Real-time object detection serves as a foundational capability in autonomous robotic systems, directly impacting their ability to perceive, navigate, and interact with dynamic environments. Traditional 2D vision-based approaches, while computationally efficient, often struggle with challenges such as depth ambiguity, occlusion, and poor spatial understanding, particularly in unstructured or cluttered scenes. These limitations hinder the reliability and precision required for critical robotic applications. To address these shortcomings, this study explores the integration of 3D vision into the object detection pipeline, aiming to enhance spatial perception and detection accuracy. The proposed framework leverages stereo vision and depth mapping techniques to enrich visual data with depth cues, thereby enabling more informed decision-making in real-time contexts. A fusion-based architecture is developed, combining RGB input with corresponding 3D point cloud or depth map representations, and implemented using state-of-the-art detection models such as YOLOv8 and optimized through hardware-accelerated platforms. Experimental evaluations conducted on both benchmark datasets and real-world robotic scenarios demonstrate significant improvements in detection accuracy and robustness, particularly in depth-critical tasks such as obstacle avoidance and object manipulation. The integration of 3D vision not only enhances detection fidelity but also supports more resilient operation under variable lighting and environmental conditions. These findings underscore the potential of 3D vision-enhanced systems to elevate the capabilities of modern robotics, paving the way for more intelligent and context-aware autonomous agents.

Keywords: Real-Time Object Detection, 3D Vision, Robotics, Stereo Imaging, Depth Perception, Sensor Fusion