Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Landmark-based Localization using Stereo Vision and Deep Learning in GPS-Denied Battlefield Environment

Ganesh Sapkota, Sanjay Madria

TL;DR

The paper tackles GPS-denied battlefield localization by integrating passive stereo vision with deep landmark recognition. It detects landmarks via a fine-tuned YOLOv8s model and estimates landmark distances from stereo disparity, then localizes the unknown node using trilateration followed by L-BFGS-B optimization. The approach outperforms DV-Hop and competes with vision-based SLAM/VO methods, demonstrated on real-world MSTlandmark datasets with high localization accuracy (RMSE near 0.0147 m). This framework provides a practical, radio-free localization solution for dynamic, sparse battlefield environments with potential for extension to path planning and hazard avoidance.

Abstract

Localization in a battlefield environment is increasingly challenging as GPS connectivity is often denied or unreliable, and physical deployment of anchor nodes across wireless networks for localization can be difficult in hostile battlefield terrain. Existing range-free localization methods rely on radio-based anchors and their average hop distance which suffers from accuracy and stability in dynamic and sparse wireless network topology. Vision-based methods like SLAM and Visual Odometry use expensive sensor fusion techniques for map generation and pose estimation. This paper proposes a novel framework for localization in non-GPS battlefield environments using only the passive camera sensors and considering naturally existing or artificial landmarks as anchors. The proposed method utilizes a customcalibrated stereo vision camera for distance estimation and the YOLOv8s model, which is trained and fine-tuned with our real-world dataset for landmark recognition. The depth images are generated using an efficient stereomatching algorithm, and distances to landmarks are determined by extracting the landmark depth feature utilizing a bounding box predicted by the landmark recognition model. The position of the unknown node is then obtained using the efficient least square algorithm and then optimized using the L-BFGS-B (limited-memory quasi-Newton code for bound-constrained optimization) method. Experimental results demonstrate that our proposed framework performs better than existing anchorbased DV-Hop algorithms and competes with the most efficient vision-based algorithms in terms of localization error (RMSE).

Landmark-based Localization using Stereo Vision and Deep Learning in GPS-Denied Battlefield Environment

TL;DR

The paper tackles GPS-denied battlefield localization by integrating passive stereo vision with deep landmark recognition. It detects landmarks via a fine-tuned YOLOv8s model and estimates landmark distances from stereo disparity, then localizes the unknown node using trilateration followed by L-BFGS-B optimization. The approach outperforms DV-Hop and competes with vision-based SLAM/VO methods, demonstrated on real-world MSTlandmark datasets with high localization accuracy (RMSE near 0.0147 m). This framework provides a practical, radio-free localization solution for dynamic, sparse battlefield environments with potential for extension to path planning and hazard avoidance.

Abstract

Localization in a battlefield environment is increasingly challenging as GPS connectivity is often denied or unreliable, and physical deployment of anchor nodes across wireless networks for localization can be difficult in hostile battlefield terrain. Existing range-free localization methods rely on radio-based anchors and their average hop distance which suffers from accuracy and stability in dynamic and sparse wireless network topology. Vision-based methods like SLAM and Visual Odometry use expensive sensor fusion techniques for map generation and pose estimation. This paper proposes a novel framework for localization in non-GPS battlefield environments using only the passive camera sensors and considering naturally existing or artificial landmarks as anchors. The proposed method utilizes a customcalibrated stereo vision camera for distance estimation and the YOLOv8s model, which is trained and fine-tuned with our real-world dataset for landmark recognition. The depth images are generated using an efficient stereomatching algorithm, and distances to landmarks are determined by extracting the landmark depth feature utilizing a bounding box predicted by the landmark recognition model. The position of the unknown node is then obtained using the efficient least square algorithm and then optimized using the L-BFGS-B (limited-memory quasi-Newton code for bound-constrained optimization) method. Experimental results demonstrate that our proposed framework performs better than existing anchorbased DV-Hop algorithms and competes with the most efficient vision-based algorithms in terms of localization error (RMSE).
Paper Structure (19 sections, 10 equations, 18 figures, 4 tables, 2 algorithms)

This paper contains 19 sections, 10 equations, 18 figures, 4 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Our Approach
  4. Stereo Vision Method
  5. Landmark Detection using CNN
  6. Trilateration and Node Localization
  7. Optimization of Estimated position:
  8. Experiments
  9. Experimental Setup
  10. Dataset
  11. MSTlandmarkv1 Dataset:
  12. MSTlandmarkStereov1 Dataset
  13. Trainnig Model
  14. Training Results and Discussion
  15. Distance Measurement Using Stereo Vision
  16. ...and 4 more sections

Figures (18)

  • Figure 1: Visualization of landmark anchor-based localization on a map. The green pin indicates the graphical landmarks detected by the landmark recognition model. The blue pin indicates the actual position of a mobile node and red solid circles indicate the estimated positions of the node using the proposed framework
  • Figure 2: Overview of DV-Hop localization algorithm
  • Figure 3: System Overview: Distance estimation and landmark recognition steps and the results after two operations are fused to calculate the position of an unknown node.
  • Figure 4: Demonstration of Disparity in Stereo image Pair
  • Figure 5: Disparity and Depth Map Generation using Stereo Matching
  • ...and 13 more figures