Progress of Machine Vision Technologies in Intelligent Dairy Farming

Round 1

Reviewer 1 Report

The article is of a review nature. It concerns the use of computer vision methods in the automation of dairy cow breeding.
The number of dairy cows around the world is increasing. To reduce the labor intensity of breeding, smart systems are being developed to support this activity. Touch sensors were used to monitor the behavior of the cows. However, they can trigger stress responses and have limitations: high cost, vulnerability, reliance on human resources for maintenance, and limited functionality. The authors describe how advances in artificial intelligence and image processing have opened up new possibilities for recognizing cow behavior using machine vision technology. The use of machine vision allows for non-contact, stress-free and economical monitoring of many types of behavior.
Traditional cow identification methods include manual tagging with ear tags or collars, and electronic RFID tags. Machine vision and deep learning identify cows based on facial features, muzzle prints, body contours and blotch patterns. The authors have compiled the results of selected studies on the identification of cows using such methods.
The best of these methods achieved an average recognition accuracy of 99.64%, but the others had much worse results (even below 84%), which probably prevents their use in practice. It is also important to identify the behavior of moving cows. It is important, among others, in disease prevention. Here it is necessary to analyze video sequences.
In conclusion, the authors predict that the future of cow farming will be largely based on digital technologies. They list the advantages of machine vision in precision cow farming. They also mention that there are still challenges in this area.

The article may be used by the authors of future works on the application of computer vision in automatic cow breeding as an introduction to the review of the field.

 

Additional comments:

1. What is the main question addressed by the research?

  The main question addressed by this research is - how can machine vision technology be applied in so-called precision breeding of dairy cows to: improve farming efficiency, reduce costs, enhance animal welfare, and provide data support for intelligent farming.    

2. Do you consider the topic original or relevant in the field? Does it address a specific gap in the field?  

The paper is of a review nature and does not describe new, original research. It contains information about results obtained by other researchers.    

3. What does it add to the subject area compared with other published material?  

Collecting results obtained by different authors in one place (results including e.g. accuracy of different cow identification methods) allows future researchers to choose a particular method as a starting point for further refinement. However, there are no new results here.    

4. What specific improvements should the authors consider regarding the methodology? What further controls should be considered?  

The authors reviewed the literature, so it is difficult to talk about the research methodology. The summary of results achieved by different authors can be possibly improved so as to group the results obtained using similar methods (now Table 2 presents various methods without grouping in terms of mode of operation).    

5. Are the conclusions consistent with the evidence and arguments presented and do they address the main question posed?  

The fifth chapter titled "Conclusions" contains considerations on the importance of information technology in cow breeding. It emphasizes the need to develop these methods for better breeding results. This is actually what the authors of the cited works claim.    

6. Are the references appropriate?  

Yes, bibliographic items are properly selected.    

7. Please include any additional comments on the tables and figures.  

The article contains three tables, all of which contain the results obtained by other authors. Two drawings are illustrative - they show the number of cows bred in the world and the methods of attaching sensors monitoring the behavior of cows.

The text is written in understandable English, although it requires minor corrections. Here are some suggested corrections (but it is necessary to read the whole text carefully):

Line 42 - redundant dot after "on"

Line 45 - end of line before the end of the paragraph

In order to obtain about every dairy cow information -> In order to obtain information about each dairy cow

it is difficulty to recognize -> it is difficult to recognize

confidence scores of SSD algorithm exceeds 90% -> confidence scores of SSD algorithm exceed 90%

These factors can lead to breeding personnel cannot timely obtain the status of dairy cows... ->
These factors can prevent breeding personnel from timely obtaining the status of dairy cows...

abnormal status of dairy cow -> abnormal status of dairy cows

measures can be taken in a timely to avoid further expansion of losses ->
measures can be taken in a timely manner to avoid further losses

machine vision monitoring of daily behavior of dairy cows can also provide reference data for the welfare breeding of dairy cows ->
machine vision monitoring of the daily behavior of dairy cows can also provide reference data for the welfare-oriented breeding of dairy cows

Author Response

Please see the attachment.

Author Response File: Author Response.doc

Reviewer 2 Report

Paper gives a decent review of literature surrounding cattle recognition. Some critical appraisal of each method is given, with a summary of results.  I would have liked to see more figures to explain/demonstrate some of the concepts (e.g., cow muzzle detection, Holstein cow patterns) but not 100% necessary.

Comments for author File: Comments.pdf

Minor typo: line 114: "title" should be "cattle".

Author Response

Please see the attachment.

Author Response File: Author Response.docx