This work plan proposes the development and analysis of a system in which a robot crosses through an environment, capturing images from a thermal câmera to create a 3D map of the environment and to track objects in the scene using monocular vision. The map created from the processing of images will be represented as a point cloud and will be used to track objects in the scene. The object tracking will be developed using pulse coupled neural networks and image matching techniques. The constructed system will be validated in simulation environment and tested in air robots.
News published in Agência FAPESP Newsletter about the scholarship: