Adaptive License Plate Character Segmentation via Iterative Threshold Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ALPR systems face challenges in character segmentation due to noise sources like shadows, non-uniform illumination, and optical geometries, leading to incomplete segmentation, which results in missing characters and reduced OCR accuracy.
Innovation Solution
An adaptive character segmentation method that iteratively adjusts segmentation threshold parameters based on OCR confidence metrics to ensure a minimum number of valid characters are identified, using a histogram projection approach and feedback loops to increase aggressiveness until a believable result is achieved, preventing over-segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a fixed segmentation threshold is used in the first stage, then computational resources are minimized and processing speed is maintained, but character segmentation accuracy deteriorates when two or more characters are missing
Solution Approach 1:
The patent implements dynamic threshold adjustment by introducing a feedback loop where the segmentation threshold is adaptively modified based on the number of valid characters detected. When OCR validation identifies missing characters, the system automatically adjusts the segmentation threshold and re-processes the image, transforming a static segmentation process into a dynamic one that responds to validation results.
Solution Approach 2:
The patent employs feedback mechanisms where OCR validation results are fed back into the segmentation process. The system uses the confidence metrics and character count from OCR to determine whether to adjust segmentation parameters and re-execute segmentation, creating a closed-loop control system that continuously improves segmentation accuracy based on downstream validation.
2Reliability
If the second stage applies additional analysis to recover missing characters, then character segmentation accuracy is improved, but computational resources increase and processing time is extended
Solution Approach 1:
The patent implements self-service by enabling the segmentation system to automatically detect its own deficiencies through OCR validation and self-correct by adjusting thresholds and re-segmenting. The system monitors its own output quality and autonomously initiates corrective actions without requiring external intervention or complex manual recovery procedures.
Solution Approach 2:
The patent applies preliminary action by performing a first pass segmentation with conservative thresholds to ensure no characters are missed, then using OCR validation to identify gaps before final processing. This preliminary segmentation with built-in validation allows the system to detect and correct issues early in the processing pipeline.
3Measurement precision
If conservative segmentation thresholds are used, then false positive segmentation is reduced, but the number of valid characters detected decreases
Solution Approach 1:
The patent transforms static segmentation thresholds into dynamic parameters that adapt based on validation feedback. The system starts with conservative thresholds to ensure precision, then automatically relaxes thresholds in subsequent iterations when validation indicates missing characters, allowing the same system to achieve both high precision and complete character detection.
Solution Approach 2:
The patent implements periodic re-segmentation with adjusted thresholds based on validation results. Rather than using a single fixed threshold, the system periodically re-evaluates and re-segments the image with modified parameters, creating a rhythmic cycle of segmentation-validation-adjustment that ensures both precision and completeness.
Data Source
AI summary
Methods, systems and processor-readable media for adaptive character segmentation in an automatic license plate recognition application. A region of interest can be identified in an image of a license plate acquired via an automatic license plate recognition engine. Characters in the image with respect to the region of interest can be segmented using a histogram projection associated with particular segmentation threshold parameters. The characters in the image can be iteratively validated if a minimum number of valid characters is determined based on the histogram projection and the particular segmentation threshold parameters to produce character images sufficient to identify the license plate.