Using the ARCore library to visualize keypoint clouds in navigation systems

Abstract

The article explores the prospects of using the ARCore to create navigation systems tailored for visually impaired individuals. The ARCore is an augmented reality platform that combines visual and inertial odometry to ensure precise device localization in space. The authors describe the key principles of the ARCore, including methods for extracting and tracking key points, integrating IMU readings, and calculating movement trajectories.

Special attention is given to the architecture of the navigation system, which relies on smartphones to perform computational tasks such as SLAM algorithm implementation and neural network inference. The system processes data locally on the smartphone and transmits the results to a tactile feedback module via Bluetooth. This approach ensures the affordability and compactness of the final product.

To improve data accuracy, the authors propose the use of clustering algorithms such as DBSCAN, Local Outlier Factor (LOF), and the Kalman filter. These methods are aimed at filtering out noise and stabilizing data, which is particularly crucial for building effective navigation systems.

In conclusion, the authors highlight that combining the ARCore with neural networks and modern data processing algorithms opens new possibilities for creating accessible and functional solutions to improve the mobility of visually impaired individuals. Future research will focus on optimizing algorithms, training neural networks on specialized datasets, and experimentally testing the system in real-world conditions.