MUMBAI, India, Oct. 5 -- Intellectual Property India has published a patent application (202541028411 A) filed by Jaishree. B; Kalimath Jumana T R; Kumaran K; and Gunalan K on March 26, 2025, for Deep Learning-Ready Dataset Creation: A Fusion Approach With Oak Vision And Imaging Radar.
Inventors include Jaishree. B; Kalimath Jumana T R; Kumaran K; and Gunalan K.
The application for the patent was published on October 02, 2026, under issue no. 40/2026.
Abstract: The present invention introduces a real-time multi-sensor fusion system integrating radar and camera modules for enhanced object detection and environmental perception. The system leverages ROS2 for seamless data acquisition, processing, and synchronization between the 4D imaging radar and the OAK-D camera. A deep learning-based detection model is employed, using YOL0v8 trained on annotated images to improve reflector identification and dynamic object tracking. The collected radar point clouds and camera frames are processed through dedicated ROS2 nodes, enabling structured data extraction. The system further supports adaptive calibration to optimize sensor alignment, ensuring robust multi-modal fusion for autonomous navigation and industrial monitoring. Additionally, the invention features an automated data processing pipeline, converting ROS2 bag files into MCAP format for visualization in Foxglove Studio and further transforming the data into CSV format for in-depth performance analysis. The fusion framework enhances situational awareness by utilizing machine learning techniques to refine detection accuracy through transfer learning. The system's modularity allows for seamless integration with additional sensors and AI-driven perception models, making it a scalable and adaptable solution for autonomous systems, robotics, and smart mobility applications. Keywords: Radar module calibration, ROS2 integration, Docker container, Radar connectivity, Rad detection data, Point cloud data, Sensor fusion, Depth AI OAK-D camera, Machine learning model, Objec detection, Reflector detection, Annotated images, YOL0v8 'training, ROS2 bag recording, MC . . conversion, Foxglove Studio visualization, CSV data processing, Data plotting, Autonomous perception Multi- sensor fusion.
Disclaimer: Curated by HT Syndication.
