Detection of Impurity Rate of Machine-Picked Cotton Based on Improved Canny Operator

Round 1

Reviewer 1 Report

This work proposed a canny operator-based image processing algorithm to identify and separating the impurities in the machine-picked cotton and YOLO V5 is then used for impurity classification.  I am mainly concerned about the novelty of this work.  First of all, an extensive literature review of the Canny operator is missing in the introduction.  I did not see a significantly difference of the proposed Canny operator with the standard one.  Change the filters for noise removing should not considered as an improvement for the Canny operator itself for edge detecting.  Secondly, the multi-thread technology is vaguely described in the contents. What does that refer to? It is not clear how many cores and what CPU type are used for multi-thread technology.  Those concerns should be addressed before it can be considered for publication.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The intent of this paper is the real-time detection of impurity rate in machine picked cotton processing based on Canny operator. This paper is well organized and the contributions are good fit for the journal. The authors used different strategies to achieve the goal including edge detection, classification, etc. The results showing the betterment in the proposed work. However, there are few correction to be made before consider this paper for publications, which are listed as follows:

1. The equations used in subsection 2.2.2 are available in existing and published papers. Recommended to provide the citations.
2. In the conclusion, the authors claim that they used neural network for classification. But in the proposed work section, the authors detailed very less.
3. Is the computational time to perform this process is better than existing ones? It is recommended to estimate the time and space complexities for the proposed work
4. There are several clustering algorithms in the literature, for example, machine learning algorithms for wireless sensor networks a survey. Why the authors choose only k-means and SVM for the comparisons?
5. What the primary reasons you noticed to achieve the superior performance in the proposed work?

Author Response

Dear Reviewer：

Thank you for your comments concerning our manuscript entitled “Detection of impurity rate of machine-picked cotton based on improved Canny operator”. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Changes are marked with red font and yellow background. The main corrections in the paper and the responds to the your comments are as follow：

Responds to your comments：

1. Response to comment：The equations used in subsection 2.2.2 are available in existing and published papers. Recommended to provide the citations.

Response：Thank you for your meticulous work, this is our negligence, we have added citations on lines 172 and 179, and marked them with a yellow background.

2. Response to comment：In the conclusion, the authors claim that they used neural network for classification. But in the proposed work section, the authors detailed very less.

Response：Thank you very much for your question. In this article, we first segment the image, and then detect the target of the segmented image. In the part of target detection, we use YOLO v5 neural network framework. Because the YOLO v5 neural network framework can meet our requirements for cotton impurity detection (including real-time and accuracy), we have not changed the YOLO v5 neural network framework. We only use the YOLO v5 neural network framework to train our own data sets, and the resulting training model is used to detect cotton impurities. Therefore, we introduced our training and testing process in Section 2.2.5, and finally used 200 sets of images to test the accuracy of the model.

3. Response to comment：Is the computational time to perform this process is better than existing ones? It is recommended to estimate the time and space complexities for the proposed work.

Response：Thank you for your meticulous work. The algorithm in this paper is divided into two tasks, including image processing and data processing. If single thread is used, the operation time is relatively long, so we use multi-thread technology to divide the two tasks into two threads for parallel calculation, which greatly saves the operation time. I'm sorry that we didn't introduce these separately, so we introduced them at the beginning of section 2.2.6 and marked them with yellow background, lines 385-389.

4. Response to comment：There are several clustering algorithms in the literature, for example, machine learning algorithms for wireless sensor networks a survey. Why the authors choose only k-means and SVM for the comparisons?

Response：Thank you very much for your suggestions. In the review section, we introduced some relevant researches on machine learning, in which machine learning is mainly used for classification recognition or target detection, which belongs to the second part of our research. In the process of segmentation, we tried many traditional methods, among which K-means, SVM and Canny edge detection segmentation performed better, while the other methods were quite different from our requirements. Therefore, we chose the best among the best, and did not choose other methods for comparison. In the classification and recognition part, we use the current popular YOLO framework to replace the traditional machine learning method, which can achieve better results.

5. Response to comment：What the primary reasons you noticed to achieve the superior performance in the proposed work?

Response：Thank you for your question, we improve the Canny operator for the edge extraction of cotton impurities, so that it can better detect the edges of impurities and segment them from the image, which lays a good foundation for the follow-up work. We use the YOLO neural network with good accuracy and real-time performance, so that it can quickly classify and identify impurities, and calculate the impurity content through the V-W model, so that the value of impurity content is closer to the real value.

Special thanks to you for your good comments.

We tried our best to improve the manuscript and made some changes in it. We appreciate for your warm work earnestly, and hope that the correction will meet with approval.

Once again, thank you very much for your comments and suggestions.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The improved version has a better shape and it is suitable for publication at Electronics.

Author Response

Dear Reviewer：

Thank you for your comments concerning our manuscript entitled “Detection of impurity rate of machine-picked cotton based on improved Canny operator”. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our research. This revision mainly modified the grammatical errors in the article, hope to get your approval.

Line 9: Aiming at the real-time detection of impurity rate in machine-picked cotton processing, a detection method of impurity rate in machine-picked cotton was proposed based on an improved Canny operator.

Line 10: According to the characteristics of different saturation between cotton and impurity, impurities were separated by extracting the image S channel.

Lines 11-14: Because of the problems existing in the traditional Canny operator edge detection, the gaussian filter is replaced by employing mean filtering and non-local mean denoising, which can effectively remove the noise in the image.

Line 15: The V-W model was established to solve the impurity rate based on mass. Compared with a single thread,

Lines 19-20: This method solves the problems of low speed, poor real-time , and easy ease to be have interfered. Then guide the cotton production process.

Line 47: Ding Xiaokang [6] et al. used the Canny operator to detect colored foreign fibers.

Line 59: For example, Roberts operator, Sobel operator, Prewitt operator ,and LOG operator, etc.

Lines 60-62: Compared with these differential operators, the Canny operator based on an optimization algorithm is widely used because of its advantages of large SNR and high detection accuracy.

Lines 76-79: However, because the traditional Canny operator is sensitive to noise and requires an artificial setting of Gaussian filter parameters, it lacks the adaptability to different images, so these methods have achieved experimental results to a certain extent.

Lines 79-80: However, there are still limitations, such as the processing time is being too long to carry out real-time detection.

Lines 100-104: In addition, as a comparative experiment, impurities were manually separated and weighed by electronic balance to get the impurity inclusion rate of a mass method, which was compared with the impurity inclusion rate based on image information, so as to compare the merits and demerits of various algorithms.

Lines 110-113: The lower computer collects the image in real-time through the camera, encodes the image information, and transmits it to the server of the upper computer through the raspberry PI.

Special thanks to you for your good comments.

We tried our best to improve the manuscript and made some changes to it. We appreciate your warm work earnestly and hope that the correction will meet with approval.

Once again, thank you very much for your comments and suggestions.

Author Response File: Author Response.pdf