Technologies of Vision
SCCAA project supported by COMUNE di TRENTO
SCOCA is a system for the automatic counting, classifying, and tracking of vehicles in a road intersection using video information only.
To achieve these goals with a real time solution, the research has been concentrated on the design, implementation, and evaluation of computer vision algorithms mainly based on the detection of moving objects in a real-world scene, trajectory tracking, and feature extraction for object classification.
Applicative goal of the system is the collection of data, for later access, such as the number of vehicles of each class that cross the intersection, their average speed, the intensity and the direction of the traffic flow with respect to the time, the origin-destination distribution map, and so on (see the SCOCA Statistic Suite package).
Furthermore, if unusual events are detected (e.g. busy intersections), in principle, the system can alarm in real-time the traffic controllers who, observing the scene in remote, could intervene.
The traffic flow information coming from different intersections can be used for a smart timing and coordination of traffic lights that minimizes stops and gives priority to public transport or emergency vehicles and permits the communication with the drivers by means of changeable message signs.
This image illustrates a frame of the mpeg sequence captured by a camera positioned near a road intersection. The camera is a standard pole-mounted surveillance camera (height: 7.8 mt). The sequence is sent to the main center, where it is decompressed and analysed by SCOCA in real-time. The system tracks the vehicles and count them when they pass through the scene. Speed of vehicles is obtained from information about the camera parameters and the tracking phase. Virtual gates are defined by means of a user interface in the set-up phase. They permit to build of the O/D map of the cross. The vehicles are classified into several categories using a model based paradigm for a rough subdivision (e.g. the same model for bicycle and motorcycle) and other more specific features for a classification refinement.
SCOCA distinguishes vehicles in seven classes: bicycle, motorcycle, car, van, lorry, urban bus, extraurban bus. Isolated pedestrian are also captured.
Other user-side information on SCOCA is available in the SCOCA working page with demonstrative screen recording sessions of SCOCA at work.
Statistic Suite - a package that analyses the rough SCOCA output to provide high level statistics: traffic composition, traffic volume and average speed through the time, distinguished for each class and for each lane; distance between vehicles with respect to safety distance, occupancy for each specified region of interest and class, anomaluos paths analysis and classification.
Traffic WebCams in Trento
SCOCA has been implemented on a movable station (a van with telescopic pole two TPZ cameras and PC) to acquire traffic data in not video surveilled junctions.
The first demo and a poster of the SCOCA project was presented at the National Congress Traffico e Ambiente, Trento, Italy, 21-25 February 2000 (in Italian).
The project was presented at the invited talk: S.Messelodi, C.M. Modena, Nuovi sistemi per il conteggio automatico e la classificazione dei veicoli, National Workshop "Noise Mapping: Quali dati per la mappatura del rumore?" - Milano, 14 March 2000 [pdf] (in Italian).
- S. Messelodi, C.M. Modena and G. Cattoni
Vision-based bicycle / motorcycle classification
Pattern Recognition Letters, Vol. 28, No. 13, pp. 1719-1726, October 2007 [abstract]
- S. Messelodi, C.M. Modena and M. Zanin
A computer vision system for the detection and classification of vehicles at urban road intersections
Pattern Analysis and Applications, Vol. 8, No. 1-2, pp. 17-31, September 2005 [abstract] [TechRep pdf]
- S. Messelodi, C.M. Modena, N. Segata, M. Zanin
A Kalman filter based background updating algorithm robust to sharp illumination changes
13th International Conference on Image Analysis and Processing - ICIAP 2005, Cagliari, Italy, September 6-8, 2005 [abstract][pdf]
- N. Segata
Sistemi di monitoraggio video: il problema dello sfondo in presenza di variazioni globali di illuminazione
Tesi di Laurea in Informatica, Universita' degli Studi di Trento, 2004 - ITC-irst P04-12-27 (in Italian).
- G. Monfardini
Recursive Neural Networks for the Classification of Vehicles in Image Sequences
ITC-irst Technical Report T04-02-06, February 2004, in Biological and Artificial Intelligence Environments, 15th Italian Workshop on Neural Networks, SIREN, 2005, B. Apolloni, R. Tagliaferri, M. Marinaro (Editors), pp. 359-368
- G. Cattoni
Utilizzo di Support Vector Machine per la classificazione di veicoli basata su caratteristiche visive: sviluppo di un classificatore di motociclette/biciclette
Tesi di Laurea in Ingegneria delle Telecomunicazioni, Universita' degli Studi di Trento, A.A. 2002/2003 - ITC-irst P03-11-03 (in Italian).
Contact person: Stefano Messelodi - messelod (at) fbk.eu
Page Maintainer: Carla Maria Modena