Validation Database New
Road Detection & Road Safety
NEXYAD tools for ADAS

NEXYAD Automotive & Transportation Newsletter #4, the 7th of September 2015

Validation database for camera-based ADAS

The company NEXYAD started building a database for validation of advanced driver assistance systems (ADAS and Autonomous car) using the methodology AGENDA published in the 90 by Gérard Yahiaoui (methodology initially developped for control construction of learning and test databases for the implementation of artificial neural networks).
This database has two essential characteristics:

1) Known life situations
Indeed, the methodology AGENDA proposes to describe potential changes of signals and images came into factors of variability and their crosses.
Example, for obstacle detection :
   . weather (dry overcast, sunny weather, rain, fog)
   . overall brightness (low, medium, high)
   . speed of the carrier vehicle (low, moderate, high)
   . type of road (highway, road with marking, road without marking …)
   . coating (bitumen 1, bitumen 2, …, cobblestones)
   . day / night (headlights and the lights switched infrastructure)
   . season (spring, summer, autumn, winter)
   . etc …

      > type of obstacle :
           – stopped
                      . infrastructure-related: work terminals, tolls, …
                      . related users: tire on the road, parcel felt from a truck lying on the road, biker following a road                       accident, disabled vehicle stopped on the floor, standing pedestrian on roadside edge (dodger /                       no sniper)
           – moving
                      . truck, car, vulnerable (pedestrian, bicycle, motorcycle) each with types trajectories (longitudinal
                      in rolling direction, longitudinally in the opposite direction of rolling side) and position (opposite
                      to right, left).
                      . Etc…

We see that if we cross these factors, we find fairly quickly a huge number of cases. However, the development of ADAS systems is complex, and it is necessary to proceed by successive iterations, starting from simple situations to move to complicated situations.
Our database allows this, since all records are described in terms of crossing the terms of the factors of variability. Thus knows exactly which cases were tested or not by the system.
Formalism ‘crossing of variability factors of the terms’ allows using design of experiments, and in particular orthogonal fractional plans to sharply reduce the number of cases to be tested while ensuring maximum coverage of life situations. One can in this context to develop a fractional ADAS on an orthogonal plan and test other hard fractional orthogonal planes for example.

2) Reality reference
This is to crop images barriers and infrastructure elements (markings, roadsides, etc.) so as to constitute a reference to measure system performance.

. Examples of life situations:
Life Situations

1.1, summer, overcast, unmarked road, moderate speed tire on the floor, dry weather
1.2, summer, overcast, unmarked road, moderate speed, parcels on the floor, dry weather
2.1, summer, overcast , unmarked road, moderate speed, standing pedestrians non ambush at the edges of the floor, dry weather
2.2, summer, overcast, unmarked road, moderate speed, lying on the floor human, dry weather
etc …

Not sure that you would meet those few cases, even with on million kilometers on open roads.

Our Goal

NEXYAD starts his collection of images and data:
      . video (towards the front of the vehicle) Color
      . accelerometers
      . gyros

The files are synchronized by RT-MAPS tool INTEMPORA society.
The files are saved as RT-MAPS format and replayable directly by this tool.

NEXYAD currently looking for contributors on this internal project. Co contributors fund and in return free access to the database, unlimited in time. This contribution will accelerate the work of collecting and labeling.
NEXYAD wishes to provide this basis before June 2016, free way to give the material to the community and the ADAS autonomous vehicle for a smaller version of the database, and pay way (as subscriptions) for complete database.
NEXYAD’s ambition is to spread its methodological expertise and allow everyone to assess the performance of vision systems for ADAS, whether systems developed by NEXYAD, or others.

“Methodology for ADAS Validation: Potential Contribution of Other Scientific Fields Which Have Already Answered the Same Questions”, Gérard Yahiaoui, Pierre Da Silva Dias, CESA congress Dec 2014, Paris, proc. Springer Verlag
“Methods and tools for ADAS validation”, Gérard Yahiaoui, Nicolas du Lac, Safetyweek congress, May 2015, Aschaffenburg

For questions, or if you wish to become a contributor, please contact NEXYAD : +33 139041360


Road detection for ADAS and autonomous vehicle :
NEXYAD module RoadNex V2.1

A useful complement to markings detection

The detection of the road is a key element of driver assistance systems (ADAS) and autonomous vehicles.
Indeed, objects, obstacles, other road users, must be detected but also positioned relatively to the road.
The detection of the entire route, that is to say not only its markings or edges, but all the way, should enable
embedded intelligence to select appropriate action.

