An Improved Method to Obtain Fish Weight Using Machine Learning and NIR Camera with Haar Cascade Classifier

Round 1

Reviewer 1 Report

The manuscript is written with clear understanding of the project addressed. However, there are major concerns that need to be addressed to enhance the quality of the manuscript. My specific comments are as follows:

Abstract:

Add methods of study

Add main finding of the study

Introduction:

L41: Explain briefly how agriculture 4.0 contribute to the development in aquaculture

Add on literatures on applications of machine vision system in aquaculture

Based on your objectives, please compare how your study is different from those that have already been published

Methods:

Explain how does the system works.

Section 2.3 should comes after 2.1

How about split ratio of training/testing/validation?

Results:

Relate how the algorithms contribute to the identification of fish; add justification

Section 3.2: The authors only mention about training model. How about testing/validation dataset

Figure 15: specify what are the two models in the figures. Add legend

Instead of mentioning the results, the authors should justify/explain the findings

 Conclusion:

Add on main finding/results of the study. What are the main outcome based on the results. The authors should highlighted this matter

General comments:

Please check the reference styles and grammar of the manuscript.

Author Response

Point 1: Abstract:

Add methods of study

Add main finding of the study

Artificial intelligence technology is found in many areas like medicine, engineering, biology, agriculture, and economy, to mention a few; specifically, machine learning is being used to solve problems or help in some tasks that the existing technology could no to do. Moreover, the technology has generated tools like Near Infrared Spectroscopy cameras to detect objects much faster and avoid light brightens and shadows that using a conventional camera was impossible to avoid. Lastly, with the help of Haar Classifier, detecting some particular objects is straightforward; this work used the technologies mentioned before to detect fish in normal conditions and estimates the length and weight, avoiding using hands. This method successfully detected the fish in the aquatic environment and using mathematical models the length and weight of the fish was obtained, the above was enhanced using near-infrared wavelength to avoid the noise caused by light. 

Response 1:  L16 – L26 The calculation of weight and mass in aquaculture systems is of great importance since with this task it is decided when to harvest; generally, the above is manipulating the body manually, which causes stress in the fish body. Said stress can be maintained in the fish body for several hours. To solve this problem an improved method was implemented using artificial intelligence, Near Infrared Spectroscopy camera, Haar Classifiers, and mathematical model. Hardware and software were designed to get a photograph of the fish in its environment in real conditions. This work aimed to obtain fish weight and fish length in real conditions to avoid the manipulation of fish with hands for the process mentioned, avoiding fish stress and reducing the time for these tasks. With the implemented hardware and software adding an infrared light and pass band filter for the camera successfully the fish was detected automatically and the fish weight and length were calculated moreover the future weight was estimated.    

Point 2: Moreover, agriculture 4.0 can help aquaculture with digital technologies like cyber-physical (CPS), artificial intelligence (AI), wireless sensor networks (WSN), big data analytics (BDA), autonomous robots systems (ARS), and ubiquitous cloud computing (UCC) [7].

L41: Explain briefly how agriculture 4.0 contribute to the development in aquaculture

Response 2: L42 - L47 Moreover, agriculture 4.0 contribute to the development of agriculture because crops can be cultivated much easier with lees effort in less time and also every time the growing variables can be monitored by sensor and displayed on a PC or cell phone screen; technologies like digital technologies like cyber-physical (CPS), artificial intelligence (AI), wireless sensor networks (WSN), big data analytics (BDA), autonomous robots systems (ARS), and ubiquitous cloud computing (UCC) [7].

Point 3: Add on literatures on applications of machine vision system in aquaculture.

Response 3:  L78 – L83 In the case of machine vision system [1] used convolutional neural network and machine vision to provide an automatic method for grading fish feeding intensity;  also [2] use a machine vision system including a Multi-column Convolution Neural Network (MCNN) and deeper dilated convolution neural network (DCNN) to counting fishes; finally [3] use machine vision, acoustics and sensors to analyze fish behavior in pro of production and management decisions.    

Point 4: Based on your objectives, please compare how your study is different from those that have already been published.

Response 4: L89 – L 95 Finally, this work is different from others in the first instance it used an infrared camera adding a pass-band filter lens to reduce the wavelength and get more focus in capturing the fish and avoid the noise that can interfere; also used Haar Classifier to identify the fish in the culture system, compared with other works they used more complex analysis like convolution neural network, multi-column convolution neural network, artificial neural network or wavelet; lastly, de system uses a mathematical model to estimate fish future weight and length.          

Point 5: Explain how does the system works.

Response 5:  L118 – L122 The operation of the system is as follows, first, the image is captured by the NIR camera this camera is connected to the PC, the information captured by the camera is used by an algorithm done with Haar Cascade Classifier and mathematical models, lastly, the information related to fish actual length, future length, the actual weight, and future weight is shown in the user interface.  

