Abstract: Compact ground wave radar is an important means for maritime target monitoring and early warning within the exclusive economic zone. However, due to its low transmit power and low signal-to-noise ratio of target echoes, the low detection probability may easily lead to missed detections of maritime vessels, making track initiation with sequential methods more challenging. In this paper, a maritime target track initiation method based on multi-frame clustering is proposed for compact ground wave radar by analyzing the different characteristics of clutter and targets in terms of position and velocity changes over time. This method leverages the discrepancy in motion features between the target and clutter across multiple consecutive frames. By incorporating motion feature constraints into the Density-Based Spatial Clustering of Applications with Noise (DBSCAN), the plots of each vessel in multiple frames are grouped into distinct clusters, thereby distinguishing the target track points from the clutter ones. The track points within each cluster are connected in chronological order to produce the initiated target trajectory. Track initiation experiments were conducted using both simulated and field data. The results show that compared with the logical method, the proposed method improves the initiation time by 5.95 minutes on average with the loss rate reduced by 12.68%. The proposed method can solve problems of delayed weak target track initiation and is suitable for initiating target tracks in cluttered areas or in long distances.