The company NEXYAD has been working on this issue for over 20 years without interruption, and has accumulated a large number of cases of road types, of coatings, in various atmospheric conditions.
This is to detect the rollable area on the road, without regard to, in a first step, lane markings.
Indeed, in Europe, there are many unmarked roads, and work on a marked road may change the markings and
make a « follow the markings strategy » dangerous.

In the images below you can see on the left a typical French countryside road with no markings, and on the right image, new markings was achieved while former markings still strongly visible.
Road without MarkingRoad with old and new Markings

These cases are quite common on our European roads and a driver assistance system, or a driving delegation
system, must at least understand such cases and if necessary tell the driver to cope with it by himself.

The NEXYAD road detection module, RoadNex V2.1 is a brick to go further to cope with these cases :
RoadNex V2.1

RoadNex V2.1 should be coupled with road signs detection, road markings detection, obstacle detection, in order to build an intelligent perception system. RoadNex is then a key module of such a system.

The road detection module NEXYAD, RoadNex V2.1 is available as a component into the asynchronous real time framework RT-MAPS : See HERE


Road Safety for ADAS and autonomous vehicle :
NEXYAD module SafetyNex running as real-time component
of Framework RT-Maps

SafetyNex (safety level estimation for ADAS)
SafetyNex Onboard is a high level functional bloc (software) of safety measurement, taking into account map and GPS geolocation (shape of the road, crossing roads, … ahead), speed, accelerations, visibility, adherence, distance to obstacle, etc.
SafetyNex measures adaptation of the driving style to infrastructure topology, and possibly Dangerous situations.
Two main applications :
_ Car industry : intelligent Navigation system providing valuable advices to keep the car in a good level of safety; sending alarms on dangers
_ Insurance : driving style measurment correlated with accidentology (insurance pricing, Pay How You Drive)

SafetyNex is now running in RT-Maps by IMTEMPORA
SafetyNex is under fusion with Ecogyzer (eco driving rating system) : this “package” will be the ultimate tool for eco and safe driving combination.

SafetyNex V2.1
SafetyNex v2.1

Using the NEXYAD road detection (RoadNex)
to make obstacles detection more robust

The detection of obstacles on the road, or even recognition of those obstacles, has become an
important issue for the next few years, in order to propose to the driver:
. Smart Systems for driving assistance : ADAS
. Delegation (partial at first, later full) of the driving task, to go step by step to the so-called autonomous

NEXYAD developed a vision-based obstacle detection system (ObstaNex) that aims to offer an alternative to
the current reference product on the market (MobilEye).

We detail the general principles of detection in a previous article.

It is interesting to note that NEXYAD also developed a road detection module named RoadNex.

RoadNex indicates the edges of the rollable way (thus highlights tighter if an obstacle is wayside, this may
be useful) and also indicates the rollable flat area (surface) in front of the vehicle.

RoadNex v2.1
RoadNex (NEXYAD): the rollable lane detection. For a clear urban lane.

RoadNex v2.1 with obstacle
RoadNex (NEXYAD): the rollable lane detection. For a urban labe cluttered by a vulnerable road user
in motion

One can see on this picture that RoadNex, even if it does not detect obstacles (because it detects
lanes), finally finds the “negative” of obstacles.

One of the uses of RoadNex is eliminating rollable areas from possibly detected obstacles by another module
(camera, radar, lidar …): the above image.
A simple confirmation that RoadNex not colored area corresponds to obstacles detection alarms
is sufficient to confirm the presence of an obstacle and to initiate, for example, the braking.

This system of cooperation between RoadNex and an obstacle detection system (ObstaNex, other
Vision-based module, radar, lidar, …) is particularly useful in city (see pictures RoadNex above)
and on the highway :

RoadNex v2.1 Highway
RoadNex (NEXYAD): the rollable lane detection. Case of clear highway lane

RoadNex v2.1 Highway with obstacles
RoadNex (NEXYAD): rollable lane detection while taking over a truck on highway

NEXYAD is currently working on a low-level fusion of ObstaNex and RoadNex in the context presented above.

