Two-Dimensional Flow Field Measurement Method for Sediment-Laden Flow Based on Optical Flow Algorithm
Round 1
Reviewer 1 Report
I’m not sure how the velocity in m/s was calculated from the optical flow filed. How the system was calibrated to move form pixels to physical units [m]?
The performance of the CC-algorithm is typically improved with use of image pre-processing. Have you done some background substation etc.? Please state it in the paper? If not my suggestion is to try to compare the Optical Flow with CC-PIV based on pre-processed images.
The references are quite old. In the introduction please elaborate more on state-of-art developments in Echo-PIV and optical flow in PIV. Such as:
Gurung, A., Poelma, C. Measurement of turbulence statistics in single-phase and two-phase flows using ultrasound imaging velocimetry. Exp Fluids 57, 171 (2016). https://doi.org/10.1007/s00348-016-2266-x
Author Response
Point 1: I’m not sure how the velocity in m/s was calculated from the optical flow filed. How the system was calibrated to move form pixels to physical units [m]?
Response 1: In this paper, we conduct calibration experiments to determine the relationship between the pixel flow rate and the physical flow rate. We clarify our approach by adding a description of the calibration experiment in lines 277~285.
Point 2: The performance of the CC-algorithm is typically improved with use of image pre-processing. Have you done some background substation etc.? Please state it in the paper? If not my suggestion is to try to compare the Optical Flow with CC-PIV based on pre-processed images.
Response 2: The manuscript has included a description of the recent progress of the CC algorithm in the introduction section, and pointed out that its shortcomings mainly lie in the correlation of the ultrasonic image sequences affected by the sand concentrations in the flow, and the interference of the multi-peak phenomenon in the intercorrelation plane. In the experimental section, the pre-processed CCPIV calculation is performed and compared with the algorithm in this paper, based on the literature [4].
Point 3: The references are quite old. In the introduction please elaborate more on state-of-art developments in Echo-PIV and optical flow in PIV. Such as: Gurung, A., Poelma, C. Measurement of turbulence statistics in single-phase and two-phase flows using ultrasound imaging velocimetry. Exp Fluids 57, 171 (2016). https://doi.org/10.1007/s00348-016-2266-x
Response 3: The recommended papers were carefully read and cited, and the analysis of the current state of research was reorganized to strengthen the description of the current state of research related to PIV and UIV.
Many thanks to the reviewers for their careful reading and valuable comments. We have gained a lot from the revision process based on the suggestions.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The authors provide a paper on flow measurements in sediment-laden flows processed by optical flow techniques. The topic is of high interest in multiphase fluid mechanics. The overall quality of the paper is ok, but some aspects need to be improved. Besides the comments and annotations I made in the attached pdf file I recommend to clarify the following points:
1. The text starting at line 253 explains the hydraulic setup and refers to Fig. 4, but Fig 4 only shows the general concept but not the experimental setup. Add a figure that actually shows what you explain in the text.
2. Are the vector fields shown in Fig 9 and 11 time averaged results over a ceratin number of frames or instantaneous vector fields? There is no explanation on that in the text.
3. Explain the reference measurements to obatin the ground truth. There is only one sentence on line 306 saying that an ADV was used. However, ADV is a pointwise measurement technique. So how are using this point information to calulate the angle error for a two dimensional area. Do you scan the same area that is seen in the ultrasound images? This question is directly realted to the previous one. In case you use instantaneous vector fields from the ultrasound measurements, you would then compare this instantaneous data with time averaged point information of the ADV. And this point data is uncorrelated in time. So provide a more detailed explanation of your reference measurements.
4. You compare the optical flow results with cross correlation results. I assume you used a PIV software for that. In the text you only mention "templates" for certain sizes of interrogation windows. Since every state of the art PIV software provides iterative cross-correlation with grid refining and filtering etc., you need to explain the software used and calculation procedures that have been applied. Otherwise the comparison of processing results is not on a fair level.
5. In Fig 8 and 9 you used a velocity of 0.35m/s for the comparison with other methods in Fig 10 and 11 you use 0.25m/s. Why do you change the velocity?
Comments for author File: 
Comments.pdf
Author Response
Point 1: The authors provide a paper on flow measurements in sediment-laden flows processed by optical flow techniques. The topic is of high interest in multiphase fluid mechanics. The overall quality of the paper is ok, but some aspects need to be improved. Besides the comments and annotations I made in the attached pdf file I recommend to clarify the following points:
Response 1: All the comments and annotations in the attached pdf have been revised and polished.
Point 2: The text starting at line 253 explains the hydraulic setup and refers to Fig. 4, but Fig 4 only shows the general concept but not the experimental setup. Add a figure that actually shows what you explain in the text.
Response 2: We have added a physical picture of the experimental setup and a brief description in lines 250~276.
Point 3: Are the vector fields shown in Fig 9 and 11 time averaged results over a ceratin number of frames or instantaneous vector fields? There is no explanation on that in the text.
Response 3: The experiments choose a high concentration ultrasonic particle image sequence of 50 consecutive frames of ultrasonic particle images to acquire a time-averaged flow field. We have reinforced the description about it in lines 334~345, 386~392.
