Abstract:

This paper proposes a eliminate method that can apply to the gross error in image matching. The whole process can be divided into three steps. First, the Systematic differences are offset. Then, Triangulated Irregular Network (TIN) of image matching points are constructed to get the partitioning local binning. Based on the normal distribution of the image matching gross error, vector descriptor is proposed in the statistics on the local binning. The Reasonable threshold is set for quickly elimination of gross error. The feasibility of the method is tested based on the experimental verification of multiple groups of data. This method has been successfully applied to processing system of satellite image and low-level image.

Key words: image matching, gross error elimination, local binning, vector descriptor