Coarse-to-fine license plate detection algorithm and coarse-to-fine license plate detection system

Abstract

The invention discloses a coarse-to-fine license plate detection algorithm and a coarse-to-fine license plate detection system in the field of computer vision, and the algorithm comprises the following specific steps: S1, carrying out the downsampling of an input image based on the size of a to-be-detected minimum vehicle; S2, detecting a vehicle area of the down-sampled image by using a convolutional network I; S3, performing downsampling on the detected vehicle area image based on the minimum license plate size to be detected; S4, detecting a candidate license plate region of the downsampledvehicle region image by using a convolutional network II; and S5: for the candidate license plate area detected in the step S4, obtaining a license plate area. Proportioning resolution, detecting a license plate by using a convolutional network III; the system comprises a data collection module, a license plate detection module, an output module and a model training module, the used convolutionalnetwork structure is relatively simple, the required calculation amount is far less than that of the existing license plate detection algorithm, and the system can be effectively applied to high-speed no-parking charging and other scenes with high requirements on the real-time performance of license plate detection.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a coarse-to-fine license plate detection algorithm and a system thereof. Background technique [0002] At present, license plate detection and recognition have been widely used in parking lots, traffic crossing violation detection and other fields. With the gradual development and application of smart cities, free parking tolls on expressways have become a trend. This application scenario has high requirements for the real-time performance of license plate detection and recognition. If the detection takes too long, the car still needs to wait for the system to respond, which cannot achieve the purpose of passing quickly without stopping, and may easily cause traffic jams. [0003] At present, most mainstream license plate detection algorithms use deep convolutional networks to detect license plates on the input image. To ensure the accuracy of detection, the network is ...

AI Technical Summary

Problems solved by technology

If the detection takes too long, the car still needs to wait for the system to respond, which cannot achieve the purpose of passing quickly without stopping, and is likely to cause traffic jams

[0003] At present, most mainstream license plate detection algorithms use deep convolutional networks to detect license plates on the input image. To ensure the accuracy of detection, the network is generally more complex, and the calculation time required for detection is also relatively long. It is difficult to achieve real-time in the case of fast passing sexual needs

Images