Point 6:  Section 2.3 should comes after 2.1

Response 6: L100 – L122 2.1 System Design, L123 – L145 2.2 Image acquisition system 

Point 7: How about split ratio of training/testing/validation?

Response 7: L207 – L208 Finally, in this training was considered training-testing-validation ratio 60:20:20.

Point 8: Relate how the algorithms contribute to the identification of fish; add justification.

Response 8: L298 – L302 The algorithm contributes to fish detection because; the fish recognition is automatic due to the camera o video [5], the algorithm can analyze a large amount of data in a short time, these data previously was saved in the algorithm [6], also the algorithm allows segmenting the object in more difficult environments [7], moreover, the algorithm can be combined other method to improve it [8].

Point 9: Section 3.2: The authors only mention about training model. How about testing/validation dataset.

Response 9:  L308 – L313 The testing dataset was done after the training, this was done using the first system configuration shown in Figure 2; several repetitions were made which consisted of detecting the fish in the small fishbowl. The validation stage was done using a 500 L fish tank, in these tanks a certain number of fish were placed, and the validation was done on different days by taking photographs from different tanks. Also was done weight and length measurements to corroborate the validation of the algorithm.     

Point 10: Figure 15: specify what are the two models in the figures. Add legend. 

Response 10:

Figure 15. Regression analysis between two models (Logistic and Michaelis-Menten)

Point 11: Instead of mentioning the results, the authors should justify/explain the findings.

Response 11: L265 - L360

Point 12: Add on main finding/results of the study. What are the main outcome based on the results. The authors should highlighted this matter

Response 12: L399 – L408 The method first allowed the recognition of the fish and then the processing to obtain fish length and fish weight to avoid the fish stress because of the handling. The measurement task by hand takes consideration time, with this method is consumed less time for this task. If the environment of the fish changes or the species is different, new training in the software will be required.

 As time goes applications related to artificial intelligence will be more common to stay in many applications than before were not taken into consideration by this kind of technology, this technology allows great savings in time and money; it is necessary to mention that is preferable that the application be portable and avoid wires and have certain sturdiness.

Reviewer 2 Report

1. Please explain the 1200 images taken by NIR cameras, such as image resolution, image format, etc.

2. This paper uses the HAAR -class linked classifier, but the HAAR class classifier is prone to the results of fake positive testing. Please explain in detail what the false positive test results are. What is the impact of the test results?

3. Please express the strong classifier used by the HAAR -class-linked classifier in this article.

4. Please explain the weight and length error analysis of the fish in the experimental results in detail.

5. There is no quantitative analysis of the image recognition results of the fish, please show the accuracy, recall rate, etc. 

6. Please demonstrate why the HAAR class classifier uses fewer resources.

Author Response

Point 1: Explain the 1200 images taken by NIR cameras, such asimage resolution, image format, etc.

Response 1: L125 – L128 To obtain the collection fish data, the fish species used was tilapia Oreochromis niloticus; 1,200 pictures were obtained with the NIR camera; resolution of 1.3 MP, 1280 x 1024, the image format used was PNG, later the software was trained whit these pictures.

Point 2: This paper uses the HAAR -class linked classifier, but the HAAR class classifier is prone to the results of fake positive testing. Please explain in detail what the false positive test results are. What is the impact of the test results?

Response 2: L203 – L208 An aspect to take into consideration is that Haar Classifier can throw false positives, in this training the false positive test was done with the camera taking objects that can be in the fish environment but are not fish, these were classified as negative images. In the first instances could be many false positives but as the training continues this false positive rate starts to reduce. The impact on the test result is that in the beginning, the software could throw false positives but with the use of the software, this will decrease.

Point 3: Please express the strong classifier used by the HAAR -class-linked classifier in this article.

Response 3: L214 - L125The strong classifier used in this work was the Adaboost for searching a small number of features related to fish and not have a significant variation.

Point 4: Please explain the weight and length error analysis of the fish inthe experimental results in detail.

Response 4: L369 – L370 In this part, the weight and length error analysis was done with statistical software, that estimates the r-square.  

Point 5: There is no quantitative analysis of the image recognition resultsof the fish, please show the accuracy, recall rate, etc.

Response 5: L319 – L321 The recognition performance obtained from the validation phase resulted in an average accuracy of 92% also a true positive rate equal to 95% and finally false positive rate equal to 12%. 

Point 6:  Please demonstrate why the HAAR class classifier uses fewerresources

Response 6: There is a misundertandig whas was meant that only one pc is needed to program or use the HAAR class classifier.

Round 2

Reviewer 1 Report

The authors have addressed all the comments. Hence, the paper can be accepted as it is. 

Reviewer 2 Report

I accept the paper for publication in the current version.