To contact us

Other NEXYAD publications about ADAS ON NEXYAD WEBSITE




Driving Delegation: key elements

Driving Delegation: key elements for an artificial perception system
Publication of September 2, 2015
Authors : Gérard YAHIAOUI & Pierre DA SILVA DIAS

The automotive industry starts offering ADAS, and plans to propose in the near future partial or total driving delegation systems.

Main cases to be processed first may be:
. Highway driving, where the number of events per kilometer is small because the infrastructure has been designed to minimize path irregularities (little or no turns, every car in the same direction, wide track, geometric visibility up to several kilometers, enough little interactions between vehicles, at least when the traffic is flowing).
. The city, where infrastructure complexity is very large, where interactions between the road users are very strong, making detection a difficult tasks, but where speed of the vehicle is low.
In all cases, these future ADAS require developing advanced systems of perception.

Perception consist in detecting objects, clustering, and possibly tracking them in their own trajectory, from selected sensors (cameras, radar, lidar, slam, ultrasound, …)

It is usually presented as several phases :
. Detection: we perceive that “something” comes off the background, but we do not know what it this is. The Johnson criteria for detection give a theoretical limit of one period, or a minimum width of two pixels to detect a stationary object.
. segmentation and tracking: when zones are detected as being detached from the background (the landscape for image processing, the cluter for a radar, …), the detection must be agglomerated to track large enough objects that may have a meaning.
. Recognition: Recognition is to be able to say what it is. The Johnson criteria for human vision is about 6 periods (for stationnary objects) which gives 12 pixels.
. identification: identification gives, in the recognized class, the precise name of the object.

Detection is by far the most complex. It is potentially based on several principles:
. breakage hypothesis : we made a number of assumptions about world geography. We choose this hypothesis and make sure they are verified for the landscape (or cluter), and not for the objects to be detected. The non-validation of assumptions corresponds to a detection.
. the confrontation of a knowledge of the landscape or cluter: Comparing the “background” as it is supposed to appear in the absence of additional objects with said background which contains objects lead to detection of those objects.
. the knowledge of the shape of the objects to be detected: in this case the detection and pattern recognition are the same. System detects an object in its environment because it recognizes this object.
Human perception jointly implements the three principles.

Perceptions systems incorporate sensors and methods of processing, and are generally effective in a frame capture conditions, and little or not effective in the other frames. For example, a camera in the visible wavelenghts (and its image processing methods), will generally not be effective at night or in fog because “you can not see anything.”

No detection system can operate in any case when dealing with a real problem in the open world.

Designing a detection system then comprises two important phases:
. extend the maximum possible number of cases where the detection system works.
. have a diagnosis that allows to know when it is or when it is not in a position to that the perception system is effective.

We talk about performance (very efficient detection of all objects of interest), strength (number of cases where the collection system remains effective), and reliability (Situational Awareness in which one is and thus the confidence that can be placed in the collection system).
These three elements, performance, robustness, reliability, should be fully known in order to cooperate collection systems (for example, a camera and a radar).

NEXYAD proposed the Methodology AGENDA for characterizing life situations, using the formalism of orthogonal plans of experiments. The recognition of cases of functioning mode can be based on the description of life situations with this methodology. This gives a theoretical and practical framework for an estimation of robustness and reliability.
Performance is measured with statistical comparison operators: in general, it is considered the output of a detection system is a categorical variable with two categories: “detected” and “not detected”. This variable must be compared to a qualitative variable of reference that also has two modalities: “Presence of an object to
be detected” and “absence of objects to be detected.” The comparison can not be made by calculating a percentage (yet it is often that performance is measured this way), but it must use tools such as contingency table, the Khi2, normalized Khi2, khi2 in the box, etc …
To extend the life situations of the domain where the system detects objects correctly, we use to make cooperate several detection systems which use complementary types of sensors (eg in fog, we will trust in radar or infrared detection, but not detection by conventional camera).
A reliable system is one that is able to answer “I do not know”: in the case of driving delegation a system that could detect all objects so powerful, robust, and reliable in 30% of the time has a great value.
The delegation of driving frees 30% time of the driver, which is a real value proposition.

