The license plate recognition system is an application of computer image recognition and processing technology, which usually includes four parts:
Image acquisition pre-processing
Positioning correction
Character segmentation
Character recognition
Now the third-party platform has realized the function of license plate recognition, and provides an API interface, we only need to call the interface they provide to achieve it, but most of the third-party interfaces are charged, and most of them do not support local recognition, and they need to be uploaded to their platform, which will also be affected by the network speed.
If you consider the financial problem, then you can use the free and open source lightweight framework HyperLPR, its total code volume does not exceed 1K lines, supports Python, Android, C++ and other programming language calls, the accuracy of the bayonet scene is about 95%-97%, and you can refer to the documentation for the complete implementation process.