Technicians from ZoheTec test an UnstructuredCam intelligent imaging system on the roof of a building in Binhai New Town in Fuzhou, Fujian province. [Photo/China Daily]

'Smart coast guard' keeps an eye on things at wetland reserve in Fujian

Binhai New Town in the southeastern part of Fuzhou, capital of Fujian province, is blessed with beautiful sea views and a 35-kilometer coastline. It is also home to a national wetland reserve, which serves as a habitat for several endangered migratory birds, including the Chinese crested tern and the black-faced spoonbill.

An intelligent imaging system was introduced last month to monitor and protect the environment in the coastal area. Using two Gigapixel-level smart cameras, an innovative unstructured camera array known as an UnstructuredCam and a set of gigapixel-level complex scene visual analysis solutions, the system now acts as a kind of "smart coast guard".

The UnstructuredCam is capable of making high-resolution, real-time and full color video recordings of the activities of large crowds of people spread over a wide area, according to Fang Lu, an associate professor at Tsinghua University's Department of Electronic Engineering. Fang leads the team working with a company called ZoheTec to develop and run the system in Binhai.

Inspired by human cognitive functions, Fang said she and the team developed the unstructured array camera system by adopting a hierarchical modular design with multi-scale hybrid cameras making up the different modules. Intelligent computation was also introduced into the system.

Unlike conventional single cameras, the UnstructuredCam removes the optical limitations of a single camera and paves the way for high-performance imaging by making use of both micro-cameras and computation to stitch together multiple images. "This system reduces hardware costs, enables high-resolution and long-term monitoring of dynamic wide-area scenes and performs multi-scale, human-centric studies for the assessment of human behavior at the individual level and crowd behavior at the mass level," Fang said.

To protect privacy, she added that the high-quality data obtained will be processed to remove faces before being transmitted to the public cloud.

Fang said that a system capable of delivering high-resolution wide-FoV video data has great potential for predicting crowd dynamics, which plays an important role in risk prevention for large-scale events.

"Based on the joint global and local observation of dynamic large-scale scenes, our system can predict the abnormal behavior of an individual, group or crowd by detecting and tracking thousands of parallel targets over long distances within 1 kilometer," she said.