1. Crowd Surveillance via Range Sensors
Comparing with traditional image and video based crowd surveillance, range sensors bring new possibilities of monitoring both global people flow and individual movement in an accurate and efficient way. Our tracking algorithm is capable of tracking more than 170 pedestrians simultaneously (processed in real-time when the number of people is less than 50) and achieves satisfactory tracking accuracy in a series of field tests.
2. Spatial Data Assimilation of People Flow
Sensing people flow via various sensors attracts increasing attention from worldwide, which is critical for many potential applications. However, considering the cost it is difficult to collect sufficient information in practice. In many cases, only partial information in adjacent areas, or data of different kinds are available. Our purpose is to estimate and assimilate people flow from such incomplete data.
3. Urban/Village Mapping
Identifying residential areas in developing countries is an important topic which helps improving the management and quality of the residents’ life. We have been assisting the remote sensing group of our laboratory in recognizing urban/village areas, by carefully combining novel machine learning techniques with multi-source satellite images.