In the Australian state of Victoria, the authorities decided to create a system for searching for stolen cars. The idea is to equip 220 vehicles with cameras that can recognize the signs of passing vehicles and check whether they are wanted. The creation of the system is estimated at $ 70 million, that is, equipping each car should cost $ 300,000.

The Australian programmer Tate Brown at leisure decided to try to do something like that. He decided that the signs of machines should be recognized locally, so as not to spend money on transferring video from the camera; The system should not use commercial programs and should recognize images of poor quality.

To recognize signs, Tate Brown used an open open-source library specially created for this. It is enough to transmit images or video from the camera, so that it translates the characters into text. The request to the library can be literally put in a couple of lines of code.

Tate Brown also wrote a program that takes the recognized license plate and punches it through an already existing system for checking cars for theft. The whole program took dozens of lines of code.

For recognition, the programmer used video from the DVR.

As it turned out, the signs were determined quite accurately, despite the fact that the openalpr library was made for American numbers, not Australian.

As Tate Brown noted in his post on Medium, he understands that the system that the authorities want to create is much more complicated than he came up with. In particular, a lot of money will have to be spent on data storage and, perhaps, on training algorithms to increase the recognition accuracy. Nevertheless, Brown believes that the system can do less than $ 70 million.