Human detection and tracking system: Law-enforcement officials are often stuck scrolling through hours of video in an attempt to catch a suspect, an activity that could be sped up with the right software. Unfortunately, current technology cannot reliably track multiple people or sort individuals from a larger group. Even a shadow can confuse programs—its movements might look like a second person. To build systems that work in more crowded environments, co-inventors Ram Nevatia, a computer scientist at the University of Southern California, and Bo Wu, now an engineer at Google, decided to zero in on body parts.
Patent No. 8,131,011 describes a software system that searches videos for human parts—the head, the torso or legs—and figures out how they fit together. Once the software finds a person in the front of a group, it can detect a second body by the head alone. The software can also track moving individuals by predicting where they might be in subsequent frames of the video. The next step is to tweak the software so that it recognizes activities—a function that could potentially catch a thief.