Research

Monocular SLAM and Moving Object Tracking

Traditional monocular simultaneous localization and mapping (SLAM) problems using a single camera assume that the objects in the environment are static to estimate the 3D position of the objects. However, the static environment assumption restrict the practicability of these methods. In dynamic environment, we hope to not only detect the moving objects but track and estimate the 3D position of the moving objects as well. This definitely extends the ability of robots. One of the challenges is that the detection of the moving object is hard since all objects look like moving in the video taken by a moving camera. Another difficulty is to estimate 3D position of the moving objects when the camera itself is moving.

Tour in Mood of Place

It would be desirable if the touring system presents conventional information as well as the memories of others visiting the same place. For example, a night market in Taiwan is populous at night and empty in the day time. If you unfortunately visit it in the morning, the system provides the night market atmosphere in the right time as these are most people's memories about this place. It's like you are interacting with a live place telling you his stories and life.