Landmark Stereo Dataset for Landmark Recognition and Moving Node Localization in a Non-GPS Battlefield Environment

TL;DR

This work tackles non-GPS localization on GPS-denied battlefields by introducing landmark anchors as virtual coordinates derived from Yolov5 landmark recognition and stereo depth estimation, enabling node localization without radio anchors. The approach relies on a mobile stereo vision node and an on-device control server to detect landmarks, estimate distances via an improved semi-global matching pipeline, and store (landmarkID, distance) tuples as virtual coordinates for subsequent trilateration. To support this, the authors create two real-world datasets, MSTLandmarkv1 for landmark detection and MSTLandmarkStereov1 for distance estimation, achieving a Yolov5mAP of 0.95 at IoU 0.5 and RMSE of 1.79 m for distances in the 33–78 m range. They demonstrate a vision-only localization pathway suitable for GPS-denied environments and outline future work on trilateration-based node localization and safe-path verification, highlighting practical implications for battlefield navigation and troop survivability.

Abstract

In this paper, we have proposed a new strategy of using the landmark anchor node instead of a radio-based anchor node to obtain the virtual coordinates (landmarkID, DISTANCE) of moving troops or defense forces that will help in tracking and maneuvering the troops along a safe path within a GPS-denied battlefield environment. The proposed strategy implements landmark recognition using the Yolov5 model and landmark distance estimation using an efficient Stereo Matching Algorithm. We consider that a moving node carrying a low-power mobile device facilitated with a calibrated stereo vision camera that captures stereo images of a scene containing landmarks within the battlefield region whose locations are stored in an offline server residing within the device itself. We created a custom landmark image dataset called MSTLandmarkv1 with 34 landmark classes and another landmark stereo dataset of those 34 landmark instances called MSTLandmarkStereov1. We trained the YOLOv5 model with MSTLandmarkv1 dataset and achieved 0.95 mAP @ 0.5 IoU and 0.767 mAP @ [0.5: 0.95] IoU. We calculated the distance from a node to the landmark utilizing the bounding box coordinates and the depth map generated by the improved SGM algorithm using MSTLandmarkStereov1. The tuple of landmark IDs obtained from the detection result and the distances calculated by the SGM algorithm are stored as the virtual coordinates of a node. In future work, we will use these virtual coordinates to obtain the location of a node using an efficient trilateration algorithm and optimize the node position using the appropriate optimization method.

Figures (19)

• Figure 1: Node to Landmark Distance Estimation and Virtual coordinate of Node obtained using Stereo Vision Method
• Figure 2: Stereo Vision Method for Depth Measurement
• Figure 3: Disparity in stereo images
• Figure 4: Area Selected For Landmark Dataset Creation. The blue pin indicates the landmarks and the red solid circle indicates the position of the stereo camera while capturing the images.
• Figure 5: Custom Stereo Vision Setup