Simultaneous Vehicle Localization and Roadside Tree Inventory Using Integrated LiDAR-Inertial-GNSS System

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Abstract line 14: how does expensive equipment result in poor real-time performance? 

P 2, line 62.  Tree occlusion is not the only thing that affects GNSS.  Other physical features--especially urban canopy will also affect PDOP.  Also, how does IMU errors accumulate over time?  IMU errors do contribute to the accuracy of every measurement. 

P3, line 116.  It is low however maybe cite? or say low because of extensive processing requirements..

P. 12, line 479, clearly state that the INS has an integrated GPS or GNSS.  Also, the 1CM +10PPM means what?  You can only get that if you are using a broadcast base station.  Did you use a base station? Also, how can a SPAN-ISA-100 be used for ground truth?  It is just another IMU--and, it is mounted on the same vehicle--thus, systematic vehicle tilt can affect both IMUs.

The detector sounds great--and very efficient.  Major questions: the paper states a real time tree inventory.  What is the context?  Trees in view including the height, width, foilage etc?  Occluded trees also?  The accuracy gives differences.  What about false positives and negatives.  Any features such as power poles or signs considered as trees?

Lastly, how does this enhance safety???  Can the AV leverage the tree information in real time in an effort ensure that it avoids a tree if it runs off the road?  I think not.  Having a fast and efficient way to create an inventory of trees is very useful from a safety standpoint but I question its use in real time.    The authors should clarify how this contributes to enhancing driving safety and "paving the way to more reliable autonomous driving systems"  I AV benefit is very unclear.  AVs try to avoid objects--what does it matter if it is a tree or not.  The extra processing to determine tree inventory info--how can an AV use that in real time?  

Note that I think this paper only needs minor revision however I selected the Reconsider option because the Accept after minor revision assumes that I don't need to see your edits.  I would like to see how you addressed my comments.  

 

Comments on the Quality of English Language

English is good but the sentence structure of the abstract can use improvement.  Also, I prefer use over utilize and position over pose.  

Some of the text "looks familiar"   Especially in the intro.  May very well be different but should run through a program to check.  

Overall, I like the paper.

Author Response

Dear reviewer,

 

Thank you for your prompt review of our manuscript, " Simultaneous Vehicle Localization and Roadside Tree Inventory using Integrated LiDAR-Inertial-GNSS System" (ID: remotesensing- 2488970). We appreciate the time and effort you have dedicated to evaluating the previous version of our paper. Based on the instructions provided in your comment, we uploaded the file of the revised manuscript.

 

Appended to this letter is our response to the comments raised by you. The comments are reproduced and our responses are given directly afterward in a different font.

 

Thank you once again for your valuable input and assistance. We look forward to hearing from you soon.

 

Yours Sincerely,

 

Wenbo Pan

 

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

 

Comment:

1) Abstract line 14: how does expensive equipment result in poor real-time performance?

Reply:

Thank you for your review and comments. Expensive equipment often relies on post-processing algorithms based on ultra-high-precision integrated inertial navigation systems, which require forward and backward propagation of the collected attitude data over the acquisition time period. As a result, such systems cannot operate in real-time.

 

2) P 2, line 62.  Tree occlusion is not the only thing that affects GNSS.  Other physical features--especially urban canopy will also affect PDOP.  Also, how does IMU errors accumulate over time?  IMU errors do contribute to the accuracy of every measurement.

Reply:

Thank you for your careful review and comments. In the revised paper, we have added other occlusion factors that can affect GNSS (please refer to the corresponding modifications). Regarding the issue of IMU error accumulation, based on the characteristics of IMU sensors, they measure the changes in acceleration and angular velocity over a certain time interval and use this data to derive the changes in the position and attitude of the vehicle during that time period. As the measurements contain errors, the calculated changes will also have errors, and these errors will accumulate over time. For a single IMU, without additional absolute observations, these errors cannot be corrected. Therefore, IMUs are subject to the issue of cumulative errors.

3) P3, line 116.  It is low however maybe cite? or say low because of extensive processing requirements.

Reply:

Thank you for your careful review and comments. In the case of MLS systems, they require offline processing to generate a complete point cloud map of the entire area. If the point cloud map is directly processed, it involves a large computational load. In contrast, our proposed method directly extracts line features from the beam information. This approach allows for a simple and rapid selection of potential tree locations from the extensive point cloud without the need to search the entire point cloud. Therefore, compared to MLS-based systems, our system offers higher real-time performance and efficiency.

