Proposal of a locating method with video of driving shot

In order to utilize digital maps more substantially and flexibly, not only the shapes of buildings and streets, but also the texture and color information is desired. To extract such information from related videos, it is necessary to associate the videos with related digital maps. Moreover for a car navigation system, it is difficult to locate the current position at the accuracy of a building due to the errors of GPS (Global Positioning System) navigators. In this circumstance, visual information can be employed for error amendment. In both cases, finding out the correspondence between a video and related digital maps is indispensable. We propose a locating method which can deal with a video of driving shot when the car moves in zigzags. At first, the street and intersection information of a specified area is represented as a digraph with multiple weights. Then, after the car movement route is obtained from the video, candidate map routes are extracted by searching the car movement route in the digraph. Finally, an end-opened DP (dynamic programming) matching algorithm is employed to find out the real route from multiple candidates. In order to show the effectiveness of this proposed method, experimental results are also presented