High-definition maps (HD maps) are crucial for precise navigation of autonomous vehicles. However, creating and maintaining HD maps involves substantial costs and challenges. Hence, the online generation of HD maps using on-board sensors has gained considerable attention. Existing methods often encounter data visibility difficulties due to factors such as occlusion, long distance, etc., which are common challenges in autonomous driving. In this paper, we propose the framework to improve the accuracy of online HD map generation by introducing implicit clues from the trajectories of traffic participants. We specifically leverage the trajectory data from the driving scene and encode trajectories as an additional branch. Although trajectory data may contain tracking errors or noises, we adopt an attention-based architecture to adaptively focus on relevant trajectory information, thus significantly improving the performance. We use nuScenes dataset to evaluate the proposed method, and the experiments demonstrate that trajectory information can enhance the performance of online map construction in rasterized formats.