Abstract: Due to the complexity of aerial image scene and many uncertain factors that interfere with line matching,we propose a line matching algorithm based on geometric constraint and MSLD in this paper.Firstly,the linear detector LSD is employed to obtain the line feature information of the image.Secondly,according to the geometric features between the different straight lines as the constraint condition,the straight line pair of the two aerial images is obtained by grouping.Then,the candidate line pairs are determined by using the kernel constraints,and the support regions of the reference and candidate line pairs are constructed in turn.The affine transformation is used to unify the size of the support regions MSLD describing the local appearance of lines.By calculating the Euclidean distance between line descriptors in aerial images,candidate lines satisfying the nearest neighbor distance ratio criterion are determined as matching results.Finally,the matching results are checked to remove redundant matching lines.The experiment uses aerial image data of different scene types.The experimental results show that the proposed algorithm can deal with the problem of the poor line matching effect of aerial images.

Key words: geometric grouping;line pairs;epipolar constraints;mean-standard deviation line descriptor;affine transformation