Talk, Lead Workshop, Budapest
In this dynamic session, Dr. Chang Liu explores the transformative role of Artificial Intelligence in cutting-edge applications involving satellites, drones, and mobile robotics—specifically autonomous vehicles and robotic dogs.
Conference proceedings talk, 17th International Conference on Signal Image Technology & Internet based Systems (IEEE SITIS 2023), Bangkok, Thailand
This talk presents a method for optimal wildfire escape route planning specifically for drones, considering dynamic fire and smoke models. The approach includes:
Conference proceedings talk, The 26th IEEE International Conference on Intelligent Transportation Systems (ITSC 2023), Bilbao，Spain
This talk presents a method for wilderness rescue using UAVs and satellite imagery. The approach integrates UAV vision technology and satellite image analysis for:
Conference proceedings talk, IMPROVE 2023, Prague, Czech Republic
This work introduces a global multi-temporal satellite dataset to remove clouds and shadows using STGAN. The baseline STGAN achieves PSNR 33.4 and SSIM 0.929, producing high-quality cloud-free images for real-world applications. The dataset is publicly available.
Conference proceedings talk, Proceedings of 2nd International Conference on Image Processing and Vision Engineering - IMPROVE 2022, Online
This work presents an intelligent UAV-based route planning system for field environments. First, roads are extracted from real-time UAV images using D-LinkNet. Then, main and secondary roads are distinguished based on width and real-time road conditions, assigning different weights. Finally, a weighted A* algorithm calculates the optimal route between user-defined start and end points. The approach is validated through simulations on publicly available datasets and provides real-time optimal routes for people in the field.
Conference proceedings talk, IMPROVE 2021, Online
This work presents a UAV-based human gesture recognition system for rescue operations. The system can detect, track, and count humans, and recognize ten basic rescue gestures. Special attention is given to dynamic gestures Attention and Cancel, which set and reset communication with the UAV. Using deep learning, the system achieved 99.47% accuracy on training data and 99.09% accuracy on testing data.