Nexyad-Logo3D



keywords : advanced drivers assistance, adas, road safety, obstacle detection, on board vision, road detection, visibility estimation, distance of visibility, fog detection, rain detection, adherence estimation, adhesion estimator, grip estimation, road tire grip, available technology, time to business, monitoring, image compression, fast image binarization, obstacle detection video, obstacle detection film, obstacle detection movie, road detection video, road detection  film, road detection movie,


Algorithms for Advanced Driver Assistance Systems (ADAS) and road safety applications


The following technologies are available. If you develop applications, think about building your solutions on the basis of existing and working subsystems ... instead of developping your own low level algorithms from the start ... Time to business should be shorter !
These on-board algorithms are available both as .dll or as RT-MAPS modules.

Advanced Drivers Assistance (ADAS) and Road Safety Applications 
NEXYAD was one of the major partners of the French research program ARCOS. We developped 3 algorithms usable in real time (on board solutions) for Advanced Drivers Assistance Systems (ADAS) or Road Safety Applications :


Obstacles detection :
Sensors : 1 regular camera (colour is not needed)
Needed : connection to CAN bus : informations used by the algorithm = speed of the wheels
Example of results : (click on the image to start the movie)
This low level artificial vision algorithm is very efficient for pedestrian, cars, and traffic sign detection. From this low level brick (that gives leads of obstacles or interesting objects) it is possible to add upper layers that filter and deliver the expected information (the upper layers are application sensitive, of course).

Road detection :
Sensors : 1 regular camera (colour is not needed)
Needed : nothing else



Example of results : click on the image to start the movie
This low level artificial vision algorithm is very efficient for road detection. Road detection is needed for at least two functionalities :
 - to detect the road (intelligent car that "knows" where the road is)
 - to eliminate from the image every area that is not "interesting" for the obstacle detection/recognition algorithms (the road detection algorithm is very fast : every area detected as "road" won't have to be screened by the pattern detection and recognition algorithms).


Visibility estimation :
Sensors : 1 regular camera (colour is not needed)
Needed : nothing else

Example of results : click on the image to start the movie
NEXYAD is involved into visibility measurement for more than ten years. We developed algorithms and a very strong know-how on this field of artificial vision. In particular, we developed a tool called VisiNex that lets test the wiper system efficiency (including rain sensors, software strategy, blade, rubbers, motors, ...). We also worked on an on-board algorithm.
You can download HERE a film of this on-board visibility estimation.



300 tracks virtual sensor for car dynamics :
Sensors : regular sensors set + speed sensor (may be replaced with a virtual sensor)
Needed : Driver's command signals (on CAN bus)
Results : estimated longitudinal and lateral grips
Minimum level of acceleration/deceleration needed : 2 m/s2
Even in the cases where the grip is not well estimated, dynamics of the model and of the real vehicle are similar enough to consider the 300 tracks of the model as 300 virtual sensors.
Average convergence time : 1 s

Example of results :
This research is made by NEXYAD in partnership with the french research company SERA CD that developed a car simulator (CALLAS). NEXYAD's algorithms re-adjust the parameters presets of the car model in real time. One of the presets concerns the grip. Once the dynamics of the simulation is similar enough to the dynamics of the real vehicle, the 300 tracks of the model are considered as 300 virtual sensors.