The invention belongs to the technical field of image processing, and relates to an urban wide baseline image feature point matching method based on local geometric structure constraint, which comprises the following steps of S1, extracting point features and linear features, and obtaining local geometric structure information of feature points; S2, constructing a feature point supporting region;S3, acquiring a feature region and a feature descriptor; S4, performing initial matching based on the similarity of the feature descriptors, and obtaining a basic matrix between the initial matching set and the stereo image pairs; S5, carrying out secondary matching on the feature points which are not successfully matched and have at least two geometrical structure directions; S6, performing feature point matching expansion on the feature points which are not successfully matched; S7, combining all the matched matching sets, removing the wrong matching in the matching sets, and obtaining a final matching set. The method solves the problems that in the prior art, wrong matching is generated for feature points located at the discontinuous parallax position, and a reliable initial matching set is difficult to obtain, so that a final matching set is unreliable.