 

4) P. 12, line 479, clearly state that the INS has an integrated GPS or GNSS.  Also, the 1CM +10PPM means what?  You can only get that if you are using a broadcast base station.  Did you use a base station? Also, how can a SPAN-ISA-100 be used for ground truth?  It is just another IMU--and, it is mounted on the same vehicle--thus, systematic vehicle tilt can affect both IMUs.

Reply:

Thank you for your valuable comments and suggestions.
In the revised paper, the description of the INS system has been clarified. For the high-precision INS system that uses RTK information, we used a base station to obtain high-precision positioning in non-obstructed conditions.

The term "1cm + 10ppm" refers to the accuracy specification of the system. The "1cm" indicates that the basic unit of accuracy is 1 centimeter, and "10ppm" stands for 10 parts per million, which corresponds to a relative error of 0.001%. The specific value of the relative error needs to be multiplied by the distance from the base station.

As for the use of SPAN-ISA-100C for ground truth, it is a combined unit that includes both an IMU and a satellite board. Through post-processing algorithms, it can achieve millimeter-level accuracy, allowing us to evaluate our centimeter-level positioning system. Additionally, both IMUs are calibrated to align the output attitude with the vehicle's reference frame, compensating for the effects of installation errors on both IMUs.

 

5) The detector sounds great--and very efficient.  Major questions: the paper states a real time tree inventory.  What is the context?  Trees in view including the height, width, foilage etc?  Occluded trees also?  The accuracy gives differences.  What about false positives and negatives.  Any features such as power poles or signs considered as trees?

Reply:

The real-time tree inventory selects only unobstructed trees, and through multi-angle observations, it can measure the complete structure of the trees. In the revised manuscript, we have added information about false positives and false negatives, which can be found in Figure 12 and Table 2 of the revised version. Additionally, we have included the visualization of tree detection and mapping results under tree occlusion and multiple rows of trees, as shown in the newly added Figure 11 and its corresponding descriptions. As for features such as power poles or signs, we will judge them based on the shape of the final recognized tree bounding boxes. However, if power poles or signs overlap with the tree canopies, it may be challenging to differentiate them accurately. Thank you for bringing this to our attention, and we hope that our revisions meet your expectations.

6) Lastly, how does this enhance safety???  Can the AV leverage the tree information in real time in an effort ensure that it avoids a tree if it runs off the road?  I think not.  Having a fast and efficient way to create an inventory of trees is very useful from a safety standpoint but I question its use in real time.    The authors should clarify how this contributes to enhancing driving safety and "paving the way to more reliable autonomous driving systems" I AV benefit is very unclear.  AVs try to avoid objects--what does it matter if it is a tree or not.  The extra processing to determine tree inventory info--how can an AV use that in real time? 

Reply:

Thank you for your review and feedback. Autonomous vehicles rely on real-time perception of their surroundings based on point cloud information from LiDAR sensors to ensure that the drivable area does not contain any obstacles. For obstacles such as trees, which are typically static objects, their point cloud information, especially the canopy part, may be sparse at longer distances due to the angles between LiDAR beams. While a single frame detection may not fully recognize the canopy part of trees, constructing detailed maps from multiple frames can provide better detection results. Once the canopy part of a tree is identified, it is compared with the planned drivable path of the vehicle to determine if there is any interference. If interference is detected, the vehicle adjusts its trajectory in advance to avoid the obstacle. This allows trajectory adjustments to be made at a considerable distance from the obstacle, enabling safer path planning.

 

7) English is good but the sentence structure of the abstract can use improvement.  Also, I prefer use over utilize and position over pose. 

Reply:

Thank you for your positive feedback on our paper and for your helpful suggestions to improve it. We have revised the sentence structure in the abstract and have tried to use "use" instead of "utilize" and "position" instead of "pose" throughout the paper.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper introduces an innovative approach that utilizes an integrated LiDAR Inertial Navigation- GNSS to achieve vehicle positioning and roadside tree inventory. The proposed method is effective, but I  have some minor concerns:

 

(1) The tree detection result in Figure 10 is ok, but when the trees distributed densely and occluded, the author should provide some experiments under this condition.

(2) There are no comparison experiments. I think the author may pay attention to verify the effectiveness of the method. Only one experimental result is not convincing enough.

