipeline as a
detection is performed using a non-parametric modelling, evaluating directly local pre-processing state. Experiments show that our technique is efficient, outperforms
descriptor statistics, and an unsupervised or semi-supervised approach. A method standard H.264 encoding at comparable bitrates, and preserves features critical for
to update scene statistics, to cope with scene changes that typically happen in downstream detection and recognition.
real world settings, is also provided. The proposed method is tested on publicly
available datasets and compared to other state-of-the-art approaches.
1
0.96
0.9
Our approach 0.94
0.8
H.264 0.92
0.7
Size %
SSIM
0.9
0.6
0.88
0.5
0.86
0.4
0.84
0.3
0.82
0.2
20 20.5 21 21.5 22 22.5 23 23.5 24 24.5 25 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
CRF Size %
Dataset
Dataset for vehicle counting and classification: Thanks to the involvement of Comune di Prato (a local municipality), we were able to collect a very
wide dataset that turned out to be key for the project activities. The dataset is made of more than 250 hours of recording taken on a well-travelled county road, with
different lighting and weather conditions. From these video sequences we have extracted an image dataset of about 1250 vehicle images. The data set, publicly
available at www.micc.unifi.it/projects/orussi, will be used to train a vehicle classifier.
References
[1] M. Bertini, A. Del Bimbo, L. Seidenari, “Dense Spatio-temporal Features For Non-parametric Anomaly Detection And Localization” in Proc. of ARTEMIS Int‘l Workshop on Analysis and Retrieval of Tracked Events and Motion in
Imagery Streams, Florence, Italy, 2010
[2] A.D. Bagdanov, M. Bertini, A. Del Bimbo, L. Seidenari, “Adaptive Video Compression for Video Surveillance Applications” in Proc. of ISM Int‘l Symposium on Multimedia, Dana Point, California, USA, 2011.