Numerical Technologies for Vulnerable Road User Safety Enhancement - New Contributions in Information Systems and Technologies - page 337

be recognized accurately. While dependency on visual light can be partially countered
by e.g. FLIR cameras, the thermo-vision has its own limitations. They are characterized
by limited discrimination of objects with temperature similar to ambient or ones well
insulated - such as pedestrians in winter clothes. Despite high difference in human body
temperature to ambient, covered body parts are not well visible. While several different
data processing methods can be found in the literature [14, 15], all of them have several
limitations. The Neural Networks (NN) and Support Vector Machines (SVM), based on
supervised learning, show high robustness. However, the effects depend on used
training data. Moreover, processing speed significantly decreases with the image size. A
pattern matching by correlation is a simple method using cross-correlation to find the
best match between selected templates and video frames. It needs no training data, but
accuracy increases with number and complexity of patterns, while processing speed
decreased. A low number of templates result in low reliability, though algorithm may be
quite fast. Heuristic methods are based on recognition of body parts or postures. The
main limitation is a need for a high resolution of the images, certain operations on
background and general computation cost.

2.4 Autonomous Emergency Braking

The Autonomous Emergency Braking System is a solution designed to limit collision
speed or, under favourable conditions, to completely avoided an accident. Typical
AEBS warns a drive if a potentially dangerous situation occurs and, if there is no
timely reaction, starts the braking procedure. Such systems are commercially
available and will be mandatory from 2015 in all newly produced or imported to EU
soil trucks and coaches (class N2/M2 and above vehicles). However, they are not
necessarily designed to protect pedestrians and bicyclists in urban environment.
Nevertheless, such systems were tested in the past [11, 13] and presently a
commercial version is also available [16].

Fig. 1. Possible layout of AEBS system for vulnerable road user active safety