Video can be a powerful and increasingly abundant resource for improving public safety, yet safety officials and investigators face numerous challenges when analyzing it. Accurate detection, identification and “re-identification” of individuals in video footage is paramount, even when the quality of the footage is very low. Poorly tuned or older legacy devices may record and store video that is inadequate for this task, creating a need for more effective analytic tools that can automatically extract actionable information from protective and investigative video.
CINA researchers at Purdue University and the University of Notre Dame are developing a system to help DHS components extract, summarize and reconstruct critical information. By using both face and full-body re-identification methods, the tool is now producing reliable results.
This system, scheduled for implementation in the fall of 2019, includes a highly integrated user interface that allows officials to interact with video data and visualize ranked matching results.
As Principal Investigator Ed Delp puts it, “We are addressing the problem of what can be done at the ‘edge’ of the network close to the video camera. This solution will allow DHS to minimize the amount of video data they need to collect.”
CINA is evaluating the system with DHS partners, who are providing important feedback to guide further development. Once the evaluation period is complete, we anticipate broader implementation as the tool is shared throughout the DHS network. The project should result in additional efficiencies, reducing the time and cost of investigations, and providing access to information that was previously unavailable.