In order to improve the efficiency of dummy generation and the query service quality in the k-anonymous location privacy protection method of dummy, and to solve the problem of that the preprocessing is complex and the geographic semantic features are not fully considered in dummy location generation, a k-anonymous location privacy protection method of dummy based on approximate matching is proposed. Firstly, the area is divided into several square grids, and the coordinates of each location are converted to Morton code. Then, through the approximate matching between the Morton codes of different locations, a candidate set of dummies is generated, and the locations in it are non adjacent to each other and distributed in different grids. Finally, by matching approximately geographic names information of locations, the semantic similarity between any two locations in the candidate set is obtained, and $k-1$ locations with the minimum semantic similarity are selected as dummies. Experimental results show that the method can ensure the physical dispersion and semantic diversity, and can improve the efficiency of dummy generation. At the same time, the balance between privacy protection security and query service quality is achieved.