WHEN IT COMES TO IMAGERY, volume is both a blessing and a curse. Thanks to satellites, drones, dash cams and body worn cameras, and IP-enabled monitoring and surveillance, government and commercial enterprises are enjoying a boom time in terms of both still images and video footage. New York City alone sought to roll out 18,000 body worn cams last year, while sales of IP video cameras are growing at 20 percent a year. For federal agencies, visual data delivers situational awareness, early warning, identity information, operational insight and a host of other valuable inputs.
At the same time, this sheer avalanche of information threatens to overwhelm traditional analytics approaches. As the intelligence community already knows, and others in government are fast learning, it’s simply beyond the ability of humans to watch and evaluate this nonstop flood of visual data. The human eye cannot pick up every subtle cue; human attention cannot remain fully focused and on task for as many hours as the job requires; nor is the present analyst workforce adequate to keep pace with the escalating demand. Something better is needed to assist humans, to help them focus on the moments and images requiring their scrutiny.
A sub-specialty within the broader field of Artificial Intelligence (AI), “computer vision” promises to cut through the clutter. With machines trained to understand visual images in the same way as humans do, government agencies could better leverage their investment in visual data while streamlining their own processes and freeing human talent for higher-level work.
While computer vision has been evolving for some time, it has lately come to the fore thanks to low-cost computing, an increase in training data and higher processing power. Here we will give a high-level overview of the technology behind computer vision. We’ll look at how computer vision is being deployed in the private sector, consider emerging federal use cases, and chart a path forward for government agencies looking to make the most of this fast emerging capability.