Inkjet printer prediction under complicated printing conditions based on microscopic image features.

Area of Science:

• Forensic Science
• Computer Science
• Image Analysis

Background:

• Accurate printer model identification is crucial for forensic document examination.
• Varying printing conditions can challenge the reliability of forensic analysis.
• Existing methods may lack robustness against diverse printing parameters.

Purpose of the Study:

• To develop and validate a novel technique for predicting the source inkjet printer model of a document.
• To assess the effectiveness of the technique under various printing conditions.
• To enhance the capabilities of forensic document examination.

Main Methods:

• Collected samples from 41 inkjet printer models under diverse printing conditions.
• Analyzed morphological features using image analysis algorithms and microscopic imaging.
• Applied statistical tests (Wilcoxon) to assess feature significance.
• Employed discriminant analysis and K-Nearest Neighbor (KNN) algorithm for printer prediction.
• Utilized a training dataset with 70% of images and tested on the remaining 30%.

Main Results:

• Significant variations in printed samples were identified across different printing conditions.
• Quadratic Discriminant Analysis (QDA) achieved 96.3% accuracy in source printer prediction.
• A combined approach using KNN and QDA reached a high accuracy of 98.6%.

Conclusions:

• The developed technique demonstrates high accuracy in identifying inkjet printer models.
• The combined KNN and QDA method offers a robust solution for forensic document examination.
• This approach can significantly aid in the forensic analysis of printed documents.