Autonomous camera-gimbal system for vision-based robotic applications

Abstract

Intelligent vision-based robots are used in various industries to provide increased flexibility and productivity. Their applications include the automation of complex tasks such as surveillance, mapping, and infrastructure inspection. Vision-based robots are equipped with an on-board camera system which plays a crucial role in performing vision-based tasks. For target tracking applications, the camera is usually attached to a gimbal device to improve its detection and tracking capabilities. The main challenge is to implement an efficient object detection method and pass the position of the detected object accurately to the gimbal control system. The objective of this project is to design a mobile robot equipped with an autonomous camera-gimbal system for real-time surveillance and target tracking applications. Object detection is implemented using deep learning-based object detection algorithms and the position of detected objects are passed to the gimbal control system. The detected targets are tracked autonomously using a gimbal control package based in the ROS environment. The gimbal control package can be implemented into any robot platform using ROS and communicate with any pre-existing ROS packages such as SLAM packages to further extend the capabilities of the mobile robot. The developed mobile robot is tested to detect and track target using different object detection models at various ranges. The results have been analyzed and the camera-gimbal system is optimized for real-time target detection and tracking applications while maintaining a good detection performance.