Abstract: The blind spot images collected on the ground often lack pose information, which will cause stratification when combined with aerial image processing. Therefore, to improve the precision of the tilt model a new method is proposed to reconstruct the position and attitude of the complementary image by constraining the pair extreme singular matrix. Firstly, feature matching is carried out to obtain the pair singular matrix, namely the fundamental matrix. The projection matrix of aerial image is assumed to be normalized, and the projection matrix of ground image is composed of anti-symmetric matrix and non-singular matrix decomposed from the basic matrix. Then, the POS information of aerial image is used to eliminate the ambiguity of the projection matrix pair, so as to obtain the accurate position and posture value of the ground image. The test results show that this method can automatically calculate the position and pose information of the ground blind area retaking image without image control, and it is well matched with the aerial image, which improves the operation efficiency under the premise of ensuring the precision of the model.

Key words: fine modeling, position and orientation, fundamental matrix, ambiguity, image-free control