A Novel Method for High-Voltage Bundle Conductor Reconstruction from Airborne LiDAR Data
Round 1
Reviewer 1 Report
This manuscript presents a novel method for classifying bundle conductors for high voltage power lines. The manuscript does a good job of justifying the importance of the work, and contextualizing it with reference to current literature in the field. Overall the manuscript is well-written and organized, though I would encourage a thorough spelling, grammar and syntax check, in addition to the corrections and suggestions I make later in this review. While I support that this method could make a worthwhile addition to the field, and see use, I have concerns over the evaluation of the method, and therefore whether the method is ready to use operationally, as discussed below. I do not think it is absolutely essential for the authors to address these concerns, but it would make it more likely that theirs is the touchstone for this process when it is applied by other practitioners.
In section 4.1, the authors state: “In order to reduce the impact of sample imbalance, this paper adopted a strategy by repeating the training samples to keep all samples’ number is in the same magnitude.” Because of this under-representation of some of the scenarios, does this mean that the testing data were drawn from the training data? Possibly even from data that were duplicated in the training set? If so, we can have the expectation that the model would be over-fitted to these training data, and then poorly evaluated by the testing data. Or was the testing data drawn from dataset II because these data were not used in the classification? Additionally, I assume the accuracy is assessed on a per-point basis. This approach will give a high sample size, but is the classification of each point truly independent, such that they are able to be treated as independent samples in the classification accuracy assessment? It seems like the classification of points are dependent on the fitting of functions to their neighbors, thus making their classifications interdependent. Together, these issues caution against the use of “proved” in the discussion, with regards to the efficacy of the method. If no further, or more robust, validation work is available then I would suggest toning down the vocabulary.
The dependence of the classification on data quality is an extremely important aspect of this proposed method. The authors state that misidentification of single conductors in dataset III is “because of the low point density and low data quality” and visually show an anecdotal example in Figure 10. However, quantitative analysis of the data quality requirements for the success of the method are not conducted. Could the authors consider controlled degradation of the higher quality data to investigate crucial thresholds? The method would be more readily operational for other users if data quality specifications were provided.
The method itself seems sensible, and implemented in pragmatically from a computational standpoint. My only question would be whether the two planes for modeling the conductors should be arranged in a fixed manner (X and Z), or whether they would be effective if one operated orthogonally to the observed local fit of the power line (a localized X and Z plane, as it were). This is just a thought, and perhaps was already considered by the authors.
Additional corrections and suggestions are below. Any entries with a "?" are instances where I was unsure whether a correction was needed, or if I had misinterpreted the intention of the sentence.
Abstract: “power transmission corridor is” to “power transmission corridors are”?
Introduction:
“Electricity, as one of the indispensable energy” to “Electricity, as one of the indispensable energies” or “Electricity, as an indispensable energy”
“in long distance and high capacity, have a great impact” to “over long distances and in high capacity, and as such have a great impact”
“most of high-voltage power lines are extending” to “most high-voltage power lines extend”
“management of smart grid” to “management of smart grids”?
“in recent year” to “in recent years”
“for urgent demand” to “for the urgent demand”
Related Work
Power line Extraction: (“Extraction” to “extraction” for continuity with subsequent subheadings)
“method for low-voltage power line” to “method for low-voltage power lines”
Power line reconstruction:
“This method can well solve the problem of break power lines.” to “This method can solve the problem of broken power lines.”
Power pylon reconstruction:
“Because of power pylon’s complexity structure, data-driven strategy is hardly used alone for its reconstruction.” To “Because of power pylons’ complex structure, data-driven strategies are rarely used alone for their reconstruction.”
“manually modelling library” to “manual modelling library”
“Power pylon was structurally” to “Each power pylon was structurally”
2.1. Power Transmission Corridor Detection and Classification
“classify points in candidate region into” to “classify points in candidate regions into”
2.3. Fitting Residuals-based Bundle Conductors Identification
“Identifying bundle conductors are the premise to extract each sub conductor.” To “Identifying bundle conductors is the premise to extract each sub conductor.”
“is in horizontal arrangement, otherwise it is in vertical arrangement.” To “is in a horizontal arrangement, otherwise it is in a vertical arrangement.”
Projected dichotomy on the XOZ plane
”this step is mianly to divid a whole into two subsets: up and bottom.” To ”this step is mainly to divide a whole into two subsets: up and bottom.”
Projected dichotomy on the XOY plane
"this step is mianly to divid a whole into two subsets: left and right.” To “this step is mainly to divide a whole into two subsets: left and right.
“otherwise, a projected dichotomy on the XOZ plane is adopted to divid it into up and bottom subsets.” To “otherwise, a projected dichotomy on the XOZ plane is adopted to divide it into up and bottom subsets.”
There are other instances of “divid” where “divide” was intended.
2.5.Double-RANSAC-based Model Fitting
“according to the optimal parameter with an aid of the start and end of power line.” To “according to the optimal parameter with the aid of the start and end of the power line.”
“where the line model is instead of the catenary model.” To “where the line model is used instead of the catenary model.”
3. Experimental Data
“Especially, there are two transmission lines stayed closely in dataset III, which containing two kinds of power lines:” stayed does not seem like the correct word here. Maybe should be changed to : “Especially, there are two transmission lines located closely in dataset III, which contain two kinds of power lines:”
4.3. Sub Conductor Extraction
“Although staying very closely, each sub-conductors are still well extracted through this method.” To “Although residing very close together, each sub-conductor is still extracted effectively through this method.”
“a projected dichotomy on the only one plane is adopted according to their arrangement.” To ‘a projected dichotomy on only one plane is adopted according to their arrangement.”
