This paper proposes an original image-matching algorithm for consecutive digital images acquired by means of the CCD sensor of a Mobile Mapping System (MMS). For such end, once a point is (manually) chosen in the “first” image, a Kalman filter based
model is applied to automatically search for the homologous point in the “second” image. The matching geometric parameters are given from the six coefficients of the affine transformation from the template to the patch window, while the radiometric parameters are the two coefficients of the linear radiometric transformation between the same windows. The usual “Least Squares Matching” condition between template/patch grey-values is applied. To improve the solution convergence, the geometrical constraint given by the well known “epipolar line” (in the second image) is employed. This line comes out from the so-called “coplanarity condition” between homologous image points. The epipolar line equation makes possible to analytically define the filter “geometrical state equation”. This is possible since, while usually as “state vector” an unknown dynamic quantity is utilised, a simplest unknown geometric quantity, suitably defined, is now employed. An iterative solution strategy is carried out in recursive steps until the convergence is fulfilled.
The implementation/experimentation of the matching model is now in progress, by using MATLAB software, and it is giving well promising numerical results.
