The existing initial filtering systems of matching road all the database based on the disk file to store road network data and grid index. Due to the big overhead between in and out memory, the traditional disk database fails to support real-time and high throughput processing of the large-scale GPS data. To address this issue, this paper imports main memory database into the system. We built a test system by integrating the existing filtering algorithm and the MM-DB. The experimental result demonstrates that our method improves the search performance of the system up to 300% and promotes the speed and throughput of the system effectively.