Study on detection of black neutral pen by hyperspectral imaging

The human eye is sensitive to light in the visible range and distinguishes materials based on color. However, humans are unable to distinguish between two identical colors. Black neutral ink stable, writing does not fade. Many important documents are written with a black neutral pen, such as contracts, receipts, certificates, checks and other documents, the numbers on these documents, time, text and so on. Easy to be added or tampered with, the identification of tampered handwriting and the reproduction of covered handwriting are important evidence in criminal proceedings, therefore, in most civil and criminal cases, a lot of document identification requires the identification of black neutral pen handwriting. There are two main methods for identifying handwriting: lossy detection and nondestructive detection. Hyperspectral imaging, an effective non-destructive tool, has been widely used in the identification of agricultural products in recent years. In this paper, 18 kinds of black neutral pens sold in the market are taken as the object to explore a more effective method of handwriting recognition, which provides a research basis for handwriting criminal investigation and identification.

In this paper, a 400-1000nm hyperspectral camera is used. FS13, a product of Hangzhou Color Spectrum Technology Co., LTD., can be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection).

1. Materials and equipment

Preparation of experimental materials and experimental samples

The experimental samples were 18 brands of black neutral pens that were popular in the market, and 18 brands of neutral pens tampered with and covered each other. After writing the number "1" with 18 brands of neutral pens, the number "40" was altered by other brands of neutral pens 24 hours later, and 306 samples of tampering experiments were made. (a) and (b) in Figure 1 are the pictures before and after pen 1 was tampered with by pen 2. As can be seen from Figure 1, after No. 1 pen is tampered with by No. 2 pen of the same color, the traces of tampering are completely invisible to the naked eye. 18 brands of neutral pens were used to write their respective numeric serial numbers, which were covered by other brands of neutral pens 24 hours later, and 306 masking experiment samples were made. (c) and (d) in Figure 1 show the pictures of No. 14 pen before and after being covered by No. 15 pen. As can be seen from Figure 1, the obscured writing is completely unrecognizable to the naked eye.

2. Results and discussion

Handwriting tampering and masking of reproduced identification results

Take pen No. 1 and pen No. 17 for example, as shown in Figure 2, (a) is a digital photo, (b) is the result of principal component analysis processing without background removal, (c) is the result of principal component analysis processing without background removal, and (d) is the result of false color synthesis processing. As can be seen from Figure 2, the processing results are clearer after the interference of background information is removed. A large amount of data analysis shows that false color synthesis has the best recognition effect on handwriting tampering. People who have not seen the original data can successfully identify the tampered handwriting, that is, the group of samples can be identified.

Taking No. 2 neutral pen as an example to cover the sample with No. 13 neutral pen, FIG. 3 (a) is the digital photo of the sample, (b) is the result of principal component analysis processing without background removal, (c) is the result of principal component analysis processing without background removal, and (d) is the result of false color synthesis processing. A large amount of data analysis shows that the principal component analysis processing with the removal of background has the best effect on the recognition of handwriting masking recurrence.

3. Conclusion

(1) In the range of 720-1000nm band, the spectral reflectance of different brands of neutral pens is very different, and it is the best band for recognizing handwriting.

(2) The recognition effect of domestic pens and Nissan pens can reach 100%, which provides a theoretical basis for the counterfeiting of goods.

(3) The research shows that after removing the background information, the identification effect is updated clearly after analysis and processing again.

(4) In this paper, the handwriting is recognized by noise reduction, IsoData, eye mask establishment, background removal and PCA analysis. After processing by different methods, different sample data will be recognized. Among 306 groups of tampered sample data of black neutral pens, 232 groups of data could be identified, with a recognition rate of 75.8%. Among 306 groups of black neutral pen masking samples, 175 groups of data could be reproduced, and the recognition rate reached 57.3%.

(5) The research results show that hyperspectral imaging technology can be used to identify the tampering and coverup between different brands of black neutral pens, which provides a research basis for the criminal investigation and identification of handwriting.