Dual-View Stereovision-Guided Automatic Inspection System for Overhead Transmission Line Corridor

Round 1

Reviewer 1 Report

Excellent paper. Well written and documented. Excellent illustrations.

Really only a need for some English Editing and Proofreading and perhaps some changes to headings to make them more specific.

Author Response

We would like to thank the reviewer for the thoughtful and thorough reviews. The red parts are the revised contents in manuscript. Please see the attachment.  Hopefully we can deal with all of your concerns. 

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper presents a dual-view stereo vision approach for transmission line corridor evaluation by detecting the power lines and the ground clearance.

I read the paper with great interest however there are points that should be improved:

Page 2. Line 55: “… replace artificial detection with intelligent means”, please clarify the difference between artificial detection and intelligent means?

Page 2. Line 64 and 67: Please improve both sentences: “On the one hand ...” and “On the other hand ...”.

Page 3. Line 85: “ … clearance measurement of in ..” improve the sentences

Page 3. Line 87: “ Inefficiency of offline ...” improve the sentences by removing “of” from sentences

Page 4. Line 144: The authors refer to a visual positioning system [20]. Is this method being used to estimate the UAV position? Or are you using RTK GPS + visual features? Please explain correctly the method used to estimate the position and attitude of the UAV, because it’s not clear what is the level of accuracy that you have in the estimated pose (position and attitude). Explain what the flight state is? And for me it’s not clear how to use visual features went you are flying 20 – 30 meters from the ground. What camera resolution are you using to detect features in the ground. What is the UAV's impact on the power lines' estimate position? How the algorithm deals to the fact that the power lines can be detected in different 3D position (errors caused by the accuracy of the estimates pose)

Pag 4. line 147. What is the upward angle. The authors detail the downward angle but doesn’t explain the upward angle.

Pag 4. Line 157 – 160. What is the embedded platform? Improve the sentence. “… UAV could be sent to the embedded platform and the embedded platform can control the UAV”. Do you mean embedded PC NVIDIA?

Pag 4. Line 168 – 172. Please explain how the authors synchronize the upper and downward cameras with the IMU? Where is the estimated UAV position? Why only the IMU? Please improve the sentence with more details about the previous questions.

Pag 5. Line 173 – 178. Please clarify the following questions: The dataset for the classification network was performed by acquiring 10Km of aerial images? Please explain more clearly what the dataset was for training? Image resolution, number of images (all of them were used for training? Or part was for training and the other for inference?)

Pag 5. In table 1, please include the estimated position/attitude accuracy.

Pag 5. The authors refer to β as default was equal to 0.5. In the case of the power lines, what was the weight edge supervision? Please improve this sentence about the value that was used. Did try to tuning this value with an iterative method to estimate β?

Pag 8. In line 269 you refer to the declination angle obtained from the magnetic compass from UAV DJI. Please describe what was the impact of flying under the power line, which is greatly affected by the magnetic field generated by high voltage cables. Please include a graphic with the results of the magnetic compass UAV during the power line inspection. What was the impact of the magnetic field on the 3D power line estimated position and orientation?

Pag 10. Line 334. Please explain the channel size of 32.

Pag 11. Please move figure 6 to page 11, it’s to far from the description in page 10 and 11.

Pag 15. Improve figure 9. From figure 9, It’s not possible to understand the 3D position of the lines. Where is the Ground Truth of the power lines, to confirm the sentence: “power lines are relatively accurate” (Line 492), and the distance between UAV and power line (Line 488 and 490)?

- Please include in the results the 3D power line reconstruction of several frames and not only from one image.

- Please include the processing time of each frame during the power line detection with the NVIDIA embedded PC

Pag 18. What was the LiDAR used in the UAV? Can you detail the accuracy of LiDAR sensor, and also, please improve the description of the outliers that you present in the LiDAR graph in figure 14. From my point of view, I don’t understand the outliers in the LiDAR data, I was expecting to have more outliers in the image-based clearance distance measure due to feature data association in stereo. 

Author Response

We would like to thank the reviewer for the thoughtful and thorough reviews. The red parts are the revised contents in manuscript. Please see the attachment.  Hopefully we can deal with all of your concerns. 

Author Response File: Author Response.pdf

Reviewer 3 Report

This manuscript developed the dual-view stereo vision guided automatic inspection system for overhead transmission line corridor. The authors put forward the architecture of the proposed dual-view stereo vision guided automatic transmission line corridor inspection system. Then the algorithms, including power line detection, power line 3D reconstruction, automatic flight strategy formulation, ground object classification, ground clearance measurement and detection, are proposed. The corresponding datasets and practical 220kV transmission line corridors are both demonstrated.

Overall, the manuscript is well organized and its presentation is good. The results show that the proposed method is effective.

However, some minor issues still need to be improved:

(1) The main contribution of other references about the dual-view stereo vision should also be mentioned briefly. It is suggested to summary the contributions briefly and list them in Introduction Section.

(2) In Section 3.11 and 3.2.2, it is suggested to change the title number 1. , 2.,…… to 1), 2)，…….

Author Response

We would like to thank the reviewer for the thoughtful and thorough reviews. The red parts are the revised contents in manuscript. Please see the attachment.  Hopefully we can deal with all of your concerns. 

Author Response File: Author Response.pdf