Deep learning assisted real-time object recognition and depth estimation for enhancing emergency response in adaptive environment

Abstract

Accurate long-range object recognition is essential in autonomous navigation and military surveillance applications. While recent advancements have improved real-time recognition, existing models, especially those focused on monocular depth estimation, face accuracy challenges due to supervised Deep Learning (DL) limitations. This study presents a robust, real-time military object recognition system that leverages temporal sequences and attention mechanisms for enhanced depth estimation. Using RGB frames along depth maps from the KITTI and synthetics dataset, along with a fine-tuned YOLOv11 model, our system achieves a Root Mean Squared Error (RMSE) of 1.24 meters, and RMSE (log) of 0.18 in-depth estimation, with object detection adequate up to 250 meters.The model maintains high precision (96.4%), recall (93.67%),and F1 score (93.33%) across various ranges, confirming YOLOv11's accuracy with an average inference time of 13 ms for short-range and 17 ms for long-range detection. These results highlight the system's potential for deployment in real-time military and adaptive response scenarios, outperforming existing models in both accuracy and computational efficiency.

Keywords

• Object detection
• Depth estimation
• Real-time systems
• Deep learning
• Temporal information
• LSTM