Fixing CCTV security systems

The problem I would like to solve is that with traditional closed-circuit television (CCTV) security systems, humans miss too many events. This is because they can’t watch everything. If a person is monitoring traditional CCTV, about 98% of the footage is never watched. After about 12 minutes, they can miss over 45% of the activity, and after 22 minutes the amount they miss goes to around 95%.

This is important because if people are missing that much of the footage, they are probably missing a lot of events. We want to address this because if we don’t people could lose lots of money and possessions, and damage to property can also happen.

I plan to develop and test an AI system to detect threats. I will compare the performance of the AI system to human performance under a variety of conditions, such as positioning the camera low versus high or using visible light versus infrared light. I can test this by giving the AI and humans multiple videos of people doing normal things and committing crimes. I will start with some obvious ones and then move to some more inconspicuous ones and see who gets more of them correct.