License Plate Classification Using Appearance and Character Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing license plate classification technologies face challenges in accurately distinguishing license plates from different nations due to similarities in design and limited data availability, particularly for geographically close nations with complex variations, leading to low classification accuracy.
Innovation Solution
A method utilizing a two-branch network combining Convolutional Neural Networks (CNN) for appearance features and Transformer for character connections, with one-hot encoding to integrate character content, to enhance classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning is used for license plate classification, then classification accuracy is improved and applicability to various types is improved, but data collection difficulty increases and generalization ability requirements increase
Solution Approach 1:
The patent segments the classification task into two distinct feature extraction pathways: one for appearance features (color, shape, layout) and another for character content features. This segmentation allows the system to process different types of information separately and fuse them later, reducing the burden on data collection for each individual pathway while maintaining high overall accuracy.
Solution Approach 2:
The patent introduces an intermediary feature fusion mechanism that combines appearance features and character content features. This intermediary layer acts as a mediator that integrates information from multiple sources, allowing the system to leverage partial data from different nations while achieving robust classification through the combined representation.
2Device complexity
If traditional image processing method is used, then data collection requirement is reduced, but classification accuracy deteriorates
Solution Approach 1:
The patent merges traditional image processing techniques with deep learning approaches by combining appearance feature extraction (using traditional CV methods) with character content recognition (using OCR and deep learning). This merging allows the system to benefit from both the robustness of traditional methods and the accuracy of deep learning, achieving high classification accuracy without requiring exhaustive data collection.
3Adaptability or versatility
If license plates of geographically close nations are classified, then classification coverage is improved, but distinction difficulty increases due to high similarity
Solution Approach 1:
The patent segments the license plate analysis into appearance features (color, shape, layout) and character content features. For geographically close nations with similar designs, the character content segmentation provides discriminative information that appearance alone cannot provide, enabling successful distinction while maintaining broad classification coverage.
Solution Approach 2:
The patent adds another dimension to the classification by incorporating character content recognition alongside appearance analysis. This dimensional expansion transforms the classification from relying solely on visual appearance to a multi-dimensional approach that includes textual information, thereby improving the ability to distinguish between similar license plates from different nations.
4Adaptability or versatility
If nations with large number of license plate types are classified, then classification versatility is improved, but classification accuracy deteriorates due to limited variations between classes and great variations within classes
Solution Approach 1:
The patent segments the feature extraction into appearance and character content components. This segmentation allows the system to capture both the limited variations between classes (appearance) and the great variations within classes (character content), thereby maintaining high classification accuracy even when dealing with nations that have numerous license plate types.
Solution Approach 2:
The patent introduces character content as an additional dimension to the classification space. This dimensional enhancement provides the necessary discriminative power to handle nations with large numbers of license plate types, where appearance alone is insufficient to distinguish between the many variations within each class.
Data Source
AI summary
A license plate classification method, a license plate classification apparatus and a computer-readable storage medium are provided. The method includes: performing a license plate recognition process on a first license plate image to obtain a license plate recognition result; performing an encoding process on the license plate recognition result to obtain a first license plate feature; performing a feature extraction process on the first license plate image to obtain a second license plate feature; and processing the first license plate feature and the second license plate feature through a classification network to obtain a first license plate classification result. In this way, an accuracy of license plate classification is improved.