A Dynamic Street View System Using Video Event Data Recorder

Abstract

In recent years, the street view system which (such as Google street view) is popular and convenient for users to observe real street environment on Geographic Information System (GIS) also helps users to search and locate their destinations more easily. However, the street view system usually takes two or three years to update the latest street view content for only several developed countries. Because high frequency update of the street view needs lots of street view cars with panoramic cameras and driving time to capture pictures or videos for the latest street views. It is a very time-and money-consuming task and thus cannot always provide users with the latest street views. The concept of this paper is to synthesize image sequences of arbitrary street view at any location captured by consumer electronic devices such as vehicle video recorders or smart phones. This concept is similar to dynamic street view images maintained by general users. The proposed scheme is used to develop a system which is able to map any image recorded from a video event data recorder (EDR) to the existing panoramic street view image database. This system does not aim to replace the panoramic image region by the EDR image. Instead, the purpose of the proposed system is to continually register each image to the corresponding panoramic image region in the database. In order to minimize the visual inconsistency between different frames, image synthesis is employed to minimize the geometric distortion within the resultant generated street view sequences.