A night license plate detection method and system based on efficient channel perception and feature fusion

By employing an efficient nighttime license plate detection method based on channel perception and feature fusion, and utilizing the ResNet101 network and feature pyramid structure, the problem of low efficiency and insufficient accuracy in nighttime license plate detection is solved, achieving efficient and accurate license plate detection.

CN122199906APending Publication Date: 2026-06-12NANJING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NANJING UNIV OF POSTS & TELECOMM
Filing Date
2026-01-27
Publication Date
2026-06-12

AI Technical Summary

Technical Problem

Existing license plate detection technologies are inefficient and inaccurate in low-light environments at night. Traditional methods perform poorly in complex scenarios, while deep learning models have high computational overhead and insufficient robustness.

Method used

A nighttime license plate detection method based on efficient channel perception and feature fusion is adopted. Feature extraction is performed through ResNet101 residual neural network, combined with an efficient channel attention mechanism, and feature fusion is performed using hierarchical feature pyramid and path enhancement module. Target recognition is performed using a sparse-constrained weighted loss function and progressive scale expansion.

Benefits of technology

It improves the accuracy and robustness of nighttime license plate detection, shortens the feature information transmission path, enhances the model's ability to handle imbalanced data, and meets the requirements of real-time detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122199906A_ABST
    Figure CN122199906A_ABST
Patent Text Reader

Abstract

The application discloses a night license plate detection method and system based on efficient channel perception and feature fusion, belongs to the field of artificial intelligence technology, and constructs a license plate detection network model, wherein the license plate detection model comprises a feature extraction module, a feature fusion module and a target recognition module; the feature extraction module adopts a Resnet101 residual network in combination with efficient channel perception to perform multi-scale feature extraction on an input image; the feature fusion module shortens deep and shallow level information paths through a double-tower structure of FPN+PA, and fuses features to output a single feature map; the target recognition module improves the stability of a convergence process through an STL loss function, expands feature map information through a breadth-first algorithm, and obtains optimal license plate detection results. The application can significantly improve model detection precision while meeting real-time detection requirements, and enhances the robustness of the license plate detection model in complex scenes.
Need to check novelty before this filing date? Find Prior Art