Data Extraction Method for Industrial Data Matrix Codes Based on Local Adjacent Modules Structure

Round 1

Reviewer 1 Report

• The research objectives and the methodology should be better explained and motivated. In my opinion, the background is not well organized. It would be better if the authors first explain the motivation for their study, then discuss the problem statement, and finally review the relevant studies.

“Our contributions include:(1) comparing different CNN models that help extract information from low-quality DM images; (2) unlike the traditional methods that predict the DM features based on pixels of each module, we use features of pairs of adjacent modules of a DM image to increase the accuracy of the data extraction; (3) edge images are constructed and corrected by the proposed correction method based on KM algorithm”.

• The authors must add updated articles 3 to 5 references from the journal “Applied Sciences".
• In the conclusions section, the authors should provide a general interpretation of the rustles, the unique contributions of the paper, limitations of the research, managerial implications, and the impact that the paper might have on future research and on policy decisions.

Author Response

Please see the attachment。

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors, according to today's message from the Assistant Editor, you have received exceptionally qualified reviews from other colleagues. Nevertheless, I consider it my duty to complete the previously begun consideration of the manuscript submitted by you in some particular moments.

The article proposed for publication is made on an exciting topic of the use of Artificial Neural Networks in the problem of Low-Quality Image Recognition to Extract Data taking into account the information of adjacent modules. One of the advantages of the presented main results of the original work is a detailed and substantive presentation of the implementation of the method. As you know, despite the widespread use of such technologies using DataMatrix, the predecessors of such tools using markers in the form of data matrices are punch cards. To date, there are a significant number of implemented data representation and processing technologies based on the use of two-dimensional labeling. Along with the tasks of marking and Processing Industrial Data, in the current challenging conditions of the pandemic, further digitalization, data-mediated remote everyday interaction using mobile photo and video cameras in the interests of faster and contactless communication of people also acquire particular importance.

Nevertheless, the predominance of the descriptive nature in this manuscript leaves not much room for discussion in the interests of expanding the problem being solved both in matters of adequate recognition of low-quality images through Artificial Neural Networks and many data representation issues encoding and extraction.

So, one of such possible discussion issues: it is proposed to clarify the possibility of a generalized representation of the described method of data extraction, taking into account the information of adjacent modules, when considering it in the general context of Extended Channel Interpretation (ECI).

[In particular, for the data interface in the "Extended Channel Mode," the interpretation of the processed data is determined by the currently enabled modes, which are activated and deactivated using "ECI indicators" included in the processed data. ECI indicators may be formed at various points in the transmitted message and may be either of "encodable" type or "non-encodable" or "signal" type].

To the extent possible, the inclusion by the authors of an additional subsection describing the conditions for reading the markings on the mentioned metal surfaces will allow for a more concrete and objective presentation of the factors preventing the improvement of the desired quality of Industrial Data Processing. You can include illustrations (photos) of markings with the results of, for example, exposure to aggressive environments, as well as characteristic examples of glare, reflective, uneven, curved, and other surfaces.

It seems appropriate to disclose the issue of the System Integration of the method proposed by the authors into the software tools they mention.

What are the features of the effective implementation of the proposed data extraction method that should be taken into account for this?

This will presumably make it possible to reasonably indicate the actual values of low contrast and those or other characteristics of the background surface that are significant for effective recognition or leveling.

It is proposed to cite additional literature sources with the most relevant and modern applications of the "Sliding Window Method" and its analogs and examples of modification and adaptation (specifically for DM) of methods for generating codes that detect and correct errors.

To what extent are the authors limited by the structure of the Data Matrix (DM)?

Can recommendations be formulated on making significant changes to the structure or maybe choosing a fundamentally different design (taking into account the extracted information of the neighboring modules)?

What aspects of the research hypothesis are original and formulated? Do the results obtain progress in modern knowledge (Recognition, Data Extraction, and other possible applied and technological aspects)?

In this regard, among other things, a broader readership will be exciting recommendations of the authors on the further development of methods and algorithms for Data Extraction to various kinds of application-related tasks with the needs of more rapid and effective processing of low-quality images.

It should be noted that the proposals are advisory, and I join the recommendations of other reviewers and their positive conclusions.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

1. In Section 4.5, it seems that unit of brightness are required for the experimental environment. A change in the recognition rate may be unavoidable depending on the brightness condition of the light.
2. Can you make a performance comparison with reference [21]? If it can be done, it seems that the objectivity of the paper can be further improved.
3. In Section 5, future research directions should be included.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Accept in present form