In environs where cameras are installed on a freely moving platform, e.g. an autonomous vehicle, object detection and traversing becomes much more difficult. Especially, human detection technique is necessary for some applications such as autonomous v...
In environs where cameras are installed on a freely moving platform, e.g. an autonomous vehicle, object detection and traversing becomes much more difficult. Especially, human detection technique is necessary for some applications such as autonomous vehicle for safety. So, we present a real time human silhouette extraction in such challenging environments. To verify human region, we use silhouettes as human features. Since parts of a human body may be occluded by background objects, the human silhouettes may be incomplete. So, we present a robust method in detecting human and silhouette excluding the background and focusing on the human region. Experiments demonstrate that our method reduces significantly the false positives rate under very challenging conditions.