(3) The method can deal with the trees in one row, how to consider multiple rows of trees in the paper? And the author may discuss the affection of different types of trees.

(4) The symbol used in the paper can be improved, e.g. ql_i.

(5) The author's English writing is relatively standardized. However, individual statements need to be modified. For example, Line 240，calculation formula--à formula.

(6) Improper citation of references, such as Line 137-138, “Chen et al. proposed a surface-based mapping method that utilizes 3D point clouds combined with semantic information to improve mapping quality.”there is no citation. There are many such situations below, please check carefully.

(7) Please provide the full name of LOAM.

Comments on the Quality of English Language

The author's English writing is relatively standardized. However, individual statements need to be modified.

Author Response

Dear Reviewer,

Thank you very much for taking the time to review our manuscript titled " Simultaneous Vehicle Localization and Roadside Tree Inventory using Integrated LiDAR-Inertial-GNSS System" (ID: remotesensing- 2488970). We are grateful for your comments and feedback on our research, and we appreciate your positive evaluation of our work.

Appended to this letter is our response to the comments raised by you. The comments are reproduced and our responses are given directly afterward in a different font.

Thank you once again for your valuable input and assistance. We look forward to hearing from you soon.

Best regards,

Wenbo Pan

++++++++++++++++++++++++++++++++++++++++++

Comment:

(1) The tree detection result in Figure 10 is ok, but when the trees distributed densely and occluded, the author should provide some experiments under this condition.

Reply:

Thank you for your review and comments. We have included experiments under densely distributed and occluded tree conditions in the revised manuscript.

(2) There are no comparison experiments. I think the author may pay attention to verify the effectiveness of the method. Only one experimental result is not convincing enough.

Reply:

Thank you for your careful review and comments. We have included experiments, such as information on false positives and false negatives, in the revised manuscript to validate the effectiveness of the proposed method. Please refer to Figure 12 and Table 2 in the revised version for details.

(3) The method can deal with the trees in one row, how to consider multiple rows of trees in the paper? And the author may discuss the affection of different types of trees.

Reply:

Thank you for your careful review and comments. We have included the results of dealing with multiple rows of trees and occlusion in the revised paper. Please refer to Figure 11 for the corresponding effects.

 

(4) The symbol used in the paper can be improved, e.g. ql_i.

Reply:

Thank you for your valuable comments and suggestions. We have carefully reviewed the paper and made appropriate modifications to improve the symbols used.

 

(5) The author's English writing is relatively standardized. However, individual statements need to be modified. For example, Line 240，calculation formula--à formula.

Reply:

Thank you for your feedback on the English writing in the paper. We have reviewed the individual statements and made the necessary modifications.

(6) Improper citation of references, such as Line 137-138, “Chen et al. proposed a surface-based mapping method that utilizes 3D point clouds combined with semantic information to improve mapping quality.”there is no citation. There are many such situations below, please check carefully. 

Reply:

Thank you for bringing this to our attention. We have carefully reviewed the paper and made the necessary corrections to ensure proper citation of all references mentioned in the text.

 

(7) Please provide the full name of LOAM. 

 Reply:

Thank you for your positive feedback on our paper and for your helpful suggestions to improve it. "LOAM" stands for " Lidar Odometry and Mapping."

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Review: ” Simultaneous Vehicle Localization and Roadside Tree Inventory using Integrated LiDAR-Inertial-GNSS System”

 

This paper proposes an innovative approach to simultaneously achieve real-time vehicle positioning and the creation of a roadside tree inventory using LiDAR technology. The authors address the limitations of existing methods that rely on expensive mobile LiDAR systems (MLS) and lack real-time performance.

 

ABSTRACT:

The abstract presents a novel approach that utilizes LiDAR technology for accurate vehicle positioning and efficient roadside tree inventory creation. However, there are a few areas that need improvement. Firstly, the abstract mentions that the system is extensively evaluated on real datasets, but it fails to specify the evaluation metrics used to assess the performance. Including this information would provide a better understanding of the evaluation process. Secondly, while the abstract highlights the limitations of existing methods, it does not explicitly mention how the proposed approach improves upon them. It would be beneficial to clearly state the advancements made by the proposed method. Lastly, the abstract briefly mentions that the proposed method reduces computational load, but it lacks specific details on how this achievement is accomplished. Providing more information on the methodology used to reduce computational load would enhance the abstract's clarity.

