• camera
• detection
• fusion
• laser
• obstacle
• RGBD
• sensor
• tracking

A self-driving vehicle, to be deployed in indoor driving environments, must be capable of reliably detecting and effectively tracking of nearby moving objects. This thesis presents a moving object detection and tracking algorithm from an RGB-D camera and a 2D Laser sensor. The system is capable to detect and tracking objects separately from a camera and a laser, then combining the lists of tracks with a simple sensor fusion approach. The images from the camera are also used to classify the objects through the YOLO network, and this information is also included in the final list of tracks. Dempster Shafer theory is used for data association and a comparison with the global nearest neighbor is presented. The algorithm was implemented on ROS and tested with Gazebo simulations, giving fulfill results.