557 
AUTOMATIC TRAFFIC MONITORING 
FROM AN AIRBORNE WIDE ANGLE CAMERA SYSTEM 
D. Rosenbaum 3 ’ 8 B. Charmette a,b , F. Kurz 3 , S. Suri 3 , U. Thomas 3 , P. Reinartz 3 
J German Aerospace Center (DLR), Remote Sensing Technology Institute, PO Box 1116, 82230 Weßling, Germany - 
(dominik.rosenbaum, franz.kurz, sahil.suri, ulrike.thomas, peter.reinartz)@dlr.de 
b École Supérieure Chimie Physique Électronique de Lyon, Domaine Scientifique de la Doua, Bât 308 - 43 bd du 11 
Novembre 1918, 69616 Villeurbanne Cedex, France -baptiste.charmette@cpe.fr 
Commission III, WG III/5 
KEY WORDS: Aerial Digital Images, Monitoring, Disaster, Automation, Tracking 
ABSTRACT: 
We present an automatic traffic monitoring approach using data of an airborne wide angle camera system. This camera, namely the 
“3K-Camera”, was recently developed at the German Aerospace Center (DLR). It has a coverage of 8 km perpendicular to the flight 
direction at a flight height of 3000 m with a resolution of 45 cm and is capable to take images at a frame rate of up to 3 fps. Based on 
georeferenced images obtained from this camera system, a near real-time processing chain containing road extraction, vehicle 
detection, and vehicle tracking was developed and tested. The road extraction algorithms handle a-priori information provided by a 
road database for a first guess of the location of the roads. Two different techniques can be used for road extraction. In the first 
method, roadside features are found by using an edge detector based on ISEF filtering, selecting the steepest edge, which is normally 
the edge between the tarry roads and the vegetation. The second method extracts roads by searching the roadside markings using a 
dynamical threshold operator and a line detector. Vehicle detection then is limited to the road areas found by the road extraction 
algorithms. It is based on an edge detector, a k-means clustering of the edges, and on geometrical constraints, e.g. vehicle size. 
Vehicle tracking is performed by matching detected vehicles in pairs of consecutive images. For this matching the normalized cross 
correlation is calculated for each detected car within a limited search area. The algorithms for road extraction, vehicle detection and 
vehicle tracking proved to be quite sophisticated, enabling car detection and tracking rates with a completeness of 70 % and a 
correctness of up to 90 % on images obtained from a flight height of 1000 m. 
1. INTRODUCTION 
In a society which relies on plenary mobility, day-to-day large- 
area traffic monitoring is a quite useful tool to exploit the 
existing road capacities sufficiently. Moreover, daily 
commuters are interested in knowing their travel times to work 
and back, being able to plan their daily business or to change to 
public transportation systems in case of extensive congestion or 
traffic jam. Furthermore, during or after mass events and natural 
disasters, security authorities and organisations as well as 
rescue forces require fast and sufficient traffic guidance over 
large areas. 
In general, traffic monitoring is mainly based on data from 
conventional stationary ground measurement systems such as 
inductive loops, radar sensors or terrestrial cameras. One 
handicap of these methods is the low spatial resolution 
depending on their distribution on the ground. New approaches 
include data by means of mobile measurement units which flow 
with the traffic (floating car data, FCD, (Schaefer et al. 2002, 
Busch et al. 2004)). In order to handle traffic monitoring by 
remote sensing, a number of projects based on optical and SAR 
airborne sensors, as well as SAR satellite sensors are now 
running at DLR. One approach currently under development is 
to use a wide angle optical camera system for near real time 
traffic monitoring. The big advantage of the remote sensing 
techniques presented here is that the measurements can be 
applied nearly everywhere (exception: tunnel segments) and 
there are no dependencies on any third party infrastructure. 
Besides, airborne imagery provides a high spatial resolution 
combined with acceptable temporal resolution depending on the 
flight repetition rate, but require complex image analysis 
methods and traffic models to derive the desired traffic 
parameters. Moreover, estimates for travel times through the 
area of aerial surveillance can directly be determined from 
extracted traffic parameters (Kurz et al. 2007b). The paper is 
arranged as follows. In Chapter 2 the hardware is described as 
well as the testing data for the processing chain. Chapter 3 
characterizes the algorithms used in the processing chain, while 
Chapter 4 deals with the results from testing this software. 
Chapter 5 gives conclusions in brief. 
2. SENSOR AND DATABASE 
2.1 The 3K-Camera System 
The 3K-Camera system (3K: “3Kopf” = 3head) consists of 
three non-metric off-the-shelf cameras (Canon EOS IDs Mark 
II, 16 MPix). The cameras are arranged in a mount with one 
camera looking in nadir direction and two in oblique sideward 
direction (Fig 1), which leads to an increased FOV of max 
110°/ 31° in across track/flight direction. The camera system is 
coupled to a GPS/IMU navigation system, which enables the 
direct georeferencing of the 3K optical images. Boresight angle 
calibration of the system is done on-the-fly without ground 
control points based on automatically matched 3-ray tie points 
in combination with GPS/IMU data (Kurz et al. 2007a).