 

INTRODUCTION:

The introduction section provides a clearer and more informative overview of the research, addressing the lack of specific details and enhancing the reader's understanding of the proposed system's contributions. It effectively highlights the need for an approach that addresses the challenges of real-time vehicle positioning and tree inventory creation.

 

The related work section offers a comprehensive review of existing methods, their applications, and their limitations, effectively highlighting the need for a new approach. This section adequately sets the stage for the proposed method.

 

Fig 1: GNSS-RKT >> RTK

Section 3.2 line 285: Why do you refer to LiDAR as RADAR? Please be careful about this throughout the manuscript.

 

Fig 12: What is APE? It would be helpful to define this acronym to aid reader understanding.

 

The results section appears to have some limitations. The evaluation of only three tree data points statistically may not provide sufficient evidence for the proposed method's effectiveness. It is recommended that the authors increase the sample size and provide a more comprehensive evaluation to enhance the reliability of the results.

 

 

Overall, while the paper had the potential to be accepted with minor revisions, the limited quality of the results section raises concerns. Therefore, a major revision is recommended to address these issues.

Author Response

Dear reviewer,

Thank you for your prompt review of our manuscript, " Simultaneous Vehicle Localization and Roadside Tree Inventory using Integrated LiDAR-Inertial-GNSS System" (ID: remotesensing- 2488970). We appreciate the time and effort you have dedicated to evaluating the previous version of our paper. Based on the instructions provided in your comment, we uploaded the file of the revised manuscript.

Appended to this letter is our response to the comments raised by you. The comments are reproduced and our responses are given directly afterward in a different font.

Thank you once again for your valuable input and assistance. We look forward to hearing from you soon.

Yours Sincerely,

Wenbo Pan

++++++++++++++++++++++++++++++++++++++++++

Comment:

The author's English writing is relatively standardized. However, individual statements need to be modified.

Comments and Suggestions for Authors

Review:” Simultaneous Vehicle Localization and Roadside Tree Inventory using Integrated LiDAR-Inertial-GNSS System”

This paper proposes an innovative approach to simultaneously achieve real-time vehicle positioning and the creation of a roadside tree inventory using LiDAR technology. The authors address the limitations of existing methods that rely on expensive mobile LiDAR systems (MLS) and lack real-time performance.

 

(1) ABSTRACT:

The abstract presents a novel approach that utilizes LiDAR technology for accurate vehicle positioning and efficient roadside tree inventory creation. However, there are a few areas that need improvement. Firstly, the abstract mentions that the system is extensively evaluated on real datasets, but it fails to specify the evaluation metrics used to assess the performance. Including this information would provide a better understanding of the evaluation process. Secondly, while the abstract highlights the limitations of existing methods, it does not explicitly mention how the proposed approach improves upon them. It would be beneficial to clearly state the advancements made by the proposed method. Lastly, the abstract briefly mentions that the proposed method reduces computational load, but it lacks specific details on how this achievement is accomplished. Providing more information on the methodology used to reduce computational load would enhance the abstract's clarity.

 

Reply:

Thank you for your careful review and comments. The evaluation metrics are divided into two parts: the positioning accuracy of the vehicle during operation and the detection accuracy of trees, which are described in the experiments. The proposed method in this paper first extracts line features from real-time LiDAR point cloud data and projects them onto a global map, which provides an initial estimation of possible tree locations for further tree detection. Additionally, a local map search and detection approach is used to extract trees and overlay them onto the global map. Compared to methods that directly search for trees in the global map, this approach benefits from having an approximate initial tree position, which improves search efficiency. Following your suggestions, we have incorporated relevant content into the abstract. Thank you again for your valuable comments and feedback.

Comment:

(2) INTRODUCTION:

The introduction section provides a clearer and more informative overview of the research, addressing the lack of specific details and enhancing the reader's understanding of the proposed system's contributions. It effectively highlights the need for an approach that addresses the challenges of real-time vehicle positioning and tree inventory creation.

The related work section offers a comprehensive review of existing methods, their applications, and their limitations, effectively highlighting the need for a new approach. This section adequately sets the stage for the proposed method.

Fig 1: GNSS-RKT >> RTK

Section 3.2 line 285: Why do you refer to LiDAR as RADAR? Please be careful about this throughout the manuscript.

Fig 12: What is APE? It would be helpful to define this acronym to aid reader understanding.

Reply:

Thank you for taking the time to review our manuscript and for your positive feedback.

