A NOVEL METHOD OF CAMERA SELF-POSITIONING - MATCHING IMAGE CONTOUR LINES OF MOUNTAINS

摘要

A great amount of outdoors photographs of mountain area have no apparent information of camera position, although that information is extremely important in application such as mobile robot navigation. Fortunately, in most cases, the image contour line, which segments the sky background and mountain, can easily fetched from those photographs. Also the DEM data of the same area can be obtained by other approaches. This paper proposes a novel method of camera self-positioning, which makes good use of the image contour line of outdoors photograph. The key idea is that if each parameter of camera position is rightly set then the image contour line fetched from the outdoor image will match contour line that constructed from the known DEM data using these parameters and men we can search the right parameters in the parameter space. In fact, it's a typical Inverse Problem. To inverse the parameters of camera position this method uses genetic algorithms, which avoids premature convergence to local extremum. The distinct merit of the method is that there are no reference objects needed. Moreover, combining with other approaches, the method we proposed can be easily used to form new method that turns into the basis of camera self-positioning tool.