This paper presents the development of optic flow techniques applied to studies on visual perception and driver behaviour. The underlying work has been done by a conjoint team of road traffic psychologists and image analysis engineers. The image analysis techniques are used to quantify motion in real and simulated environments. In several of our studies the optical flow techniques had proven to be useful in describing the visual information processing related to motion detection and in predicting driver's performance in simulated road environments. In a fundamental study the optical flow computation allowed us to understand and to quantify the inhibition effects of self motion in object's motion detection. In applied simulated situations, computation of the environmental optical flow predicted the performance of drivers in a task of vehicle's motion detection. At the present, the optical flow algorithm has been improved and tested in more complex situations, both real and simulated. Several new applications are discussed: real time analysis of images, captured from the driver perspective, to support navigation tools and to improve human-machine interfaces, offline analysis to classify road types and to support driver's performance predictions (e.g., driving speeds, impairment of vehicles and pedestrians detection) and simulator validation. For the covering abstract see ITRD E113725 (C 22328 CD-ROM).
Abstract