A-Priori Calibration of a Structured Light Underwater 3D Sensor
Round 1
Reviewer 1 Report
In this study, a new underwater calibration method is proposed.
This method is new in two ways.
(1) Replacing the radial strain function with a variable principal distance.
(2) To estimate the underwater parameters in advance.
The calibration refinement was achieved by determining a three-dimensional correction function over the measurement volume.
[Comments]
1. This paper deals with underwater calibration for structured light.
In the chapter on experimental methods, it would be easier to understand if there are pictures of the structured light being exposed to the measurement object in water.
2. In order to compare the conventional calibration method with the proposed method, please add the results of the 3d model reconstructed using structured light with each method.
3. We recommend that you compare your results with the calibration methods proposed by other research groups.
Author Response
The following revision steps were performed by the authors
- English language and style were improved by another person
- The description of the methods was partly extended and refined
- Figures 1 and 2 were improved and refined
- For more clarity, new Figures were included, especially for illustration of the results
- One image showing the sensor during fringe projection was inserted (comment 1)
- Table 3 was improved in order to better represent the results of the different calibration methods
- Table 4 was included in order to compare the results to calibration methods proposed by other researchers
- Chromatic aberration was included into the text and mentioned as possible error source
- For more clarity, section 4 was divided into Discussion and Conclusion and partly extended
- Conclusions are derived from the presented results
Reviewer 2 Report
In this paper, a new calibration method for underwater optical stereo scanners is introduced.
It is a well written paper with a complete state of the art. Methods are also clearly explained. Authors state that this method continues former work detailed in references [26, 30] and the comparison included in the results is made against these methods, which include special distortion correction.
In order to validate the proposed results it should be necessary to compare results against the state of the art (not only former work).
Author Response
The following revision steps were performed by the authors
- English language and style were improved by another person
- The description of the methods was partly extended and refined
- Figures 1 and 2 were improved and refined
- For more clarity, new Figures were included, especially for illustration of the results
- One image showing the sensor during fringe projection was inserted (comment 1)
- Table 3 was improved in order to better represent the results of the different calibration methods
- Table 4 was included in order to compare the results to calibration methods proposed by other researchers
- Chromatic aberration was included into the text and mentioned as possible error source
- For more clarity, section 4 was divided into Discussion and Conclusion and partly extended
- Conclusions are derived from the presented results
Reviewer 3 Report
This paper deals with the improvement of a new camera model previously published by almost the same authors (ref [26] in this paper). This is a topic of great interest but the authors should emphasize the real improvements provided by this paper over the camera model of [26].
The beginning of the paper is quite redundant with the author's previous paper and maybe they should more briefly recall their previous results and focus on the real improvement brought in this paper.
The authors should recall that this methodology can only be applied to flat port housing and is essentially useful for a low-cost system. Indeed, high-end vision systems use dome port with corrective lenses (correcting for spherical aberration introduced by underwater refraction) and hence can use a classical pinhole plus radial distortion (this product as an example: http://www.deepsea.com/portfolio-items/apex-seacam/).
Some variables are defined in Fig1 but never used in the paper, like P2' (defined at two different places in the figure), deltaP2. Figure 2 can also be improved for visibility.
The authors use the word "border" for what is called "interface" for refraction.
I find that the experimental part is too small. This is a point of prime interest as only physical experiments can validate the model. Not many details are given on the number of measurements done to evaluate the measurement error, nor any standard deviation for all measurements. It would also be appreciated that the authors talk about the improvements compared to their previous model.
Images have been taken close to the camera, what happens for further objects?
Also, the paper addresses a model that accounts for refraction but it is never said that the refraction index can be highly dependant on the wavelength. In the case of flat port without corrective lenses (scope of the paper) this leads to non-negligible chromatic aberration. Did the authors try to account for it? It should at least be discussed in 3.3 and be given as perspective.
Author Response
The following revision steps were performed by the authors
- English language and style were improved by another person
- The description of the methods was partly extended and refined
- Figures 1 and 2 were improved and refined
- For more clarity, new Figures were included, especially for illustration of the results
- One image showing the sensor during fringe projection was inserted (comment 1)
- Table 3 was improved in order to better represent the results of the different calibration methods
- Table 4 was included in order to compare the results to calibration methods proposed by other researchers
- Chromatic aberration was included into the text and mentioned as possible error source
- For more clarity, section 4 was divided into Discussion and Conclusion and partly extended
- Conclusions are derived from the presented results
-
Additionally, concerning your question
- There is a statement concerning the expectations for sensors with larger measurement distances in section 4. Discussion (lines 368-372)
Round 2
Reviewer 1 Report
The authors have made appropriate corrections to the reviewers' suggestions in the paper. The usefulness of the proposed method is made clear by the results of 3D reconstruction of objects.
Reviewer 3 Report
The authors addressed most of the suggestions I've made so I recommend the paper for publication. I highly encourage the authors to have a look at the effect of chromatic aberration with a flat port (without corrective lenses as in this study) in their future work. It causes color fringes and blurs at the edges of the sensor and even though I'm not aware of any study that had a look at it, my field experiments make me believe it should not be negligible.