Right track - A google map companion using junction recognition

Tan, Yong Ming (2023) Right track - A google map companion using junction recognition. Final Year Project, UTAR.

Preview

PDF Download (2021Kb) | Preview

Abstract

The challenge of navigating complex road systems, especially in urban environments, necessitates innovative solutions to enhance driver safety and confidence. This project explores the development of a computer vision system aimed at detecting road junctions in real-time, providing drivers with timely and accurate guidance. The system comprises two key components: junction detection and image similarity comparison. Traditional object detection metrics, such as Intersection over Union (IOU), are ill-suited for the intangible nature of junctions. As a solution, we propose the use of accuracy as an alternative evaluation metric to assess the model's ability to classify frames as 'junction present' or 'no junction.' Ground truth labeling of test images as '1' or '0' is performed, facilitating accuracy evaluation. This project's computer vision model demonstrated significant progress in junction detection accuracy, enhancing driver safety and navigation. Challenges encountered provide valuable insights for future refinement, particularly in optimizing cloud-based processing efficiency. The findings contribute to the advancement of intelligent navigation systems in complex urban environments.

Actions (login required)