Chapter 3 Understanding driver attention to objects for ADASs: what do drivers see?

Le résumé fourni par la source

A driver’s actions and intent can be important factors in enabling advance driver assistance systems to assist drivers and avoid accidents. Having some knowledge of what traffic relevant objects the driver has seen or not seen could provide indications of how aware a driver is about the environment or their attentiveness. Determining what a driver has seen, i.e., recognized, is challenging, requiring determining whether the individual cognitively identified the object. Studies of human perception, however, have determined that as a precursor to recognition the individual’s gaze must fixate on an object for a period of time, estimated to be 250-500 ms. Determining a driver’s gaze can then be used to determine what objects a driver could have seen and what objects they did not gaze at, i.e., missed or ignored. Gaze, therefore, can provide insight into the driver’s intent or awareness of situations. This work presents a study of what driving relevant objects a driver gazes at during an actual drive. The data consists of a sequence of images from a stereo camera on an instrumented vehicle and a measurement of the driver’s point of gaze (PoG) for each frame. The analysis relies on object recognition of typical objects, such as traffic lights, vehicles, and traffic signs, and determining the driver’s PoG on a frame-by-frame basis. This enables a multiframe analysis to determine the length of time that a driver’s gaze fell on a particular object. The computations rely on two thresholds, one for determining whether two objects from consecutive frames can be considered to be the same object and another for the minimum number of consecutive frames. Different values for these thresholds are explored in the analysis. The methods presented can be useful in other scenarios involving the analysis of driver gaze and have implications for the design of future advanced driving assistance systems and for understanding of driver gaze and awareness.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.