If a person taking a bicycle doesn't match an image of who parked it then a theft might be in progress.

PhD student Dima Damen, from the University’s Faculty of Engineering has developed a computer system that detects individuals parking their bicycles and can automatically warn security staff if it appears that someone other than the owner retrieves the vehicle.

...

Currently at prototype stage, Damen’s system takes colour information from CCTV images when a bike is parked and stores it until the bike is retrieved. It then marries the stored information with the new image and where there are significant differences, it can raise an alert to CCTV operators. In initial tests using a camera located above a bike rack at the University of Leeds, eleven out of thirteen simulated thefts were detected.

This approach seems like it might work for cars as well. Extended further, cameras trained on a street with image processing algorithms could alert humans when someone enters a building who has never been recorded entering that building before.

I am reminded of science fiction movies where small flying police monitor cameras watch people. When it comes to suspected thieves such mini flying cameras could be dispatched to record higher resolution images and even to follow a driver of a car or bicycle to track where they go.