The mobility of the robot bring new challenge to computer vision. The objective is to research the real-time and robust vision system for robot. The main contents include:

• Target Recognition and Tracking

• Visual SLAM 

• Human-Robot Interface

Representative work:

Visual localization and navigation of mobile robot (National Natural Science Foundation and Joint Lab project with Microsoft Research Asia): A land-mark recognition algorithm based on local feature matching is proposed for localization and navigation of mobile robot. Visual navigation system with high robustness in natural unstructured environment is developed. This technique can be applied in mobile agents equipped with single camera or multi cameras, including robotics, intelligent vehicles, unmanned airship and group robots. 

Outdoor visual positioning and navigation (Joint project with Shandong Electric Power Research Institute): Develop a near-infrared based robust and precisely positioning and navigation system, for autonomous inspection robot working in outdoor environment. The system is equipped with near-infrared panoramic camera and panoramic infrared illuminator, Bar code land-mark based positioning and navigation method is proposed. The system has high positioning accuracy and it is real-time. The system has strong adaptability to outdoor environment and can work day and night, even in misty weather.

Ceiling feature based visual odometry

Lank mark based self-positioning and navigation system with mono-cular

visal localization and navigation for outdoor robot