投影不变约束的矿区遥感影像特征点匹配

In mining area remote sensing images, the phenomenon of "homogeneity of different points" and the ground surface features changing is evident. The traditional SURF (speeded-up robust features) algorithm would be inapplicable in mining area remote sensing images mosaic. A projective invariant constrained SURF point matching method is proposed . Feature points are extracted from the SURF points on the large scale. Based on the principles of the projective invariant, calculate the initial projective transformation model of the unregistered image; The transformation model was then used to estimate the overlap region and locations in the base image corresponding to points in the unregistered image. Feature points are extracted in small scales and matched using spatial constrained method. The projective invariant are then used to eliminate the wrong matched. Finally, the obtained feature points are used for image mosaic. The experiment on Juye coal mine satellite image matching validated that the method is better than traditional SURF algorithm both in speed and in accuracy. There are no obvious dislocation in the mosaic image which can meet the needs of the mining area image mosaic.