Circuit Board Text Recognition Using Gradient Structural Concavity Algorithm
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical character recognition (OCR) techniques struggle with recognizing damaged or unreadable characters on printed circuit boards due to dirt and other forms of damage, resulting in lower detection accuracy and increased production costs from human intervention.
Innovation Solution
The implementation of a Gradient, Structural, Concavity (GSC) algorithm combined with a k-Nearest Neighbor Classifier using Euclidian Distance, which extracts features from digital images of characters by analyzing stroke shapes, trajectories, and relationships, improving character recognition efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If existing optical character recognition (OCR) techniques are used to recognize characters on printed circuit boards, then the recognition process can be automated, but the detection accuracy decreases when characters are damaged or unreadable due to dirt
Solution Approach 1:
The patent transforms the character recognition approach by changing parameters from simple pixel-based OCR to a multi-feature parameter system including gradient features (first-order derivatives), structural features (second-order derivatives), and concavity features. This parameter transformation enables the system to recognize damaged characters by analyzing multiple characteristics simultaneously, resolving the contradiction between automation and accuracy.
Solution Approach 2:
The patent segments the character recognition process into distinct feature extraction stages: gradient feature extraction, structural feature extraction, and concavity feature extraction. Each segment focuses on specific aspects of character morphology, allowing the system to handle damaged characters by relying on intact feature segments even when parts of the character are obscured.
2Productivity
If existing OCR techniques are used, then production costs can be reduced through automation, but human intervention is still required for damaged characters increasing costs
Solution Approach 1:
The patent implements a self-service mechanism where the character recognition system automatically handles damaged characters through its multi-feature analysis capability. The system performs self-correction by utilizing gradient, structural, and concavity features to reconstruct character information even when parts are damaged, eliminating the need for human intervention and maintaining both high productivity and reliability.
Solution Approach 2:
The patent incorporates feedback mechanisms through its comprehensive feature analysis system. By evaluating multiple features (gradient, structural, concavity) and comparing them against reference data, the system provides feedback to verify recognition accuracy and automatically correct errors, ensuring reliable operation without human intervention.
3Device complexity
If simple OCR algorithms are used, then device complexity is reduced, but the ability to recognize damaged characters is insufficient
Solution Approach 1:
The patent segments the complex recognition task into three manageable feature extraction modules: gradient feature extraction (first-order derivatives), structural feature extraction (second-order derivatives), and concavity feature extraction. This segmentation makes the overall complex system implementable by breaking it into standardized, reusable components that can be processed systematically.
Solution Approach 2:
The patent creates a composite recognition approach by combining multiple feature types (gradient features, structural features, concavity features) into a unified recognition framework. Similar to composite materials, this combination leverages the strengths of each feature type to achieve superior recognition accuracy for damaged characters while maintaining a structured implementation approach.
Data Source
AI summary
Techniques and systems for recognizing characters on a circuit board are discussed herein. For example, a digital image of a character on a circuit board can be obtained. The digital image can be processed with a gradient, structural, and concavity algorithm. The processing can include identifying a gradient feature based on a stroke shape, identifying a structural feature based on a stroke trajectory, identifying a concavity feature based on a stroke relationship, and so on. The character can be classified based on the gradient feature, the structural feature, and/or the concavity feature. The classifying can be performed using a k-nearest neighbor classifier algorithm and/or a distance metric.