To navigate in space, high-resolution (HD) maps are essential for autonomous vehicles operating at level 3 and above. Improving the accuracy during HD map construction is a crucial research topic, involving various methods such as developing mapping algorithms, noise filtering, and preprocessing data before map creation. In this study, a model for identifying moving objects during map construction is investigated to eliminate unnecessary information prior to map building. The process of object detection and segmentation before and after map construction is evaluated and compared. Real-world testing results demonstrate that by identifying and removing moving objects, HD maps achieve increased accuracy, enhancing the vehicle’s position determination with the mean error improving by approximately 30% in various scenarios. This research also opens up new possibilities for real-time data collection and HD map construction in complex traffic environments, It demonstrates significant potential for improving autonomous vehicle navigation and map accuracy in challenging conditions, with a focus on enhancing practices in Vietnam.