Matching probability of visible image based on the side information of buildings

The current matching probability analysis does often not take into account the side information of a building, but only evaluates the image itself, which leads to significant differences between the reference map and the detector image. A new method that evaluates the effects of side information of every building in a scene for matching first and then evaluates matching probability according to the directional gradient entropy and the texture low frequency energy ratio of the visible image is presented. Experimental results show that this method obviously improves the reliability of the reference map of areas with more tall buildings with an average matching probability more than 90%.