Real-time license plate detection and recognition method and device based on color augmentation

A real-time license plate detection and recognition method based on color augmentation, comprising the following steps:

• obtaining car images containing license plates as a license plate detection training set, transforming images containing blue license plates in the detection training set into images containing yellow license plates and green license plates through color augmentation, and then using a license plate detection model based on deep convolutional neural network for training to obtain a trained license plate detection model, inferring license plate detection results through the trained license plate detection model;
• performing view correction transformation on the detected license plates to obtain frontal view images of the license plates;
• using the obtained frontal view images of the license plates as a training set for license plate recognition, using a license plate recognition model based on a deep neural network for license plate recognition to obtain license plate recognition results by the following sub steps:
• collecting license plate recognition samples and creating the recognition training dataset;
• processing color augmentation on the recognition training dataset;
• training the license plate recognition model by using the color augmented dataset;
• testing the license plate recognition model and obtaining the license plate recognition results;
• displaying the license plate detection results and the license plate recognition results on the original test images or outputting the license plate detection results and the license plate recognition results, completing the detection and recognition of the license plates in the images;
• wherein, the color augmentation comprises the following process:
• reading the RGB color space of frontal view images M2 of the blue license plates in the recognition training dataset;
• subtracting each value in the RGB color space of images M2 from 255 to obtain images M3;
• at this time, the frontal view images M2 of the blue license plates is converted into the frontal view images M3 of the yellow license plates;
• converting the RGB color space of images M3 to the HSV color space;
• randomly generating a number n1 within a range of [9,51];
• adding n1 to each number in Hue (H) space and then modulo 180 to obtain H2;
• replacing H in the images M3 with H2 to obtain converted images M4;
• at this time, the frontal view images M2 of the blue license plates is converted into the frontal view images M4 of the green license plates.