4.4. Power Line Model Fitting
“Typical results are shown in Figure 12, where each fitted power lines are colored in different color.” To “Typical results are shown in Figure 12, where each of the fitted power lines is colored in a different color.”
“It can be seen that the fitted power lines are parallel and disjoint, basically consistent with the distribution of power lines.” To “It can be seen that the fitted power lines are parallel and disjointed, basically consistent with the distribution of power lines.”
“the fitting residual defined in Section 2.3 are calculated. Experiment shows that,” To “the fitting residuals defined in Section 2.3 are calculated. The experiment shows that,”
“For directly comparison, three typical fitted power lines are magnified and overlay with original points in Figure 13,” to “For direct comparison, three typical fitted power lines are magnified and overlaid with original points in Figure 13,”
5. Discussion
“The robustness to noise, which major own to the” to “The robustness to noise, which is mostly owed to the”
“which thanks to the projected dichotomy” to “which is thanks to the projected dichotomy”
5.1 Robustness to Noise
“there are inevitably some noise” to “there is inevitably some noise”
“there are a large continuous noise cross the two power line spans, which will bring additional difficulty for model fitting in high precision.” To “there is continuous noise across the two power line spans, which will bring additional difficulty for model fitting with high precision.”
“This is mainly own to the advance of the RANSAC algorithm:” to “This is mainly owed to the advance of the RANSAC algorithm:”
“and the affect of noise can be reduced when calculating” to “and the effect of noise can be reduced when calculating”
“resulting in the linear model basically accord with the line characteristics of the power line” accord is not the correct word to use here. Perhaps “agreeing” or “acting in accordance with”?
“power line’s catenary characteristics are obvious.” To “the power line’s catenary characteristics are obvious.”
5.2 Robustness to Breakage
“Sparseness and large gaps often occur when a section of power line obscured by vegetation” to “Sparseness and large gaps often occur when a section of power line is obscured by vegetation”
“leading a small number of power lines split into several parts or undetected” to “leading to a small number of power lines being split into several parts or being undetected”
“where both two sub-conductors are disappeared.” To “where both two sub-conductors have disappeared.”
“the power line are divided into several parts” to “the power lines are divided into several parts”
“It also works for breakage with single sub-conductors are disappeared.” To “It also works for breakages where single sub-conductors have disappeared.”
“there is only one sub-conductor of a single span is disappeared.” To “where only one sub-conductor of a single span has disappeared.”
“Although the points in breakage section are wrongly extracted” to “Although the points in the breakage section are wrongly extracted”
6. Conclusions
“For power transmission line inspection, power line is a basis of hazards detection” to “For power transmission line inspection, power lines are the basis of hazard detection” and “basis” could be exchanged for “focus” or similar.
“However, most of existing researches” to “However, most of the existing researches”
“After single power line span is extracted, bundle conductor is firstly identified by analyzing their fitting residuals;” to “After a single power line span is extracted, each bundle conductor is firstly identified by analyzing their fitting residuals;”
“each sub-conductor of bundle conductor is extracted” to “each sub-conductor of each bundle conductor is extracted”
“which is robustness to breakage;” to “and is robust to breakage;”
“well-modeled power lines, related application,” to “well-modeled power lines, related applications,”
“will combine with reconstructed power pylon to add contextual information.” To “will be combined with reconstructed power pylons to add contextual information.”
Author Response
Thank you for considering our manuscript for review and giving the opportunity for revision. We appreciate your time to process this manuscript and give review comments immediately. According to review comments and “Special Instructions to Authors”, we have made some revisions for this manuscript. The details of revisions and responses to reviewers’ comments are listed as following pages. In addition, we polished the manuscript under the help of a native English speaker.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Manuscript: A novel method for high voltage bundle conductor reconstruction from airborne LiDAR data.
It is a well written manuscript, with an adequate structure as a scientific paper demands.
Just only a few minor corrections:
Section 2.1. Why did not you valorate to classify ground / not ground points and them continuos processing with your methodology?
Table 1 Please include density points.
Author Response
Thank you for considering our manuscript for review and giving the opportunity for revision. We appreciate your time to process this manuscript and give review comments immediately. According to review comments and “Special Instructions to Authors”, we have made some revisions for this manuscript. The details of revisions and responses to reviewers’ comments are listed as following pages. In addition, we polished the manuscript under the help of a native English speaker.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper is globally sound, clear and well written. The introduction justifies adequately the topic studied and in particuler the emphasis on bundle conductor (instead of single) reconstruction. The previous work is sufficiently covered.
However, the method is quite heuristic, with nmany thresholds to tune. Also all your geometric criteria should be expressed in 3D instead of working on xy and xz planes.
2.2 the correct term is probably line breaks rather than beakage lines. The descritption of the criteria to reconnect is unclear.
2.4 this step is mianly to divid -> this step is mainly to divide (twice)
Also I think "dichotomy" is not the proper term.
What happens if the two lines are not exatly in the xy or xz plane ? Couldn't you define a more general 3D method not requiring two sub methods ?
2.5: it is unclear how you compute the model parameters from the RANSAC samples, it does not seem straightforward to me given the complexe equation (7).
Author Response
Thank you for considering our manuscript for review and giving the opportunity for revision. We appreciate your time to process this manuscript and give review comments immediately. According to review comments and “Special Instructions to Authors”, we have made some revisions for this manuscript. The details of revisions and responses to reviewers’ comments are listed as following pages. In addition, we polished the manuscript under the help of a native English speaker.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Thank you for the thorough consideration of my comments, and the excellent additions to the manuscript. The addition of the experiment exploring data quality was most welcome. I think the manuscript has reached its potential, and will garner interest from readership.
Author Response
Thank you for considering our manuscript for review and giving the opportunity for minor revision. We appreciate your time to process this manuscript and give review comments immediately.