Safety is a discipline that encompasses many issues with the objective of ensuring the proper functioning of the system in all cases.
In particular, we must be vigilant concerning detection systems which require to have several measurement channels, such as stereovision.
If detection works only when you can have both cameras, then safety experts refuse such a system because two cameras means 2 times more likely that one fails.
We then see that perception system must have quite still usable “degraded mode” when simulating glitches sensors. A good design of a perception system for ADAS incorporates all these elements.

The race for performance that interests the engineers is rarely the real issue in industrial systems. A system that allows to delegate the driving in 30% of cases (eg clear overcast day dry) and “knows” when there is a case for which it works or does not work, can delegate driving and release the driver for 30% of the time.
This is a proposal for a very high value for the driver.
A system that works effectively in 99% of cases without knowing precisely when it works is absolutely unusable. No manufacturer will put such a system in operation for road safety applications.
The company NEXYAD has been working on these issues for twenty years, especially on road detection, obstacles detection, measurement of visibility (to describe cases where the detection is reliable, for example), the estimation of road safety (suitability driving style with the infrastructure).

NEXYAD developed:
. efficient and very robust basic bricks: RoadNex, ObstaNex, VisiNex onboard, SafetyNex
. a methodology for characterizing life situations in which it develops and tests an ADAS: AGENDA (Improvement performance, the recognition of cases of good performance, and validation of ADAS).
. know-how in collaboration between multiple perception systems.

Base de données de validation des ADAS utilisant des caméras

NEXYAD Automotive & Transportation Newsletter n°4, le 24 août 2015

Base de données de validation des ADAS utilisant des caméras

La société NEXYAD démarre actuellement la construction d’une base de données pour la validation des systèmes d’aides à la conduite et de délégation de conduite (ADAS et Autonomous car) en utilisant la méthodologie AGENDA publiée dans les années 90 par Gérard Yahiaoui (méthodologie au départ destinée à maîtriser entre autre la construction des bases de données d’apprentissage et des tests pour la mise en œuvre des réseaux de neurones).
Cette base de données a deux caractéristiques essentielles :

1) Situations de vie
En effet, la méthodologie AGENDA préconise de décrire les variations possibles des signaux et images d’entrées en facteurs de la variabilité et leurs croisements.
Exemple, pour de la détection d’obstacles :
   . météo (temps sec couvert, temps ensoleillé, pluie, brouillard)
   . luminosité globale (faible, moyenne, forte)
   . vitesse du véhicule porteur (faible, modérée, grande)
   . type de route (autoroute, route avec marquage, route sans marquage, …)
   . revêtement (bitume 1, bitume 2, …, pavés)
   . jour / nuit (phares et éclairages de l’infrastructure allumés)
   . saison (printemps, été, automne, hiver)
   . etc …

      > type d’obstacle :
           – arrêté
                      . liés à l’infrastructure : bornes de travaux, péages, …
                      . liés aux usagers : pneu sur la chaussée, colis tombé d’un camion, motard allongé sur la
                      route suite à un accident, véhicule en panne arrêté sur la chaussée, piéton immobile sur le
                      bord de la chaussée (embusqué / non embusqué)
           – en mouvement
                      . camion, voiture, vulnérable (piéton, vélo, moto) avec à chaque fois les trajectoires types                       (longitudinale dans le sens de roulage, longitudinale dans le sens inverse du roulage, latérale)                       et la position (en face, à froite, à gauche).
                      . Etc…

On constate que si l’on croise ces facteurs, on trouve assez rapidement un nombre de cas énorme. Or, la mise au point des systèmes ADAS est complexe, et il est nécessaire de procéder par itérations successives, en partant de situations simples pour aller vers les situations compliquées.
Notre base de données permet cela, puisque tous les enregistrements sont décrits en termes de croisements des modalités des facteurs de la variabilité. On sait ainsi exactement dans quels cas on a testé ou pas le système.
Le formalisme de ‘croisement des modalités des facteurs de variabilité’ permet d’utiliser les plans d’expériences, et en particulier les plans fractionnaires orthogonaux pour réduire fortement le nombre de cas à tester tout en garantissant une couverture maximale des situations de vie. On peut dans ce cadre mettre au point un ADAS sur un plan fractionnaire orthogonal et le tester dur d’autres plans fractionnaires orthogonaux par exemple.

