This paper presents a new method for updating high-resolution (HD) maps crucial for autonomous vehicle localization and environmental awareness. The method addresses inaccuracies in pre-built HD maps due to environmental changes by using real-time data from self-driving vehicles. It aligns new and old images, segments them to identify differences, and projects these changes onto a 3D point cloud. Experiments with data from the self-driving car Phenikaa-X show the method's effectiveness, with position deviations between 0.015 and 0.08 m. Performance metrics for vehicle detection include 89.4% accuracy, 75.6% recall, and 87.7% mAP, while object detection with protective boxes, which are the subjects for evaluation change shows 99.1% accuracy, 94.7% recall, and 97.4% mAP, enhancing the safety and reliability of HD maps for autonomous vehicles.