Lacks of visibility detection (fog, rain, …) by Computer Vision

. Detects bad visibility weather conditions and this can be shared through a cloud to other drivers (geolocalized)

. Detects when visibility conditions are not appropriate to camera-based detection (reliability of detection and sensor fusion : if Visibility is poor, do not trust camera-based detection), and also when driver is not good at detection (adapted HMI to bring trust in the cockpit)

. Computes visibility level. Bad visibility score is sent to SafetyNex as a factor of driving risk




software component


Visibility(t) :
. Visual Quality Score
. Distance of possible detection
. Distance of possible recognition



NEXYAD algorithm VisiNex

Estimate Visibility in front of a car
Estimate Visibility from a highway monitoring camera
Alert drivers before they get inside fog
Lines give distance of visibility. Color spots estimate if quantity of information inside the area is usable or not to do objects detection (theory of detection), with human eye and brain, or with computer vision detectors. When all colour spots are red, visibility is inappropriate to computer vision based detection and you know it ! (you can react properly in case of « white truck on a white sky background »)