References

References for basic understandings

Note: These can help you to understand the basic information of currency detection.

1. Features of Korean Won

Intaglio latent Images

• Description: When viewed at an angle (askance) from the position of the eyes, the word 'WON' appears.
• How it works: This effect is achieved using a special intaglio printing method, which creates fine grooves and patterns on the note.
• Purpose: It helps verify the authenticity of the currency because the latent image is difficult to reproduce in counterfeit banknotes.

Windowed Security Thread

• Description: The security thread is a plastic film that includes hologram letters.
• How it works: It is partially exposed at regular intervals on the left side of the portrait (obverse side of the note).
• Purpose: Prevents counterfeiting as hologram technology is challenging to replicate. Acts as a verification feature for checking the authenticity of banknotes.

Color-Shifted Ink

• Description: Special ink is mixed with materials that have different light reflection properties.
• How it works: When the note is tilted at various angles, the face value number changes color from green to blue.
• Purpose: This feature makes counterfeiting more difficult because reproducing color-shifting ink requires advanced technology.

See Through Register

• Description: The Taeguek symbol (a traditional Korean symbol) is divided and printed on both sides of a banknote.
• How it works: When the note is held up to light, the two divided shapes align perfectly to form a complete Taeguek.
• Description: This feature is used to verify the authenticity of the banknote since precise alignment is difficult to replicate in counterfeit notes.

Micro Lettering

• Description: Tiny letters or patterns are printed on the banknote that are difficult to see with the naked eye but can be observed using a magnifier.
• How it works: When counterfeit banknotes are created using color printers or copiers, the micro lettering often appears as a solid or dotted line instead of clear letters.
• Purpose: Micro lettering prevents counterfeiting since it requires advanced printing technology to replicate with precision.

Watermark

• Description: When the banknote is held up to light, a hidden reverse image portrait becomes visible in the blank space on the left side of the note.
• How it works: This effect is created by variations in the thickness of the paper during the manufacturing process.
• Purpose: Watermarks are difficult to counterfeit as they require specialized printing and paper manufacturing processes.

SPAS (Special Press and Soldering)

• Description: Without holding the note up to the light, the hidden face value number can be observed directly with the naked eye due to enlarged thickness differences in certain areas of the note.
• Purpose: It enhances security and allows for quick verification of the note's authenticity.

link: https://www.komsco.com/guide/sub4.html

2. Currency detection for visually impaired

Purpose

Currency recognition systems aim to help visually impaired individuals identify banknotes easily and accurately. These systems typically provide audio or tactile feedback to communicate the banknote's denomination.

Key Techniques

1. Computer Vision
   • Uses image processing and deep learning models to analyze captured images of banknotes.
   • Features like color, patterns, text, holograms, and unique visual markers are extracted.
2. Machine Learning Models
   • Convolutional Neural Networks (CNNs): Classify and recognize denominations based on banknote images.
   • YOLO (You Only Look Once): Detect and recognize banknotes in real-time video streams.
3. Optical Character Recognition (OCR)
   • Extracts and processes text (e.g., serial numbers, denominations) for identification.
4. Edge and Feature Detection
   • Algorithms like SIFT (Scale-Invariant Feature Transform) detect patterns, edges, and unique markers to differentiate banknotes.

Workflow

1. Image Capture:
   • A smartphone or camera captures the image of the banknote.
2. Image Preprocessing:
   • Noise reduction, grayscale conversion, and alignment ensure clarity.
3. Feature Extraction:
   • Color shifts, security features, and text are analyzed for accurate classification.
4. Recognition and Feedback:
   • The denomination is identified and communicated to the user via audio output (e.g., "10,000 won").

link: https://www.ijres.org/papers/Volume-11/Issue-4/1104732738.pdf

3. Currency Recognition Using Image Processing (Detecting Counterfeit)

Purpose

The article highlights various applications and goals of currency recognition systems:

• Automation: Streamline cash handling processes in banking, retail, and vending machines.
• Accessibility: Assist visually impaired individuals in identifying denominations.
• Counterfeit Detection: Identify fraudulent banknotes by examining security features.
• Efficiency: Reduce human error and enhance the speed of transactions in financial sectors.
• Global Applicability: Facilitate currency exchange and recognition across multiple currencies.

Key Techniques

The system relies on advanced image processing and machine learning methods: a. Preprocessing

• Noise Reduction: Filters to remove irrelevant data (Gaussian smoothing, median filtering).
• Image Enhancement: Improves clarity through normalization and histogram equalization.
• Edge Detection: Identifies contours and boundaries in images to extract relevant features.
• Sharpening: Highlights details for better feature extraction. b. Feature Extraction
• Security Features: Analyzes visible elements like watermarks, microprinting, and holograms.
• UV/IR Imaging: Uses ultraviolet and infrared sensors for hidden features detection.
• Texture and Color Analysis: Recognizes patterns, shades, and unique designs on banknotes. c. Classification Algorithms
• Machine Learning:
  • Support Vector Machines (SVMs).
  • Decision Trees.
• Deep Learning:
  • Convolutional Neural Networks (CNNs): For pattern recognition and counterfeit detection.
  • YOLO (You Only Look Once): Real-time object detection for multiple denominations.
• Anomaly Detection: Detects counterfeit bills by identifying discrepancies in texture, design, or security features.

3. Workflow

The currency recognition process is divided into systematic steps:

1. Image Acquisition

• Capture Method:
  • High-resolution cameras or scanners collect images of currency.
  • Controlled lighting conditions ensure clarity and consistency.
• Input Sources:
  • Currency notes are fed into systems via cameras (e.g., ATMs, vending machines, or mobile apps).

2. Preprocessing

• Images are:
  • Resized to uniform dimensions for standardized processing.
  • Normalized to enhance brightness and contrast.
  • Filtered to reduce noise and highlight key patterns.

3. Feature Extraction

• Algorithms detect:
  • Contours, textures, and edges.
  • Security elements (e.g., holograms, watermarks, UV-sensitive markings).

4. Classification

• ML/DL Algorithms:
  • Classify denominations based on extracted features.
  • Detect counterfeits by comparing extracted features to genuine banknote datasets.

5. Output

• Recognition Results:
  • Denomination and authenticity are displayed or announced.

link: https://saiwa.ai/blog/currency-recognition-using-image-processing