Software for Car License Plates Recognition with Minimal Computing Resources

Abstract

The paper proposes a software for recognizing car license plates. The software is intended for integration into an autonomous module for installation on a gate at the entrance to the protected area. A feature of the software is the use of computer vision algorithms, the implementation of which requires minimal computing resources, since the video stream is being processed by a low-end CPU. The software is executed in the form of the caused library in language C ++ with use of standard functions of library of computer vision OpenCV. The software consists of following steps: image preprocessing and binarization, license plate localization, plate rotation and normalization, segmentation inside the license plate, segmentation result validation, text recognition. To verify and test the software, a camera was installed on the gate with the subsequent processing of the received video data. As a result of testing probability of correct recognition 0.96, probability of recognition error - 0.004, probability of missing - 0.035, probability of false recognition - 0.015, the frame processing time at a frame resolution of 3 MPix on an Orange Pi Pc 2 CPU with an Allwinner H5 Quad-Core ARM Cortex-A53 64 bit processor is 1.2-1.5 s.