Palm Print Recognition System: Need, Uses, and Working

A palm print recognition system captures a person’s palm area and compares it with the stored information in the database to establish or authenticate their identities. It is also called a palmprint biometric system.

The most common palmprint recognition system is the Automated Palmprint Identification System (APIS) which is able to store, retrieve and match hand prints.

Need for Palmprint Recognition System

• Accuracy and Reliability: A biometric system based on the palm is a reliable piece of information.
• Security: palmprint biometrics is a secure and reliable authentication functionality compared to password and other token-based systems.
• Speedy Investigation: Investigators don’t have to go through all palm prints, they only have to feed them and a system that gives a final possible list of prints
• Identify a person: Helps to filter out the person whose prints are not matched with a set of template data.

Application of Palmprint Recognition System

The following are the common uses of palmprint biometric systems:

• Person Identification: background checks and verification.
• Criminal Investigation: help in solving cases by left of palm prints or missing reports
• National ID cards: Driver’s licenses, voter registration, border crossing, or passports use a palmprint recognition system.
• Commercially: ATMs, Access control, and logins.
• Computers: for accessing computers and administrative logins.
• Network login: Specifically to access a network or admin access.
• Time and Attendance: Some high-end companies use palmprint biometric systems for attendance.

How Palm Print Recognition System Work? (With Architecture)

There are six basic design specifications in the palm print recognition system including:

1. Image Acquisition using sensors
2. Preprocessing and enhancement
3. Feature Extraction
4. Filtering (Classification/Indexing)
5. Matcher 
6. Decision Module

Following is the image of the Palmprint recognition biometric system design and workflow.

Step 1: Image Acquisition (Recording and Collection)

Goal: To capture palm prints in a comparable resolution.

Before starting the enrollment process make sure your hands are clean and dirt-free. For recording palm prints there are four major ways. These are using:

• Regular inking and cylindrical roller: most common for taking prints from suspects.
• LiveScan fingerprint technology: most common for direct authentication.
• Developing Latent Palm Prints
• Adhesive Lifting Material: commonly used to lift palm prints from the crime scene.

For onsite palmprint recognition, Live Scan is used to capture the image by maintaining quality.

You can even read the dedicated post on How To Record And Collect Palm Prints? Best Strategies And Guide.

Step 2: Preprocessing

Goal: Enhancement of the quality of the image and extraction of Region of Interest (ROI).

The following are the activities used in the preprocessing phase:

• Compression: Lossless or lossy compression.
• Quality Assessment: an algorithm that calculates and assigns a quality score.
• Define ROI/Segmentation: Removes the unnecessary elements of the image by setting boundaries for the palm region (also called Region of Interest (ROI) and discarding the background.
• Increase Signal to Noise ratio: Enhances the image quality by removing the noise and increasing the signal-to-noise ratio.
• Normalization: an additional algorithm to improve the clarity or desired ridge details.

After the preprocessing phase, the enhancement image is handed over to the feature extraction module.

Step 3: Feature Extraction

Goal: Extraction of salient and discriminatory ridge details from palm prints.

Feature extraction module extracts lines, ridges, and wrinkles from the palmprint’s ROI based on the algorithm specifically designed for a classification system. 

The system sets the principal lines and major ridges or wrinkles and then sets a geometric measurement for all the features based on the algorithm.

With set parameters and algorithms, the system identifies sets of ridges and retains the most discriminatory information possible. The feature extraction algorithm greatly controls the performance of matching and accuracy of any palmprint recognition system.

The extracted details are then sent to the filtering module before sending to the matching phase.

Step 4: Filtering (Classification/Indexing) Classification Systems of Palm Prints

Goal: Selectively save data by classification or indexing based on a specific template.

Filtering can be achieved by two different approaches: (a) Classification algorithms and (b) Indexing algorithms 

A. Classification Algorithms

They are also called classifiers. They divide the extracted features into discrete sets of classes based on classification systems. Some common palmprint classification systems are:

• Western Australian Palmprint Classification System
• Liverpool Palmprint Classification System
• Automated Palmprint Classification System 

Each class set different sets of data or palm ridges make it narrowing down (during search) easy and more specific categories. Once the template in a database is classified, values are sent to a matching module.

B. Indexing Algorithms

It is also called a continuous ordering system. Unlike classification algorithms, where data is partitioned into classes, here in indexing algorithms, the data is stored in a continuous vector representation as a string code.

It also serves an extremely fast matching process by simply comparing the query string code with that one saved in the database.

Step 5: Matching

Goal: Compare features extracted from the input query with the stored database.

The matching module used an algorithm to compare the features extracted from the input. It then compares with the stored templates in the database to produce the final result as a score of similarity and dissimilarity.

There are two types of palm print matching techniques: (a) geometry based and (b) feature based.

A. Geometry-Based Matching

Further sub-classified as point and line based method

A1. Point-based Matching: an average filter is used for the matching of feature points with corresponding orientation with the palm line. The two sets of feature points are geometrically aligned and interesting point detectors.

A2. Line-based Matching: It is more informative than point based because of the rich line features in palm prints. 

In this, the extracted line segments are determined by a series of endpoints and then compared with the database based on respective curves, lines, and lengths.

B. Feature-Based Matching

In feature-based matching of palmprints, two sets of values are used– magnitude and orientation data of palm ridges. This includes principal lines, creases, wrinkles, or other textures for general matching purposes.

The values are extracted by using statistical and algebraic techniques to represent a set of features. To compute the magnitudes common algorithms are:

• Principle Component Analysis (PCA)
• Independent Component Analysis (ICA)

For the computation of orientation information of palm lines, Gabor filters are used.

Step 6: Decision Module

Goal: Output the decision as validating the claimed identity or information of enrolled templates for further one-to-one comparisons.

The palmprint biometric system verifies a person’s identity by:

• If Identity is claimed: One-to-one matching of claimed identity with the stored identity in the database. It is called palmprint verification.
• If no Identity is claimed: the system compares palm prints with all templates stored in the database to establish an identity which is called palmprint identification.

Challenges In the Palm Print Recognition system

1. Usability is usually compensated over other recognition systems such as fingerprint systems. 
2. Effectiveness depends on the algorithm and classification being used.
3. The speed of matching and authentication may vary from system to system.
4. Comfortability is usually observed over fingerprints.
5. All the components required to make palm print recognition systems works usually cost much higher than fingerprints.
6. Regular maintenance and upgradation are required.
7. The number of authentication queries in a time period depends on the computational power and server performance.

Read More: Henry Fingerprint Classification System: Key, Major, Primary, Secondary, Subsecondary and Final

References:

• A Comparative Study on Palmprint Recognition [Research Paper]
• Advanced Biometric Technologies by Girija Chetty [Book]
• Palmprint Authentication by Xingyi et.al. [Book]
• Biometric Systems: Technology, Design and Performance Evaluation [Book]