Development of the Detecting System of the Landmark Tags to Increase the Navigation Accuracy of an Unmanned Vehicle in a Known Location

Abstract

This paper proposes the landmark detection system for use in unmanned vehicle based on NVIDIA Jetson embedded device. The paper describes the synthetic dataset generation and training YOLOv4 neural network using data augmentation and transfer learning techniques to detect the landmark in incomplete data conditions. The real-time landmark detection module was implemented on the NVIDIA Jetson Xavier. The final mean average precision was resulted as 83.8%

Keywords

• computer vision
• data processing
• artificial intelligence
• nvidia jetson