Point 4: Explain the reference measurements to obatin the ground truth. There is only one sentence on line 306 saying that an ADV was used. However, ADV is a pointwise measurement technique. So how are using this point information to calulate the angle error for a two dimensional area. Do you scan the same area that is seen in the ultrasound images? This question is directly realted to the previous one. In case you use instantaneous vector fields from the ultrasound measurements, you would then compare this instantaneous data with time averaged point information of the ADV. And this point data is uncorrelated in time. So provide a more detailed explanation of your reference measurements.
Response 4: We use adv measured flow velocities to compare and calculate the error with the depth-averaged flow velocity of the flow field measured by the method in this paper. Since the real underwater flow field is difficult to acquire, this verification is only used as a consistency verification.
Point 5: You compare the optical flow results with cross correlation results. I assume you used a PIV software for that. In the text you only mention "templates" for certain sizes of interrogation windows. Since every state of the art PIV software provides iterative cross-correlation with grid refining and filtering etc., you need to explain the software used and calculation procedures that have been applied. Otherwise the comparison of processing results is not on a fair level.
Response 5: Since the experiments in this paper use high concentration ultrasonic particle images, the direct use of PIV software is prone to matching failure, so we adopt the approach in the literature[4] and use some preprocessing such as window distortion correction to the correlation matching of different templates. We have indicated it in lines 336~342.
Point 6: In Fig 8 and 9 you used a velocity of 0.35m/s for the comparison with other methods in Fig 10 and 11 you use 0.25m/s. Why do you change the velocity?
Response 6: The comparative experiments were conducted at a flow rate of 25 cm/s. In order to observe the effect of increasing the number of layers on the adaptation to larger flow velocities, additional experiments were performed at a flow velocity of 35 cm/s. The experiments are described in the corresponding section in lines 336~339, 386~387.
Many thanks to the reviewers for their careful reading and valuable comments. We have gained a lot from the revision process based on the suggestions.
Author Response File: Author Response.docx
Reviewer 3 Report
The Authors present a methodology based on ultrasounds and L-K optical flow to extract velocity fields from high-density water-sand mixture. They compare the velocity results to another optical-flow based algorithm and to a cross-correlation based algorithm.
In my opinion there are several issues that should be addressed before this manuscript can be considered for publication.
-Organization of the manuscipt is not clear: the introduction section contains information (lines 54 to 67) that should be moved in the next sections. The sections containing well-known information (optical flow theory for example) should be shortened or moved to an appendix. Paragraph 4.1 is placed before the apparatus description, it should be placed after it.
-Introduction should be extended so as to highlight the state of the art in optical flow experiments so that the novelty presented in this work are clearly stated.
-The experiment should be explained better; the facility is depicted in a schematic way, please provide a more detailled scheme or pictures. Also experimental parameters are described in different sections of the manuscript (the flow speed is in the Results section).
-The blocks in figure 3 are not all explained.
-The manuscript is focused on the comparison of a particular optical flow methodology (L-K) to cross-correlation and optical-flow H-S algorithm applied to ultrasound images.
In my opinion the title should be changed since the actual measurements are limited and the main subject of the paper is methodology. This should be also clearly stated in the abstract.
Author Response
Point 1: Organization of the manuscipt is not clear: the introduction section contains information (lines 54 to 67) that should be moved in the next sections. The sections containing well-known information (optical flow theory for example) should be shortened or moved to an appendix. Paragraph 4.1 is placed before the apparatus description, it should be placed after it.
Response 1: The paper has been reorganized by moving the description of the ultrasonic image characteristics of the turbid water flow to be measured to lines 105~115, simplifying the theoretical description, and changing the order of the description of the experimental conditions to: 4.1. the experimental environment and facilities of the water tank; 4.2. the ultrasonic water flow imaging system; and 4.3. the water flow conditions in lines 250~299.
Point 2: Introduction should be extended so as to highlight the state of the art in optical flow experiments so that the novelty presented in this work are clearly stated.
Response 2: Introduction has been extended and reorganized to strengthen the description of the current state of research related to optical PIV, cross-correlation algorithm, and UIV.
Point 3: The experiment should be explained better; the facility is depicted in a schematic way, please provide a more detailled scheme or pictures. Also experimental parameters are described in different sections of the manuscript (the flow speed is in the Results section).
Response 3: The physical picture of the experiment is provided, and the deployment of the physical picture is briefly introduced in lines 250~276. The experimental parameters are uniformly presented in the experiment section. The comparison experiments were conducted at a flow rate of 25 cm/s. In order to observe the effect of increasing the number of stratifications on the adaptation to larger flow velocities, additional experiments were performed at 35 cm/s flow velocity. Therefore, the corresponding parameters are described in the corresponding section, lines 336~339, 386~387.
Point 4: The blocks in figure 3 are not all explained.
Response 4: We have strengthened the interpretation of the blocks in lines 88~95.
Point 5: The manuscript is focused on the comparison of a particular optical flow methodology (L-K) to cross-correlation and optical-flow H-S algorithm applied to ultrasound images. In my opinion the title should be changed since the actual measurements are limited and the main subject of the paper is methodology.
Response 5: The title has been changed to “Two-dimensional flow field measurement method for sediment-laden flow based on optical flow algorithm”.
Many thanks to the reviewers for their careful reading and valuable comments. We have gained a lot from the revision process based on the suggestions.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
The authors have addressed the concerns and improved the paper. It can be accepted for publication.
Reviewer 3 Report
Tha Authors have addressed my concerns. The new version is suitable for publication.
