Payment Card Character Recognition via Image Derivative Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional optical character recognition (OCR) techniques are not effective for recognizing embossed text on payment cards, especially under poor lighting conditions or with low contrast, and are hindered by decorated backgrounds and varying fonts, leading to low recognition accuracy.
Innovation Solution
A computer-implemented method involving image processing techniques, including determining a derivative of the payment card image, identifying Regions of Interest (ROIs) by horizontal sum of pixel values, extracting characters using peak values in a histogram, and recognizing characters using a trained Artificial Intelligence (AI) technique, such as a convolutional neural network-based deep learning algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional OCR techniques are used for character recognition, then the system is simple to implement, but the recognition accuracy is low for embossed text under poor lighting conditions
Solution Approach 1:
The patent applies preliminary image processing actions including determining derivatives of the image, computing horizontal sums of pixel values, and identifying Regions of Interest before character recognition. These preliminary steps enhance the quality of input data for the recognition algorithm, improving accuracy for embossed text under various lighting conditions while managing complexity through structured preprocessing.
Solution Approach 2:
The patent segments the payment card image into multiple Regions of Interest (ROIs) based on horizontal sum of pixel values and derivative analysis. Each ROI contains potential character regions that are processed separately, allowing the system to focus computational resources on relevant areas and improve recognition accuracy by treating different parts of the image independently.
2Reliability
If traditional OCR is used, then the processing time is short, but the recognition rate is very low for images with low contrast
Solution Approach 1:
The system performs preliminary actions by determining image derivatives and computing horizontal sums of pixel values across multiple rows before character recognition. This preprocessing enhances contrast and identifies character regions in advance, improving recognition reliability for low-contrast images while managing processing time through efficient computational steps.
Solution Approach 2:
The patent transforms the 2D image data into a 1D horizontal sum profile by aggregating pixel values across rows. This dimensional transformation highlights character regions that may have low contrast in the original image, enabling more reliable detection and recognition while reducing the complexity of direct 2D pattern matching.
3Measurement precision
If embossed text with gold or silver foil is used, then the card appears visually appealing, but the text becomes difficult to distinguish when foil is peeled off
Solution Approach 1:
The system performs preliminary derivative analysis and horizontal sum computation to identify character regions before recognition. This preliminary processing captures the structural information of embossed text that remains detectable even when foil is partially peeled, improving text distinguishability by focusing on the underlying embossed structure rather than relying solely on foil reflectivity.
4Measurement precision
If decorated images and patterns are added to the card background, then the card design is enhanced, but character identification becomes more difficult
Solution Approach 1:
The patent segments the image into multiple Regions of Interest by analyzing horizontal sums of pixel values and derivatives. This segmentation isolates character regions from the decorated background, allowing the recognition system to focus on character-containing areas while excluding background patterns, thereby improving character identification accuracy despite elaborate card designs.
Solution Approach 2:
The system applies different processing qualities to different parts of the image by identifying specific Regions of Interest. Character regions receive focused attention through derivative analysis and horizontal sum computation, while background areas are effectively excluded from detailed processing, enabling accurate character identification even with decorated backgrounds.
Data Source
AI summary
The present disclosure relates to a system and computer-implemented method for character recognition in a payment card. The method includes receiving an image of a payment card and one or more details associated with the payment card. Further, a derivative of the image is determined based on the one or more details and a horizontal sum of pixel values is determined for a plurality of rows in the image. Furthermore, one or more Regions of Interest (ROIs) are identified in the image by comparing the horizontal sum of pixel values with a predefined first threshold. Subsequently, one or more characters in the one or more ROIs are extracted using one or more peak values in a histogram of the one or more ROIs. Finally, each of the one or more characters extracted from the one or more ROIs is recognized using a trained Artificial Intelligence technique.