Multi-scale license plate precise locating method based on affine correction

Abstract

The invention discloses a multi-scale license plate precise locating method based on affine correction. Firstly, candidate license plate plaques are detected by using texture information and color features of license plates, then affine correction is carried out on each candidate license plate, an energy function is defined to accurately locate the positions of the license plates, noise is removed, and the positions of license plates in original images are obtained through back calculation. According to the multi-scale license plate precise locating method based on the affine correction, the license plates can be accurately located from multi-scale images in a complex environment, high robustness, accuracy and high-speed performance are achieved, and the omission rate of the license plates is close to zero. The affine correction is uniquely introduced, the license plate locating problem brought by view angles of the images is well solved, and a basis is established for the accurate locating of the license plates. Meanwhile, the energy function is originally defined to evaluate and accurately locate the license plates, more deep information beneficial to the locating of the license plates is used in the method, and the locating of the license plates is made to be more accurate, more stable and more reliable.

Description

technical field [0001] The invention belongs to the field of pattern recognition and computer vision, and in particular relates to a method for precise positioning of multi-scale license plates based on affine correction. Background technique [0002] License plate detection and positioning technology is one of the important research topics in the application of computer vision, digital image processing and pattern recognition technology in the field of intelligent transportation. It is widely used in road traffic monitoring, expressway automatic toll collection system, parking lot management system and many other traffic management system. With the rapid development of street view map products in recent years, it is necessary to detect, locate and blur license plates for the sake of privacy protection. Therefore, more efficient and robust license plate detection and positioning technology has become a hot research topic. [0003] At present, there are two main technical ro...

AI Technical Summary

Problems solved by technology

This method is greatly affected by lighting, weather, and shooting angles, and there are many colors and types of license plates. This method has obvious limitations.

[0006] At the same time, some methods based on statistics and learning have also been proposed, but the efficiency and accuracy are difficult to balance, and the missed detection rate and error rate are high, which is difficult to meet the actual needs

Because the actual environment of license plate detection and positioning is relatively complex, and there are interferences such as license plate viewing angle changes, lighting changes, weather conditions, motion blur, etc., the rapid and accurate detection and positioning of license plates is facing great challenges.

At the same time, there are multiple license plates in some images and the image size is large. The area occupied by the license plate has low resolution and partial occlusion, which also brings difficulties to the license plate detection and positioning.

Images