2) Réalité terrain
Il s’agit de détourer sur les images les obstacles et éléments de l’infrastructure (marquages, bords de route, etc) de manière à constituer une référence permettant de mesure la performance du système.

. Exemple de situations de vie :
Life Situations

1.1, été, temps couvert, route sans marquage, vitesse modérée, pneu sur la chaussée, temps sec
1.2, été, temps couvert, route sans marquage, vitesse modérée, colis sur la chaussée, temps sec
2.1, été, temps couvert, route sans marquage, vitesse modérée, piétons immobiles non embusqués au bords de la chaussée, temps
2.2, été, temps couvert, route sans marquage, vitesse modérée, humain allongé sur la chaussée, temps sec
etc …

Il n’est pas certain que l’on puisse rencontrer ces quelques cas, même en roulant 1 million de km sur route ouverte !


NEXYAD démarre son recueil d’images et de données :
      . vidéo (vers l’avant du véhicule) couleur
      . accéléromètres
      . gyromètres

Les fichiers sont synchronisés par l’outil RT-MAPS de la société INTEMPORA. INTEMPORA.
Les fichiers sont enregistrés au format RT-MAPS et directement rejouables par cet outil.

NEXYAD cherche actuellement des contributeurs sur ce projet interne. Les contributeurs co financent et ont en retour un accès gratuit à la base de données, illimité dans le temps. Cette contribution permettra d’accélérer le travail de recueil et d’étiquetage.
NEXYAD souhaite mettre à disposition cette base avant Juin 2016, de manière gratuite pour donner de la matière à la communauté des ADAS et du véhicule autonome, pour une version réduite de la base, et de manière payante (sous forme d’abonnements) pour la base complète.
L’ambition de NEXYAD est de propager son expertise méthodologique et de permettre à chacun d’évaluer les performances des systèmes de vision pour les ADAS, qu’il s’agisse des systèmes développés par NEXYAD, ou d’autres.

“Methodology for ADAS Validation: Potential Contribution of Other Scientific Fields Which Have Already Answered the Same Questions”, Gérard Yahiaoui, Pierre Da Silva Dias, CESA congress Dec 2014, Paris, proc. Springer Verlag
“Methods and tools for ADAS validation”, Gérard Yahiaoui, Nicolas du Lac, Safetyweek congress, May 2015, Aschaffenburg

Pour toute question ou pour devenir un contributeur, contactez NEXYAD : +33 139041360

NEXYAD vision-based ADAS modules for a demo car of “Université de Haute Alsace”

NEXYAD delivered a camera and two high-tech software modules :
. RoadNex : vision-based road detection in front of a car (for ADAS and Autonomous Vehicles)
. ObstaNex : vision and inertial navigation – based obstacle detection (for ADAS and Autonomous Vehicles)

Those two modules are integrated as component into the real time environment RT-MAPS (Intempora), and communicate with de neural network – based pattern recognition module Neuro-RBF (GlobalSensing Technologies).

NEXYAD will support Université de Haute Alsace during 2015 to integrate ObstaNex with their intertial
navigation unit.

NB : the 3 companies Nexyad, Intempora, GlobalSensing Technologies are members of the Mov’eo Groupement ADAS :



A version of RoadNex (road recognition module, by NEXYAD) will be soon available in the Real Time development tool RT-MAPS (by INTEMPORA) and in smart phones (B2C App).

Watch a demo :
[flashvideo filename= width=”600″ height=”450″ image=”” displayheight=”450″ /]

Link :

Presentation of NEXYAD products for European Car Industry (February 4, 2014)

Here are the powerpoint sildes of presentation of NEXYAD Automotive & Transportation : PDF Here

– Visibility measurement in your lab (Testing Application : for wipers, washer, windshield, rain sensor, lighting, demist, …) VisiNex

– Onboard Software (Artificial Vision) blocs for your ADAS development : (road detection) RoadNex, (road safety assessment) SafetyNex, (drowsiness detection) DrowsiNex, (visibility measurement) VisiNex Onboard, (obstacle/pedestrian detection) ObstaNex