The visually impaired find it extremely difficult to navigate in strange environments at the time when door detection and localization are life-critical in the case of earthquake or fire events, not only in everyday life. VisionGuard is a new computer vision system which provides real-time door detection and navigation with a single-camera depth estimation. The system incorporates a user-trained YOLOv8 object detection model, using monocular depth estimation techniques to detect doors and obstacles, estimate their distance, and give voice guidance. Our upgrade solution achieves 94% accuracy in door detection and its separate parts (handle, knob, hinge). The system has incorporated superior door state detection (open, closed), complete obstacle avoidance, and ergonomics. Physical placement of the devices to provide comfort and safety to the user. We look at the way the system works under different indoor settings and lighting systems and offer it as a practical aide technology. VisionGuard is a convenient, affordable, and easy-to-use solution to improve the portability and independence of visually impaired users without the need for special equipment or environmental changes.