License plate number recognition method and device

Abstract

The invention discloses a license plate number recognition method and device. The method comprises the steps: carrying out the license plate domain recognition of an image in a video of a law enforcement instrument through a trained license plate detector, and carrying out the rough correction of a license plate domain through Radon transformation; inputting the roughly corrected license plate domain into a license plate key point feature model for license plate key point feature positioning to obtain a precisely positioned license plate; performing secondary accurate correction on the license plate in a sequential affine transformation and projection accurate correction mode; and carrying out license plate number correction recognition on the corrected license plate by using an end-to-end license plate character recognition model and a license plate type matching template. According to the invention, the license plate is corrected through Radon transformation, affine transformation and projection precision correction modes, the defects of single correction are overcome, and the correction efficiency is improved; through a matching mechanism of an end-to-end license plate character recognition model and a license plate type matching template, the problem of wrong generation or missing generation of characters is avoided, and the complexity of multiple cyclic matching of license plates is reduced.

Description

technical field [0001] The invention belongs to the technical field of license plate number recognition, and relates to a license plate number recognition method and device. Background technique [0002] The intelligent license plate recognition method is a technology for license plate number recognition based on the license plate features in video images. This technology is widely used in traffic vehicle violation supervision, traffic accident forensic review and other fields. The vehicle license plate recognition technology route of the conventional bayonet can be summarized as follows: for an image or a video that is input and processed, the vehicle detector is first used to detect the vehicle in the video image at the first level, and then the vehicle area is classified as the second-level license plate area. Then input the license plate area to be checked into the key point feature model of the license plate to extract the key point features of the license plate area i...

AI Technical Summary

Problems solved by technology

[0003] However, the case handling environment of the mobile wearable body enforcement device is complex and changeable, and the video shooting is easily disturbed by factors such as light, shaking, wearing height and angle, which lead to serious lack of vehicle features in the video image, large deviation angle and deformation of the license plate, and increased Difficulty of license plate recognition in video images of law enforcement instruments

Accurate identification of license plate numbers is crucial to traffic violations and accident verification and evidence collection. However, relying on the traditional method of license plate feature extraction, the conventional bayonet vehicle first-level to license plate domain level-two detection mode and the single-slope license plate angle correction method are no longer effective. To solve the problems encountered in the license plate recognition of body-enforcement devices, there is an urgent need for an intelligent license-plate recognition technology solution for body-enforcement devices

Images