NEC technology locates similar objects across multiple videos
Via NEC News room
Nov 11, 2015
Tokyo, Japan, 11 November, 2015 - NEC Corporation (NEC; TSE: 6701) announced today the development of a "Profiling Across Spatio-Temporal Data" technology that rapidly classifies similar objects in multiple videos, allowing near real-time search of objects using criteria such as time, place and behavior.
In combination with NEC's facial recognition solutions, this technology can analyze more than 1 million facial images from a 24 hour-long surveillance video within just 10 seconds, and detect similar people appearing frequently and for long periods in the same place.
The new technology groups objects from multiple videos, such as people, based on a "level of similarity" and organizes the group with a tree-shaped data management structure. This tree-shaped data management structure allows high-speed search of similar objects by simply comparing the level of similarity of each group, instead of comparing an object one by one. In addition, this technology instantly classifies objects from new video footage and decides which groups the newly captured objects belong to.
Tree-shaped Data Management Structure
"This innovative technology allows public safety agencies to obtain new insight from a large amount of video data," said Yuichi Nakamura, General Manager, Green Platform Research Laboratories, NEC Corporation. "For example, this technology can help to prevent crime and speed up criminal investigations by detecting unauthorized people who appear frequently near restricted areas. We are planning to commercialize the technology within the next two years, while extending its application from video data to a variety of other data, including sound and text."