Abstract:
A recent development of 3D lidar technology have deepened challenges in the field of point cloud processing. Many established algorithms and methods handle the task of detection and tracking of moving objects, simultaneous localization and mapping and combination of the two. Nowadays, there exists a wide variety of approaches mostly depending on the assumed environment (e.g. indoor, outdoor, on roads, cross-country, airborne, underwater) and the expected velocity of the platform with mounted sensor. Furthermore, ensuring real-time execution and high efficiency of such algorithms and methods, enabling potentially high velocities of platforms represents quite a challenging task. This talk gave a short overview of state-of-art in the field of detection and tracking of moving objects. Afterwards, a developed application for this task using 3D laser range sensor was presented. The proposed detection pipeline consists of ground extraction, downsampling of the point cloud and the detection of dynamic parts of space. Upon this, the dynamic objects are extracted. The following step, mainly the tracking task, uses joint probabilistic data association filter and Kalman filtering with entropy based track management. Within proposed tracking approach an adaptive process and measurement noise, that inherently take into account characteristics of used sensor, are modelled. At last, some experimental results were shown.
CV:
Josip Ćesić has received his BSc and MSc degree in electrical engineering and information technology from University of Zagreb, Faculty of Electrical Engineering and Computing (UNIZG-FER) in 2011 and 2013, respectively. He fisnished the third semester of his master program at Chalmers University of Technology, Sweden, as an exchange student. He is currently employed at the Department of Control and Computer Engineering at UNIZG-FER as a research engineer within AMOR research group. His main research interests are in the areas of mobile robotics, estimation theory and image processing.