While great progress has been made in recent years, there is still much room for

improvement in ADAS systems. The frontal collision warning systems and the pedestrian

detection systems (and also of cyclists, motorcyclists etc.) are a good example of this, since

these systems present difficulties when they have to operate in certain driving conditions

(high-speed driving, environments with poor or changing lighting or visibility etc). In addition,

most advanced current systems use sensors with reduced or very specialized functionality

(RADAR, LIDAR, LASER), which aditionally become relatively expensive. As an alternative,

several consolidated companies and start-ups are currently working on improving the

functionalities of image-based systems (cameras), foccusing in new algorithms and

developing more efficient and capable processors, hence trying to improve their robustness

and expanding their functionalities. Such new systems -if achieved- could dominate in the

future the ADAS because of its extensive functionality and reduced cost.

The present project is proposed in this context but with a different approach: the use of

spectrometers (multispectral cameras) of small size for the development of ADAS systems of

wide functionality and greater robustness. There are several motivations that encourage us

to this approach at this precise moment. On the one hand, the spectacular results that the

use of the rich information acquired from hyperspectral sensors (data on hundreds of

spectral bands of reflected light), has provided in various fields of science and technology, as

already mentioned in previous sections of this memory. It seems clear that the application of

this technology to the ADAS, if carried out successfully, can lead to a small revolution in the

sector in the near future, and this is precisely the first point that this project wants to

influence: explore how and to what extent the use of accurate multispectral information can

be useful in ADAS applications and more specifically in the detection, classification and

tracking of moving objects, since to date this has not been properly investigated.