| In the process of the large-scale orthogonal multiple screen special film making, the integration of the geometric rectification , splicing , and fusion is very important for the item . And it will research the geometric rectification in the integration mainly in this paper . The basic foundation of geometric transformation is introduced, and after intensive research of those existing approaches for geometric rectification, a novel algorithm based on BP neural network is presented. The method which this paper presents requires the input and output without considering the over-many uncertain factors. The method train the network with fastest-gradient decent algorithm, and can set up the mapping relations between space points and image points, then adopts a bicubic interpolation method for gray level determination. Simulation results in Matlab reveals that the algorithm proposed is both has good correction effect and of higher precision. |