Accidentology shows that accidents can be explained by various sources of failures of this system, at the level of the perception of the environment, or at the decision level, or implementing the action decided. These different failure levels are addressed by various research initiatives, detecting potential threats or conflicts, or conveying important information to the driver, or taking the correct decision, or concretely triggering specific mechanical devices. Moreover, some projects address more than one level of a given application, which can resort to intersection management, or longitudinal control etc. PReVENT/COMPOSE subproject focuses on pre-crash and collision mitigation. Therefore, in relevant situations, collision can no longer be avoided, but active measures can still decrease the severity of damages. The overall idea behind dedicated devices is to decrease the amount of cinematic energy at the time of impact when it involves Vulnerable Road Users (VRUs), as active bumpers do for example. Along with reducing the amount of energy, the geometry of such devices aims at reducing the severity of consequences, trying not to impact fragile and vital parts of the VRU's body. Deciding to trigger such a device is an important and often irreversible choice. Taking such an option must rely on a comprehensive perception of the overall environment and detection of VRUs. A main trend within the perception community in various applications includes considering an array of different sensors to properly assess the environment. Long range radars, visible, near infrared or far-infrared (FIR) cameras, short range radars, laser-scanners are only examples of such sensors being scrutinized. Managing information coming from different sensors implies architecture, communication, registration issues and fusion algorithms which are not considered here. We concentrate on FIR imagery, as this sensor proves to be very fit for VRU detection. After establishing the practical and geometrical context for taking images, different algorithmic approaches relevant for VRU detection will be reported.
Far infrared detection algorithms for vulnerable road users protection
Weitsicht-Infraroterkennungsalgorithmen zum Schutz von gefährdeten Verkehrsteilnehmern
2006
15 Seiten, 11 Bilder, 26 Quellen
Aufsatz (Konferenz)
Englisch
Advanced protection for vulnerable road users
Kraftfahrwesen | 2005
|Advanced Protection for Vulnerable Road Users
SAE Technical Papers | 2005
|Vulnerable Road Users: Cyclist
British Library Conference Proceedings | 1992
|Vulnerable Road Users Detection Using V2X Communications
IEEE | 2015
|Advanced Protection for Vulnerable Road Users: A Case Study
SAGE Publications | 2006
|