A license plate removal blur detection method based on deep learning

Abstract

A license plate deblurring detection method based on deep learning. This method is inspired by the YOLOv5 network and designs a brand new network: YD‑NET deblurring detection network. Use a high-speed camera to shoot videos of vehicles driving on multiple roads, and use the average continuous short-exposure frames to approach the long-exposure frames, so as to obtain a large number of image pairs containing clear license plates and blurred license plates, and obtain a data set after data processing; use transfer learning After loading some pre-trained weights of YOLOv5, train the YD-NET defuzzification detection network, and continuously adjust the hyperparameters until the optimal network is obtained; test the trained YD-NET model, and use the experimental indicators to evaluate the model performance. The invention can not only quickly detect the position of the license plate in the image in the natural scene, but also effectively remove the blur caused by the high-speed driving of the vehicle. It has been proved by theoretical analysis and experiments that the method has good performance and does not depend on the high price The high-end graphics card equipment can also meet the requirements of real-time processing and has promotional value.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a method for detecting license plate blur removal based on deep learning. Background technique [0002] Automatic license plate recognition (ALPR) is an important task with various related applications in the field of intelligent transportation and surveillance, such as automatic traffic enforcement, detection of stolen vehicles, violation charges, traffic flow control, etc. The ALPR problem can be divided into the following three subtasks: license plate detection (LPD), license plate segmentation (LPS) and character recognition (CR). [0003] At present, most license plate detection systems are only suitable for fixed scenarios, such as license plate detection at toll stations, parking lot management, etc. However, in natural scenes, there will be complex and changeable situations, and the captured license plate images will contain a lot of noise. It is difficult for ...

AI Technical Summary

Problems solved by technology

However, when the vehicle is driving at a high speed, the captured license plate image may be blurred, which will undoubtedly bring great difficulties to the license plate recognition system.

Images