We have made the following changes based on your suggestions:

In Figure 1, we have corrected "GNSS-RKT" to "RTK."

Throughout the manuscript, we have replaced "RADAR" with "LiDAR" after carefully reviewing the paper.

In Figure 12, "APE" stands for "Absolute Positioning Error," which refers to the absolute positional difference between the real-time positioning output of our proposed system and the ground truth positioning.We appreciate your valuable input and have incorporated the necessary modifications accordingly.

 

Comment:

(3) The results section appears to have some limitations. The evaluation of only three tree data points statistically may not provide sufficient evidence for the proposed method's effectiveness. It is recommended that the authors increase the sample size and provide a more comprehensive evaluation to enhance the reliability of the results.

Overall, while the paper had the potential to be accepted with minor revisions, the limited quality of the results section raises concerns. Therefore, a major revision is recommended to address these issues.

Reply:

Thank you for your valuable feedback. We have carefully considered your suggestions and made significant improvements to the results section. Specifically, we have increased the sample size of tree data points in Table 1. Additionally, we have included the visualization of tree detection and mapping results under tree occlusion and multiple rows of trees, as shown in the newly added Figure 11 and its corresponding descriptions.

Furthermore, we have conducted a detailed analysis and description of false positives and false negatives, providing point cloud visualizations as presented in the newly added Figure 12 and Table 2.

We believe that these revisions have substantially enhanced the quality and reliability of the results section. We appreciate your input, and we hope these modifications effectively address your concerns.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The author have revised the paper intensively. However, the authors did not compare the proposed method with existing approaches in this paper. It is recommended that the authors include comparative experiments with state-of-the-art methods.

 

 

 

Author Response

Dear Reviewer,

We sincerely appreciate your thorough review and valuable insights. Your recognition of the substantial revisions made to our manuscript is highly valued.

Regarding your suggestion to compare our proposed method with existing approaches, we would like to provide further clarification. Our algorithm aims to optimize positioning performance based on existing SLAM algorithms while incorporating tree detection capabilities, thereby enabling real-time positioning and the generation of roadside tree inventories. However, it is important to note that existing SLAM algorithms do not inherently possess tree detection capabilities. As a result, our focus in this paper is primarily on conducting a comparative analysis of our algorithm's performance in the context of positioning. We have selected the widely recognized Fast-Lio algorithm as a benchmark for comparing our positioning results. As illustrated in our modified Figure 15, our algorithm and Fast-Lio demonstrate comparable positioning performance. Notably, our approach leverages a front-end odometry based on the error-state Kalman filter (ESKF) and a back-end optimization framework based on factor graphs. The updated poses from the back-end are employed to establish point-to-line residual constraints for the front-end within the local map. Additionally, we have enhanced the weighting of point cloud constraints related to trees and minimized false matches, thereby enhancing the algorithm's robustness. As a result, our algorithm achieves a maximum error that is 6cm smaller than that of Fast-Lio.

Given the unique nature of our integrated algorithm, we did not perform direct comparisons with other tree detection methods. Nevertheless, we acknowledge the significance of comparative experiments, and we intend to explore opportunities for such comparisons in our future work, investigating advanced techniques in the field of tree detection.

Once again, we extend our gratitude for your invaluable feedback, which has greatly contributed to elevating the quality of our work.

Best regards, 

Wenbo Pan

Reviewer 3 Report

Comments and Suggestions for Authors

Acceptable

Author Response

Dear Reviewer,

Thank you for your prompt review of our manuscript. We are pleased to see that our submission has been found "Acceptable." We appreciate your valuable feedback and constructive comments, which have undoubtedly contributed to the improvement of our work.

Best regards,

Wenbo Pan

Round 3

Reviewer 2 Report

Comments and Suggestions for Authors

The author provides a clear explanation,  I have no further comments on this paper.  The paper can be accepted in the present form.

Author Response

Dear Reviewer,

Thank you for your prompt review and valuable feedback on our manuscript titled "Simultaneous Vehicle Localization and Roadside Tree Inventory using Integrated LiDAR-Inertial-GNSS System." We appreciate your time and effort in evaluating our work.

We are pleased to hear that you found the paper's explanation clear, and we are grateful for your positive assessment. Your feedback and acceptance of the paper in its present form are greatly appreciated.

Thank you once again for your time and consideration.

Best regards,

Wenbo Pan

Author Response File: Author Response.pdf