Prediction of driver's pedestrian detectability by image processing adaptive to visual fields of view
Ryunosuke Tanishige, Daisuke Deguchi, Keisuke Doman, Yoshito Mekada, Ichiro Ide, Hiroshi Murase · 2014
Recently, pedestrian detection technology using in-vehicle cameras or sensors are being developed, which supports safety driving by notifying the drivers of the existence of pedestrians. However, warning of all existing pedestrians would interfere with the driver's concentration. Therefore, the driver should only be alerted of pedestrians with low detectability to avoid distraction of his/her concentration. To achieve this, it is necessary to develop a method to predict the detectability of a pedestrian by the driver. This paper proposes a method for predicting the pedestrian detectability adaptive to the characteristics of the human visual field. We prepared image features effective for the different regions of the human visual field; central and peripheral, in order to predict the pedestrian detectability correctly. To obtain the ground truth of the pedestrian detectability, we conducted an experiment by human subjects using image sequences captured by an omnidirectional camera including pedestrians. From the comparison between the output of the proposed method and the ground truth of pedestrian detectability, we confirmed that the proposed method significantly reduces the prediction error in comparison with the existing methods.