Vision assisted object detection In LIDAR point cloud

• Main Author: Yuen, Wei Chee
• Other Authors: Xie Lihua
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/158702
• Institution: Nanyang Technological University

In the age of Industry 4.0, the usage of autonomous guided robots has become a commonplace, especially in logistics, transport and manufacturing. With more labor-intensive operations being highly automated as a solution to improve the efficiency of manufacturing processes, autonomous guided vehicles (AGVs) are deployed to facilitate transportation of materials and finished products. The key to success is to develop a robust and secure avoidance policy for robots, therefore ensuring the safe maneuverability of the robot. The usage of a 2D LiDAR only for object collision avoidance results in frequent stop in navigation due to lack of target identification. Therefore, this project focuses on the training of a lightweight object detection model and the alignment of a LiDAR point cloud with the object detection model based on the live video footage from the RGB camera to provide a vision assisted object detection to enable the AGV to navigate and avoid obstacles in a dynamic environment. It is able to continuously track the target object , while navigating through a dynamic environment, avoiding obstacles.