Map-matching is a fundamental issue for location-based applications and traffic pattern analysis. Widely utilized approach considers map-matching process as a Hidden Markov Model (HMM). However, it usually produces roundabout paths and takes redundant computation since Markov assumption restricts that the algorithm cannot make a full use of context information and it is easy to be misled. Therefore, in this paper, to achieve a higher running efficiency as well as matching precision, we propose a novel Seed Growing Matching (SGM). The main idea is that SGM appoints several road intersections as seeds to be starting points and each seed will grow interactively along with the fitted polyline of GPS trajectory. SGM ensures the continuity of matched path inherently. It also makes growing decisions based on the contextual information so that it gets rid of time-consuming shortest path computation which is necessary for HMM-based approaches. We conduct the real dataset-based experiments to demonstrate the performances of SGM. The results show that SGM outperforms HMM-based approach and achieves about twice the matching precision of baseline and runs about five times faster on our dataset.